Does the Spatial Proximity Between Legislators and Voters Affect Voting Decisions in U.S. House Elections?∗ Chris Tausanovitch† Department of Political Science UCLA Christopher Warshaw‡ Department of Political Science Massachusetts Institute of Technology February 2015 Abstract: The theory of spatial voting has dominated scholarship on voting and elections. The spatial voting theory’s most important implication is that candidates’ ideological positions should influence voters’ decisions at the ballot box. A number of recent studies find that the spatial voting theory’s predictions are borne out in U.S. House election. This suggests that legislators should be highly responsive to the preferences of their constituents. However, there is only a very modest relationship between district preferences and legislators’ roll call votes. We explain this puzzle using a new dataset of over 75,000 voters in the 2006-2012 congressional elections. In contrast to previous studies, we find that the spatial positions of legislators have a negligible impact on citizens’ voting behavior. Instead, our results suggest that citizens cast their votes based on proximity to parties rather than individual legislators. Our findings help explain the polarization and lack of responsiveness in the contemporary Congress. ∗ We are grateful to Devin Caughey, Robert Erikson, Anthony Fowler, Justin Grimmer, Seth Hill, Stephen Jessee, Jeffrey B. Lewis, Howard Rosenthal, and seminar participants at MIT’s American Politics Conference, Princeton University, the University of California-Berkeley, UCLA, and UCSD for feedback on previous versions of this manuscript. We also thank Stephen Ansolabehere and Phil Jones for generously sharing data on constituents’ perceptions of legislator positions, and Adam Bonica for sharing data on campaign contributions. This paper was previously circulated under the name “Electoral Accountability and Representation in the U.S. House: 2004-2012.” † Assistant Professor, Department of Political Science, UCLA, [email protected]. ‡ Assistant Professor, Department of Political Science, Massachusetts Institute of Technology, [email protected]. 1 The theory of spatial or proximity voting (Black, 1948; Downs, 1957; Enelow and Hinich, 1984) is one of the central ideas in scholarship on voting and elections. The spatial voting theory’s most important prediction is that the ideological positions of candidates and parties should influence voters’ decisions at the ballot box. This electoral connection is said to help ensure that legislators are responsive to the views of their constituents (Mayhew, 1974). In recent years, a number of prominent papers have examined the implications of the theory that citizens vote spatially based on candidate positions (e.g., Ansolabehere and Jones, 2010; Jessee, 2009, 2012; Joesten and Stone, 2014; Nyhan et al., 2012; Shor and Rogowski, 2010; Simas, 2013). These studies conclude that voting behavior in congressional elections is associated with the spatial proximity between the policy views of legislators and voters. However, this result leads to a puzzle. If citizens are voting spatially for individual candidates, then legislators should have a strong incentive to cast votes that represent the median voter in their districts (Black, 1948). As a result, there should be a very tight relationship between the views of constituents in each district and the roll call voting behavior of their representative. But a large body of work shows that legislators do not converge on the position of the median voter (Ansolabehere, Snyder Jr, and Stewart III, 2001; Levitt, 1996). In addition, there is only a modest relationship between district preferences and legislators’ roll call voting behavior (Clinton, 2006; Lee, Moretti, and Butler, 2004; Tausanovitch and Warshaw, 2013). Spatial voting with regard to parties has quite different implications from spatial voting with regard to individual legislators. This theory suggests that voters use party labels to figure out the policy positions of elected officials (Conover and Feldman, 1989; Popkin, 1991; Snyder and Ting, 2002). This enables voters to make a Downsian calculation about which candidate is better from a policy perspective. They can cast their ballots rationally without knowing much about the positions of individual candidates by simply voting for the candidate whose party reflects their policy views (Sniderman and Stiglitz, 2012). We show that the implications of candidate and party-based theories of proximity voting 1 can be deduced theoretically and differentiated empirically. They can also be compared to affective theories of partisan identification. In this paper, we use a novel survey-based dataset with the ideal points1 of over 75,000 voters in 1,100 electoral contests between 2006 and 2012 to adjudicate between candidate and party-based theories of proximity voting. We test the implications of each theory using a non-parametric model, which does not rely on functional form assumptions. In order to make clear the connection between the theory and our empirical analysis, we also use a parametric model that is derived explicitly from the theory. Both approaches yield similar results. In contrast to previous studies, we find that the roll call positions of legislators (as measured by DW-Nominate scores) have almost no impact on the voting decisions of most citizens.2 Instead, we find evidence that most citizens cast their votes based on their spatial proximity with the parties: liberal voters almost always support Democrats and conservative voters almost always support Republicans. Additional analysis indicates that one plausible explanation for this finding is that voters appear to be only vaguely aware of the ideal point of their individual legislator. Only high information voters seem to take legislators’ positions into account in congressional elections, and these effects are small. In the last section of the paper, we use our large sample of voters to show that legislators’ roll call positions have only a very modest impact on their vote-shares in each contest. This suggests that the ideological characteristics of individual incumbents have a negligible influence on electoral outcomes, net of party. Our results have broad implications for representation and democratic accountability in the United States. Most importantly, our results show that incumbent legislators face few electoral consequences for unrepresentative positions (cf. Wilkins, 2012). Whatever leads 1 An “ideal point” in this context is a measure of an individuals’ ideal policy position in a liberalconservative spectrum. 2 Our study focuses on measures of legislative positions that are based on roll call voting. Our focus on roll call votes follows from our interest in representation. It may be the case that voters are sophisticated enough to differentiate legislators on some basis, such as their publicly stated positions or some other position-taking behavior, and vote spatially on this basis (see for instance, Grimmer, 2013; Cormack, 2013; Henderson, 2013). However, if voters do not condition their voting behavior on roll calls, then legislators may vote as they please. Voter behavior could be quite sophisticated, but policy representation would not be achieved. 2 more extreme candidates to lose more often (Canes-Wrone, Brady, and Cogan, 2002; Hall, 2014) probably has more to do with the correlates of extremity than policy positions per se. Legislators do not appear to be bound by the policy views of their particular constituents, as long as they can claim to belong to the political party their constituents prefer.3 This helps explain the broad patterns of divergence between the parties (Poole and Rosenthal, 2000; Lee, Moretti, and Butler, 2004), and very weak responsiveness to the preferences of constituents (e.g., Clinton, 2006), that we observe in the contemporary Congress. Theories of Proximity and Party Voting In An Economic Theory of Democracy, Downs (1957) argues that vote choices are a function of the spatial proximity between the ideal points of voters and parties. This spatial voting model was easily extended to the proposition that citizens should be more likely to vote for legislators and other elected officials that share their ideological preferences, spawning a long literature in spatial voting theory (e.g., Enelow and Hinich, 1984). In a related line of research, called directional voting theory, scholars argue that voters support candidates whose spatial positions are on the same side of the political spectrum as their own positions (Rabinowitz and Macdonald, 1989; Macdonald, Listhaug, and Rabinowitz, 1991).4 The common element of both of these theories is that they imply that individual candidates’ positions should influence citizens’ voting decisions. For many years, there was “surprisingly little direct evidence supporting [the spatial voting model’s] main assumptions–namely, that voters have preferences over the issues before the legislature, hold beliefs about their representatives’ decisions on those questions, and use that information in deciding whether to support the legislator” (Ansolabehere and Jones, 2010, 583). However, the explosion of large-sample surveys in recent years has facilitated a 3 Note that our findings do not suggest that legislative candidates can take any position at all. For instance, ideologically extreme candidates that take positions far outside the bounds of their party’s platform may still face electoral consequences (Hall, 2014). 4 Tomz and Van Houweling (2008) use survey experiments to adjudicate between theories of spatial and directional voting. They find that spatial voting is four times more common than directional voting. 3 renaissance in scholarship on voter behavior in congressional elections. Over the past few years, a number of prominent studies have found support for the spatial voting model. Jessee (2009, 2012) uses a novel survey design to estimate the ideology of voters and the major party presidential candidates in 2004 and 2008 on the same spatial scale. Jessee finds that liberals are more likely to support the Democratic presidential candidate and conservatives are more likely to support the Republican presidential candidate. Shor and Rogowski (2010) extend Jessee (2009)’s framework to the level of congressional elections. They find evidence that vote choice in congressional elections is associated with the voters’ spatial proximity with congressional candidates. In two similar studies, Ansolabehere and Jones (2010) and Jones (2011) evaluate the impact of incumbents’ issue positions on citizens’ voting behavior. They find that respondents are more likely to vote for incumbents that share their issue positions. Simas (2013) finds evidence that voters consider their ideological proximity to congressional candidates and “punish candidates who take positions that are too far out of line.” Joesten and Stone (2014) use district experts to place candidates and survey respondents onto the same ideological scale. They conclude that “proximity voting is common” among voters in congressional elections. Finally, Nyhan et al. (2012) find that legislators’ positions on health care reform and other salient votes affected voters’ decisions in the 2010 congressional elections. These papers revolutionized the study of spatial voting by bringing it into the arena of empirical study. However, an important limitation of their findings is that they fail to distinguish candidate-centered versus party-centered accounts of spatial voting. Political scientists often forget that early spatial voting theorists such as Downs (1957) and Hotelling (1929) focused on parties rather than individual candidates and legislators. In these theories, voters took proximity into account, but only of the national political parties. The partyfocused perspective is not unique to spatial voting theory either. The traditional theory of partisan identification holds that party is an enduring attachment, much like religion, with primarily affective roots (Campbell et al., 1960; Lewis-Beck et al., 2008). Party attachments 4 are formed early in life and are very stable thereafter. Policy views and vote choices are both determined by party identification, which is the dominant force in voters’ political lives. Policy determinations are more or less epiphenomena of party, and voters are relatively ignorant about policy and legislative activities. Green, Palmquist, and Schickler (2004) ground the importance of party identification in a concept of social identification. People identify with parties because they think of themselves as being similar to other people in their party. One element of early theories of party identification that has earned more emphasis over time is the notion of party as a cue or heuristic. More recent work argues that party labels help voters figure out the policy positions of elected officials (Conover and Feldman, 1989; Popkin, 1991; Snyder and Ting, 2002). This enables voters to make a Downsian calculation about which candidate is better from a policy perspective. They can cast their ballots rationally without knowing much about the positions of individual candidates by simply voting for the candidate whose party reflects their general policy views, rather than making detailed evaluations of particular candidates (Sniderman and Stiglitz, 2012). Indeed, the results of several of the spatial voting studies discussed earlier actually subtly support this line of reasoning. For instance, the descriptive tables provided by Joesten and Stone (2014, 750) and Simas (2013, 712) examine how voters cast their ballots when their spatial positions are closer to the other party’s candidate. These tables indicate that voters continue to cast their ballots for their own party between 70-89% of the time when their spatial positions are closer to the other party’s candidate.5 Both the affective theory of party identification, as well as the notion of party identification as a heuristic, imply that voters with more extreme policy views should be more reliable 5 Table 5 of Joesten and Stone (2014, 750) shows that 71% of Republicans and 89% of Democrats in 2006 House elections voted for their party’s candidate when their spatial positions were closer to the opposing party’s candidate. Likewise, column 3 of Table 1 in Simas (2013, 712) shows that citizens voted for their party’s candidate when their spatial positions were closer to the opposing party’s candidate roughly 80% of the time. Put differently, these studies find that between 70-89% of partisan voters whose spatial positions are closer to the opposing party’s candidate violate the prediction of candidate-centered spatial voting theory when they cast their vote. 5 voters for their respective parties. From an affective perspective, stronger partisans are more likely to have “correct” policy views, as passed down from their respective parties. From a heuristic perspective, party is a more effective heuristic for those with more extreme policy views. Therefore, both perspectives lead us to expect that more extreme ideologues will also be more loyal partisans. On average, liberals and conservatives will vote for the legislator of their party at higher rates than moderates. Distinguishing Alternative Theories of Voting How can we distinguish candidate and party-based theories of proximity voting? Theories of proximity voting based on candidate positions have a simple premise: unless utility functions are very flat or there is very little variation in legislator positions, then citizens’ votes must depend strongly on legislator positions. If they do not, then either citizens are not voting spatially/directionally or the role that these considerations play in their decision is small.6 Theories of party voting hold that there should be large cross-party differences in voting but that these differences should not depend on the position of the particular legislator. Further, if party attachment is primarily affective, voters’ policy positions should be largely irrelevant. All of the previous studies, however, fail to distinguish candidate-centered versus partycentered accounts of spatial voting. For instance, Jessee (2009) shows that liberals are more likely to vote Democrat and conservative are more likely to vote Republican, but fails to demonstrate that spatial proximity is the feature that determines this relationship. Similarly, many scholars use spatial proximity as their key independent variable without accounting for the direct effect of voter ideology, thereby confounding these explanations (e.g., Joesten and Stone, 2014; Shor and Rogowski, 2010; Simas, 2013). Finally, while many existing papers control for the effect of party identification, they fail to account for party loyalty. The literature indicates that partisanship is likely to interact with policy positions: more liberal 6 Of course, it is always possible that voters are capable of using a proximity voting rule, but that the use of such voting rules is not prevalent enough to matter. It is also possible that they use a proximity voting rule, but with respect to an orthogonal space or notion of position. 6 Democrats will tend to be more loyal Democrats, and more conservative Republicans will tend to be more loyal Republicans. Rather than accounting for this possibility, past work has tended to attribute all of the effect of ideology to the spatial proximity between voters and candidates. In the next section, we derive a simple model of voting behavior in congressional elections, and then discuss how we translate this model into an empirical model of voting behavior. A Model of Voting Behavior in Congressional Elections Consider a voter i with ideal point xi , and election j where the incumbent has ideal point rj and the challenger has ideal point rj . According to the candidate-centered notion of spatial voting, voters should vote, with error, for the candidate who has an ideal point “closer” to their own. Although voters cast votes with error, they are more likely to vote for their favored candidate as the spatial advantage of their favored candidate grows. Equations 1 through 3 spell this out mathematically. P (yij = 1) = f (d(xi , rj ) − d(xi , cj )) (1) = Λ((xi − rj )2 − (xi − cj )2 ) (2) = Λ(−2xi rj + rj2 + 2xi cj − c2j ) (3) This very simple algebra shows that the spatial voting logic leads directly to a logistic regression model if we assume a logistic link function and quadratic utility. We exclude challengers from our analysis and focus on the effects of xi rj and xi . We will return to the consequences of excluding challengers shortly. For now, note that we can rewrite this as a more general logistic regression equation, where c2j is subsumed in the constant term and cj is assumed to be orthogonal to the other variables. 7 P (yij = 1) = Λ(β0 + β1 xi rj + β2 xi + β3 rj2 ) (4) The most important prediction from this model is that β1 will be significantly different from zero. If β1 is substantively and statistically significant, then the probability of supporting the incumbent depends on the distance between the incumbent’s ideal point and the voter’s ideal point. H1 (Candidate-Centered Spatial Voting): The distance between the voter and the incumbent’s ideal point should have a substantive effect on the probability of supporting the incumbent. Distance is captured by the interaction between the incumbents’ ideal point and the voter’s ideal point. In terms of the model: β1 6= 0. What does H1 imply in terms of the effect that different incumbent positions should have on the likelihood that different constituents will support the incumbent? Empirically, we will not be able to say whether particular voters are to the left or right of particular legislators.7 However, we can safely assume that Republicans in Congress are in the most conservative third of the electorate and that Democrats in Congress are in the most liberal third. We can classify voters as either liberals, moderates or conservatives based on their policy views, with a third of the public in each group. If a Democratic member of Congress becomes more moderate, then their conservative and moderate constituents should be more likely to support them. Whether liberals support them more or less depends on exactly how liberal the member is and how liberal those voters are – the prediction is ambiguous. If a Republican member of Congress becomes more moderate, their moderate and liberal constituents should be more likely to support them. In this case, the prediction for conservatives is ambiguous. 7 The task of estimating voter positions in the space of legislators is a difficult one. It requires assuming equivalence between some set of behaviors that are driven by policy position: for instance, that casting roll call votes in a legislature can be considered equivalent to answering survey questions about roll call votes, or that campaign contributions are given to more spatially proximate candidates. Lewis and Tausanovitch (2013) find that existing attempts to jointly scale voters and legislators in the same space fall short. 8 These theoretical predictions are displayed in Table 1. Table 1: Theoretical Effect of Incumbent Moderating their Ideal Point on Pr(Vote for Incumbent) Voter Ideology: Theoretical Prediction Democratic Incumbent Liberal Moderate Conservative ambiguous + + Republican Incumbent Liberal Moderate Conservative + + ambiguous The predictions yielded by a party-centered notion of spatial voting are different than the predictions yielded by a candidate-centered model. Equations 5 through 7 show the algebra for a party-centered notion of spatial voting, where p1 is the ideal point of the incumbent’s party and p2 is the ideal point of the challenger’s party. P (yij = 1) = f (d(xi , p1 ) − d(xi , p2 )) (5) = Λ((xi − p1 )2 − (xi − p2 )2 ) (6) = Λ(−2xi p1 + p21 + 2xi p2 − p22 ) (7) The most important difference between equations 5 through 7 and equations 1 through 3 is that in the former, the position of the two parties is a constant for all voters. As a result, equation 7 can be simplified to a very parsimonious logistic regression. p21 and p22 are subsumed by the constant, and xi is multiplied by the unknown constant −2p1 + 2p2 . We can capture this with the following equation: P (yij = 1) = Λ(β0 + β2 xi ) (8) The party-centered notion of spatial voting yields a single important prediction. As long as the two parties do not adopt the precise same location, the position of the voter should affect the voter’s likelihood of voting for the incumbent. 9 H2 (Party-Centered Spatial Voting): The voter’s ideal point should have a substantive effect on the probability of supporting the incumbent. In terms of the model: β2 6= 0. Note that H2 and H1 are not the same. H1 says that the distance between the legislator and the voter should matter. But according to H2, only the position of the voter matters. Because voters all experience the same party locations, incumbent positions are immaterial. To set these hypothesis against each other, we formulate a logistic regression specification that includes both, as below. We control for contest-specific random effects, the direct effect of the incumbent legislator’s ideal point, and we include quadratic terms for both the incumbent and the voter. P (yij = 1) = Λ(β0 + β1 xi rj + β2 xi + β3 rj + β4 rj2 + β5 x2i ) (9) In order to capture the classic hypothesis that voting is based on an affective party attachment, we will estimate equation 9 within-party for the six possible matches between legislator and voter party. Legislators are either Democrats or Republicans, and voters are either Democrats, Republicans, or Independents. The within-party specification ensures that the incumbent party is the same for all voters, and allows the intercept to account for traditional party voting. Unlike some other recent studies (e.g., Joesten and Stone, 2014; Shor and Rogowski, 2010), we have chosen in this paper to focus explicitly on incumbent positions and eschew any attempt to estimate the policy positions of challengers. As a result, we cannot rule out the possibility that challengers position themselves in such a way as to nullify the effects of incumbent positioning. We find such a possibility highly unlikely, because if true, it has a counterintuitive implication: that voters reward candidates for moderation, but candidates respond to extreme opponents by diverging further.8 We elaborate on this argument formally 8 We acknowledge that there are theories that rationalize candidate divergence in the face of candidatespecific spatial voting by voters. The most famous example is from Calvert (1985) and Wittman (1983). 10 in Appendix A. A substantial benefit of excluding challengers from the analysis is that it enables us to avoid some stubborn methodological problems. We are interested in whether voters constrain the roll call voting behavior of their representatives. For challengers who fail to unseat the incumbent, roll call voting itself is a counterfactual, and estimating ideal points is difficult. Survey Data In order to examine whether voters hold members of the House accountable for their roll call behavior, we need surveys with information on voters’ policy preferences and vote choices in congressional elections. To this end, we pool together the 2006, 2008, 2010, and 2012 Cooperative Congressional Election Surveys (Table 2).9 Each of these surveys asked between 14 and 32 policy questions to 30,000-55,000 Americans.10 This enables us to jointly scale each respondents’ ideal point using an approach similar to the one taken by Tausanovitch and Warshaw (2013). Moreover, each of these surveys include self-reported vote choice in congressional elections as well as the party and congressional district of each respondent. To estimate voters’ ideological positions, we assume that all survey respondents have a quadratic utility function with normal errors (Clinton, Jackman, and Rivers, 2004). Each item presents individual with a choice between a “Yes” position and a “No” position. We use the two-parameter IRT model introduced to political science by Clinton, Jackman, and Rivers (2004), which characterizes each response yij ∈ {0, 1} as a function of subject i’s latent ability (xi ), the difficulty (αj ) and discrimination (βj ) of item j, and an error term (eij ), where Pr[yij = 1] = Φ(βj xi − αj ) (10) 9 We use the pre-election survey for respondents’ policy questions, and the post-election panel for their vote choice. Note that each of these surveys name both the challenger and incumbent candidates in each contest. 10 Appendix D shows all of the questions that we use in our model. We dichotomize each of the survey questions for our scaling model. 11 where Φ is the standard normal CDF. βj is referred to as the “discrimination” parameter because it captures the degree to which the latent trait affects the probability of a yes answer. The “cut point” is the value of αj / βj at which the probabilities of answering yes or no to a question are 50-50. We assume a one-dimensional policy space because a twodimensional model shows little improvement in terms of model fit. The ideal point, x, for individual i signifies the “liberalness” or “conservativeness” of that individual. We orient our values so that lower values are associated with politically left preferences and higher values with politically right preferences. We approximate the joint posterior density of the model parameters using a Markov chain Monte Carlo (MCMC) method. To validate our estimates, Table 3 shows the strong relationship between symbolic ideology and our scaled measure of citizens’ ideal points. Table 2: Data Sources for Voters Survey CCES 2006 CCES 2008 CCES 2010 CCES 2012 Total Total Sample Size Post-Election Sample of Respondents that Indicated a Vote Choice 35,919 22,046 32,800 17,895 55,488 32,004 54,535 29,999 178,742 101,944 Post-Election Sample in Contested Race 16,842 14,584 26,922 22,755 81,104 In all, we have information on the ideal points of 178,000 survey respondents (Table 1). We have information on self-reported vote choices for approximately 80,000 of these respondents in contested races with incumbents running for re-election. Our estimates of legislators’ party and roll call positions come from Poole and Rosenthal’s DW-NOMINATE scores (Poole and Rosenthal, 2000). Data on legislators’ incumbency status are derived from Gary Jacobsen’s data on congressional elections and research by the authors. Finally, we classify “leaners” (those who identify themselves as Independents but say they lean towards one party or the other) as partisans for all of the substantive analyses that follow. This choice does not significantly affect the results. 12 Table 3: Symbolic Ideology and Citizen Ideal Points Symbolic Ideology Mean Ideal Point Very Liberal -1.30 Liberal -1.03 Moderate -0.31 Conserative .83 Very Conservative 1.34 Visualizing Legislators’ Positions & Constituent Voting Do voters hold members of the U.S. House accountable for their roll call positions? As a first-cut, we use a simple non-parametric framework. We separate our data into voterlegislator pairs, one for each combination of voter and legislator partisanship (DemocraticDemocrat, Independent-Democrat, Republican-Democrat and so on).11 For each pair, we separate voters into five groups based on their ideology. In each of these categories, we graph a loess curve of the percent voting for the incumbent across the range of incumbents’ ideal points (DW-Nominate scores).12 This is similar to simply graphing a point for each category of voter ideology and each category of legislator ideology. Each of the panels in Figure 1 subset our data based on respondent and legislator party identification. The second row shows Democratic voters, the third row shows Independent voters, and the fourth row shows Republican voters. The theory of proximity voting based on the spatial distance between voters and candidates has a simple prediction: liberals should be more likely to vote for more liberal legislators and conservatives should be more likely to vote for more conservative legislators. Moderates should be more likely to vote for more moderate legislators. In other words, each of our lines should have a slope representing the sensitivity of the vote choice to legislator opinion. In the case of directional voting, the slope should be even steeper: as legislators go from the 11 All of the analyses that follow focus on contested races. But the results are the same if we analyze all races. 12 All of the curves are weighted using respondents’ survey weights. 13 “wrong” side of the “neutral point” to the “right” one, the voters should switch en masse from voting against them to for them. If the neutral point is between the two parties, then voters should always vote for the party on their side of the neutral point (all lines should be at 100% or 0%), and voting should be completely determined by ideology, not party. Looking first at the graphs for Democratic voters (top row), the most salient pattern is that all of the curves are generally flat. Indeed, over 98% of liberal voters support Democratic incumbents, and upwards of 90% oppose Republican incumbents, virtually regardless of the legislators’ positions. Liberal Republicans are the only group in these graphs that appears to be responsive to the position of legislators. For instance, there is roughly a 40% probability that Liberal Republicans support liberal Democratic legislators, and a 75% chance that they support the most moderate Democratic incumbents. However, we should keep in mind that there are few liberal Republicans. This group represents only 2.3% of all Republicans in our sample. Next, we examine the graphs for Independent voters (2nd row). Several recent, prominent papers suggest that independents are highly responsive to legislators’ roll call positions (Jessee, 2009, 2012; Shor and Rogowski, 2010). However, Figure 1 indicates there are only very modest associations between the vote choices of independents and legislators’ roll call positions in our data. Finally, the bottom row of Figure 1 shows the association between legislators’ positions and constituents’ decisions on election day for Republican voters. The plot shows that Republican voters are slightly more likely to support moderate Democratic incumbents. However, there is no consistent association between the probability that Republican voters support Republican incumbents and the incumbents’ ideology. Over 97% of Republicans support Republican incumbents, and over 90% oppose Democratic incumbents, virtually regardless of the legislators’ positions. Looking across the plots, a remarkable feature of these results is the strength of both respondents’ party and ideology as a predictor of vote choice. The effect of ideology is captured by the differences in the levels of the lines within each panel, and the effect of party is cap- 14 tured by the differences in the lines going down the plots in each column of graphs. A cursory glance shows that these effects are substantial. Even moderate Democrats overwhelmingly support Democratic incumbents, and moderate Republicans overwhelmingly support Republican incumbents. These individuals have the same ideology and differ only in party identification. However, individual ideology also has a substantial independent effect. For instance, Democratic voters who are very conservative support Democratic incumbents less than 60% of the time. Likewise, Republican voters who are very liberal support Republican incumbents less than 65% of the time. Overall, Figure 1 indicates that the direct effects of party and voter ideology dwarf the effect of legislator position. The difference in the levels of the lines within and across panels is vastly greater than the difference in the slope of the lines. The fact that individual ideology has a strong independent effect on vote choice is not evidence for the proximity model, because it contains no notion of distance. However, it does provide evidence that party attachment may not be purely affective. If voters’ policy positions drive the extent to which they reliably support their party, then the spatial distance between the voter and the party is a sensible explanation. It may be the case that voters think spatially with reference to parties, but not candidates. 15 Democratic Voter/Democratic Leg. Pr(Vote for Incumbent) Pr(Vote for Incumbent) 1.0 0.8 0.6 0.4 0.2 0.0 −0.6 −0.5 −0.4 −0.3 −0.2 1.0 Democratic Voter/Republican Leg. 0.8 0.6 0.4 0.2 0.0 −0.1 0.4 1.0 Independent Voter/Democratic Leg. 0.8 0.6 0.4 0.2 0.0 −0.6 −0.5 −0.4 −0.3 −0.2 1.0 Pr(Vote for Incumbent) Pr(Vote for Incumbent) 0.4 0.2 0.0 −0.3 −0.2 0.9 0.4 0.2 0.0 0.5 0.6 0.7 0.8 0.9 Incumbent Position 0.6 −0.4 0.8 0.6 0.4 0.8 −0.5 0.7 0.8 −0.1 Republican Voter/Democratic Leg. −0.6 0.6 Independent Voter/Republican Leg. Incumbent Position 1.0 0.5 Incumbent Position Pr(Vote for Incumbent) Pr(Vote for Incumbent) Incumbent Position −0.1 1.0 Republican Voter/Republican Leg. 0.8 0.6 0.4 0.2 0.0 0.4 Incumbent Position 0.5 0.6 0.7 0.8 0.9 Incumbent Position Figure 1: Spatial Voting in the U.S. House: 2006-2012 – This graph shows non-parametric loess curves of the relationship between legislators’ DW-Nominate scores and the probability that respondents at various ideological levels support them on election-day. The y-axis is the probability of voting for the incumbent and the x-axis is the incumbent’s DW-NOMINATE score. Each line is a loess plot for a set of voters with a given quintile of ideology. The most liberal quintile is shown using long-dashes, the next most liberal quintile is shown using dashes, moderate voters are shown using a solid line, more conservative voters are shown using a dot-dash line, and the most conservative voters are shown using a dotted line. The top row of the graph shows loess fits for Democratic respondents, the second row is for Independent respondents, and the last row is for Republican respondents. The first column is for Democratic legislators and the second column is for Republican legislators. 16 Results: Legislators’ Positions and Constituent Voting While our non-parametric analysis suggests little reason to believe that the roll call positions of legislators influence voters decisions on election day, the link between the graphs and the theoretical predictions are somewhat loose. To make a clearer connection between theory and evidence, we next turn to our parametric framework. Table 4 shows the results of a logistic regression model based on Equation 9. We control for contest random effects (not shown). Each column of the table corresponds with a different within-party specification. Column 1 shows the cases where all incumbents are Democrats and all constituents are Democrats, column 2 shows the specification where all incumbents are Democrats and all constituents are Independents, and so forth. Contrary to a purely affective notion of party identification, the ideology of the voter is significant in every specification in the predicted direction. More liberal voters are more likely to support Democrats and more conservative voters are more likely to support Republicans. However, party has an independent effect as well. The intercepts in columns 1 and 6 are substantially higher than the models where the party of the legislator and the party of the voter do not match. As in the non-parametric case, this is the biggest story our table has to tell: the partisanship and ideology of voters matters. Evidence in this table is very weak for the idea that citizens vote spatially based on their proximity with individual legislators. Only one column, column 4, has significant coefficients that correspond with Hypothesis 1. In particular, the key coefficient is the interaction between legislator ideology and citizen ideology. This is the interaction that allows vote to depend on the distance between citizens and legislators, not merely the position of the citizen or legislator alone. This interaction is always substantively small, and only significant in one out of our four cases: the case of Democratic constituents and Republican incumbents. Instead, the coefficients for voters’ ideology are the most significant set of variables in each regression, which lends further support to the notion that voters think spatially with reference to parties rather than candidates. 17 Table 4: Proximity Behavior in Congressional Elections Dependent variable: Vote for Incumbent (1) (2) (3) (4) (5) (6) Legislator: Dem/ Dem/ Dem/ Rep/ Rep/ Rep/ Constituent: Dem Ind Rep Dem Ind Rep Citizen Ideology ∗∗ 0.83 (0.12) ∗∗ −1.62 (0.13) −1.55 (0.18) −1.33 (0.10) 1.19 (0.20) 1.30∗∗ (0.15) Citizen Ideology Sq. −0.11∗∗ (0.04) −0.03 (0.05) 0.05 (0.03) −0.06∗∗ (0.03) 0.05 (0.04) −0.02 (0.03) Legislator Ideology −0.31 (1.84) 0.51 (2.16) 1.42 (1.69) −4.33∗∗ (1.60) 1.56 (2.41) −1.62 (2.03) Legislator Ideology Sq. 1.48 (2.68) −0.53 (3.10) −0.69 (2.47) 2.74∗∗ (1.28) −1.77 (1.92) 1.66 (1.63) Leg. Dem x Citizen Ideology −0.27 (0.36) 0.04 (0.47) 0.22 (0.28) 0.39∗∗ (0.19) 0.19 (0.32) 0.04 (0.25) (Intercept) 2.11∗∗ (0.30) 0.76∗∗ (0.36) −0.55∗ (0.28) 0.69 (0.48) 0.20 (0.73) 2.53∗∗ (0.61) 18,723 −2,422.36 4,858.72 4,913.58 3,105 −1,276.93 2,567.87 2,610.15 14,798 −3,398.16 6,810.32 6,863.54 16,296 −5,178.87 10,371.73 10,425.62 3,588 −1,572.98 3,159.95 3,203.25 21,285 −2,822.45 5,658.89 5,714.65 Observations Log Likelihood Akaike Inf. Crit. Bayesian Inf. Crit. ∗∗ ∗∗ ∗∗ ∗ p<0.1; ∗∗ p<0.05 Note: Coefficients are from the logistic regression model in equation 9, plus controls for contest-specific random effects (not shown). Standard errors in parentheses. Each column restricts the set of observations to only constituents and incumbents of a particular party. So column 1 is Democratic constituents who are voting on Democratic incumbents and so on. The unit of observation is the constituent or citizen. 18 The Effect of Incumbent Positioning on Constituting Voting It is difficult to interpret the magnitude of the effect of the spatial distance between voters and candidates in our analysis, because distance is embedded in the specification but not included as a single coefficient. In order to examine the effects of distance directly, we calculate the predicted change in the probability that voters in a given group would support a legislator that moderated their ideal point by one standard deviation. Table 5 shows these results. The left side of the table shows the results for Democratic incumbents, and the right side of the table shows the results for Republican incumbents. Each column shows the ideology of the constituent, split into three equal groups (Liberal, Moderate, and Conservative) and each row shows the party of the constituent. So for instance, for liberal Democrats with Democratic incumbents, a one standard deviation move to the right by the incumbent is associated with a -0.1 percentage point change in their probability of voting for the incumbent. This is a percentage point change, so the vote probability is reduced by 0.1% out of 100%. The top row of Table 5 provides a theoretical prediction based on spatial voting theory for each group. We assume that moderate and conservative members of the public are to the right of Democrats in the House of Representatives, and that liberals and moderates are to the left of their Republican legislators. If this assumption is correct, then a one standard deviation move towards the middle should always be a move closer to these constituents, and therefore it should increase their probability of voting for the incumbent. This is reflected in the theoretical prediction in the top row: for these rows, predicted changes in vote probability should be positive. For liberals who have a Democratic incumbent and conservatives who have a Republican incumbent, we are less certain of whom is to the right or left of whom, and so our theoretical prediction is ambiguous. Evidence for the candidate-centered spatial voting model is very weak. Of the twelve cases where we have a theoretical prediction, the theoretical prediction of spatial voting theory (H1) is only born out significantly in five. Of these cases, the greatest change in vote probability is 5.4%, for moderate Republicans voting on a Democrat. The rest are smaller 19 than 3%. In one case, the coefficient is negative and significant. Indeed, moderate Democrats appear to prefer liberal Democratic incumbents. Also, note that half of the estimated effects across the entire table are less than 1% in magnitude. Democratic Incumbent Liberal voter Mod. voter Conserv. voter Candidate-Centered Theoretical Prediction (H1) Republican Incumbent Liberal voter Mod. voter Conserv. voter ambiguous + + + + ambiguous Democratic voter -0.09%** (.04) -1.14%** (.26) -5.34% (3.34) 1.10** (.16) 3.18%** (.80) 0.66% (1.96) Independent voter .39% (.31) 2.74%* (1.57) .76% (.50) 1.85% (1.26) 2.45% (1.54) 0.12% (.32) Republican voter 2.87%* (1.60) 5.36%** (1.14) 0.72%** (.12) -1.20% (3.46) -0.68% (.51) -0.07% (.06) Table 5: Effect of incumbent moderating their ideal point by one standard deviation on the probability that citizens vote for the incumbent in congressional elections. Effects that are significant at the .05 level are shown with two stars and effects that are significant at the .1 level are shown with one star. Two objections that may arise to this analysis are that (a) our results may be be driven by the fact that we focus on overall ideological scores rather than important votes in Congress, and (b) our results may be driven by districts where incumbents face weak or no challengers. In Appendix B, we show that our substantive results are similar for important votes, and in Appendix C we show that our results are similar in competitive and uncompetitive districts. Finally, all our results are substantively similar using both static and dynamic ideal point estimates (e.g., DW-Nominate vs W-Nominate scores). Constituent Perceptions of Roll Call Positions Spatial voting is a theoretically and intutively appealing idea that has motivated a wide body of work in political science for decades. Why don’t voters make choices on the basis of policy 20 proximity with individual legislators? One answer comes from seminal research on representation in the 1960s, which found that citizens have only vague notions of legislators’ roll call positions (Miller and Stokes, 1963). For instance, only 56 percent of ANES respondents correctly identified their representative’s position on the resolution in 1991 authorizing the first President Bush to conduct the Persian Gulf War (Alvarez and Gronke, 1996). Moreover, only 63 percent of ANES respondents correctly identified their representatives’ votes on the budget-balancing Budget Resolution of 1993 (Lipinski, 2001). However, more recent research suggests that most contemporary voters may have accurate perceptions of legislators’ roll call positions (Ansolabehere and Jones, 2010). To our knowledge, though, no study has systematically evaluated how well citizens know the ideology of members of Congress using a large sample of voters and legislators. We take two approaches to examining whether citizens understand legislators’ roll call positions. First, we examine the relationship between constituents’ perception of legislators’ symbolic ideology on a 7-point scale and their actual roll call positions (top-panel of Figure 2). We have data for over 150,000 Americans’ perceptions of their legislators’ ideology from the 2006, 2008, 2010, and 2012 CCES. We find that citizens are very capable of differentiating between Democrats and Republicans. Overall, there is a correlation of .71 between the perceived ideology of legislators and their actual DW-Nominate scores. However, our findings suggest that while voters are capable of differentiating between parties, they are less capable of differentiating within parties. When we subset the data by legislators’ party, there is only a correlation of .14 between the perceived ideology of Democratic legislators and their actual DW-NOMINATE scores and an even weaker correlation of .09 between the perceived ideology of Republican legislators and their actual DW-Nominate scores. One possible explanation for the weak within-party correlations between actual and perceived legislator ideologies we observe is that the 7-point scale on the CCES is not granular enough to distinguish legislators within each party. For instance, it’s possible that most Democrats are a “2” or a “3”. To examine this possibility, we re-examined data on voters’ 21 Perceived Legislator Ideology 7 6 5 4 3 2 1 −0.5 0.0 0.5 1.0 Perceived Legislator Position (Scaled from CCES) Incumbent Position (DW−Nominate Score) ● ● ● ● 1 ● −1 ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ●● ● ●● ●● ●● ● ● ● ● ● ●●● ●● ● ● ●● ● ● ● ● ● ●● ● ●● ● ●● ● ● ● ● ●● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●●● ● ●● ●●● ● ● ●● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ●● ●● ●● ● ● ● ●●●● ● ●● ●●● ● ● ● ●● ● ● ●● ●● ●● ● ●● ●● ●● ● ●● ● ● ● ● ●● ●●● ● ● ●● ● ● ● ●●●● ● ● ●● ● ● ● ●●● ● ●●●● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ●● ● ● ● ●● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ● ●●●●● ● ●● ● ● ● ● ●● ●● ● ●●● ● ● ● ●●●● ●● ●● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● −1 ● ● ●● ●● ● ● ●● ● ●●● ● ● ● ● ● ●● ● ● ● ● ● ●● ●● ● ● ●●● ● ● ● ●●● ●●● ●● ● ● ● ● ● ● ●● ● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●●●● ●● ●● ● ● ● ● ●●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●● ● ● ●● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●● ●●●● ●●● ●● ● ● ●● ● ●● ● ●● ● ● ●● ● ● ● ● ● ●●● ●●●●● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ●● ● ●● ●● ● ●● ●●● ●●●● ● ● ● ●● ● ●● ● ● ● ●● ●●● ●●● ●● ●● ●● ● ●● ●● ●●● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ●● ●●● ●● ●● ●● ●● ● ● ●● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ●● ● ● ●●●●● ● ●● ● ● ●●● ● ●●● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ●● ● ● ●● ● ● ●●● ● ● ● ● ● ● ●● ●●● ●● ● ● ● ● ● ● ●● ● ●● ● ●●● ●● ●● ●●● ● ● ● ● ●●●● ● ● ●●● ●● ● ● ●● ● ●●● ● ●● ● ● ● ● ● ●● ● ●● ●●● ● ●●● ●● ● ● ● ●● ● ● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ●● ● ● ● ●●● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● 0 Legislator Position (8 Roll Calls in CCES) 1 Figure 2: Perceived vs. Actual Ideal Points – The top panel in this graph shows the relationship between the perceived ideology of House members on a 7-point scale and their actual ideal points. The grey region in the middle of the top panel shows the ideological space where few legislators reside. The bottom panel shows the relationship between the perceived ideology of House members based on respondents’ perceptions of their votes on eight salient roll calls and their actual ideal points based on their votes on these bills. perceptions of legislators’ positions on eight roll call votes from the 2006 CCES (Ansolabehere and Jones, 2010).13 We scaled both the perceived and actual positions of legislators on these eight votes using an IRT model (Clinton, Jackman, and Rivers, 2004).14 The bottom panel of Figure 2 shows the results of this analysis. There is a clear rela13 The 2006 CCES asked about voters’ perceptions of legislators’ roll-call votes on (1) a ban on partial-birth abortion, (2) federal funding for stem cell research, (3) extending capital gains tax cuts, (4) ratifying CAFTA, (5) immigration reform, (6) bankruptcy reform, (7) tax breaks for energy companies, and (8) reauthorizing the Patriot Act. See Ansolabehere and Jones (2010) for more details. 14 Note that given that we only have eight items in this scale, there is probably significant measurement error in our estimates of both perceived and actual legislator positions. However, the measures of actual positions are correlated with legislators’ DW-Nominate scores at .90. 22 tionship between the perceived and actual ideal points of legislators on these eight roll calls. However, there is a considerable degree of error in respondents’ perceptions of legislator positions. The overall correlation between the perceived and actual positions of legislators on these eight roll calls is .66. When we subset the data by legislators’ party, however, there is only a correlation of .28 between the perceived positions of Democratic legislators and their actual ideal points on these roll calls and a correlation of .10 for Republican legislators. We should note that these 8 items distinguish more effectively between Democrats than Republicans, possibly explaining the difference between these two correlations. Overall, the evidence suggests that most voters understand the party label of their representative and they can effectively differentiate between Democrats and Republicans in Congress. But voters only have a dim awareness of ideological differences between legislators within each party. Condorcet’s jury theorem shows that sometimes only a dim awareness is necessary for voters to make good decisions under some circumstances (Condorcet, 1785), and it has been shown that these dim perceptions are good enough on average that political scientists can use them to identify the actual positions of legislators (Aldrich and McKelvey, 1977; Ramey, 2014). Nonetheless, each individual is only slightly better than a coin flip at telling whether one Democrat is more or less liberal than another. Even if individuals are trying to make proximity-based judgements, lack of knowledge adds noise to individual spatial decision making, to the point where it may be easily swamped by other considerations. Does Political Knowledge Affect the Relationship Between Legislators’ Positions and Constituent Voting? Maybe, given the results in the last section, we should not expect most voters to vote based on legislators’ spatial positions. Perhaps this requires too much information or cognitive capacity. Indeed, several previous studies have found evidence that politically knowledgeable citizens are much more likely to vote based on legislators’ spatial positions than low information citizens (Jessee, 2009, 2012; Shor and Rogowski, 2010). To examine this possibility, we 23 used a battery of political knowledge questions to examine whether high information voters were more likely to punish legislators that took extreme positions than low-information voters. We scaled the political knowledge of respondents to the 2006, 2008, 2010, and 2012 Cooperative Congressional Election Studies (CCES) using responses to eight knowledge items.15 On the basis of respondents’ answers, we characterized their political knowledge as “high”, “medium”, and “low” (Shor and Rogowski, 2010). Not surprisingly, this measure is highly correlated with respondents’ education levels. Next, we ran models like the ones in Table 4 for voters at low and high levels of political information. We used these models to predict the changes in the probability of voting for the incumbent that would result from a one standard deviation move in the moderate direction by Democratic and Republican legislators. The results are shown in Table 6. Table 6 is laid out exactly as in Table 5, but with with an upper half for low information voters and a lower half for high information voters. As in Table 5, many of the coefficients do not bear out the predictions of the model of spatial voting based on candidate positions. This is particularly true for low information voters. Only three cases are significantly positive, and two are significantly negative. However, for high information voters the evidence is somewhat more mixed. No cases that are predicted to be positive are significantly negative, and seven out of twelve cases that are predicted to be positive are significantly so, some with substantively large effects. High information voters may be somewhat attuned to the policy distance between their own preferred policies and the votes cast by their incumbent legislators. 15 To scale political knowledge, we use items on respondents’ knowledge about the majority party in the House, the majority party in the Senate, the major party in their state house, and majority party in their state senate, the party of their governor, the party of their incumbent representative, and the party of their incumbent senators. 24 Democratic Incumbent Liberal voter Mod. voter Conserv. voter Candidate-Centered Theoretical Prediction (H1) Low Information Voters Democratic voter Republican Incumbent Liberal voter Mod. voter Conserv. voter ambiguous + + + + ambiguous -0.39%** (.08) -2.0%** (.32) -13.68%* (8.09) 2.67%** (.66) 3.47% (1.23) -2.44% (6.11) Independent voter -0.27% (0.83) 1.84% (2.62) 2.61% (2.03) 2.73% (6.56) 4.29% (3.02) 1.38% (2.01) Republican voter 14.97%* (8.87) 7.05%** (2.37) -0.5% (.84) 4.59% (5.90) -0.78% (.44) -0.52%** (.12) 0.02%** (0.01) 0.25% (0.35) -0.04% (2.27) 0.34%** (0.09) 3.87%** (1.06) 6.03%* (3.07) Independent voter .93%** (0.23) 4.46% (4.97) -0.98%* (0.51) 1.66%** (0.56) 3.26% (3.30) -0.13% (0.09) Republican voter 2.34%* (1.35) 7.43%** (2.33) 0.54%** (0.08) 1.15% (3.41) 0.78% (0.97) 0.05% (0.04) High Information Voters Democratic voter Table 6: Effect of incumbent moderating their ideal point by one standard deviation on the probability that citizens vote for the incumbent in congressional elections. Effects that are significant at the .05 level are shown with two stars and effects that are significant at the .1 level are shown with one star. Extrapolating Legislator Vote Shares Having examined what these results mean for theories of electoral accountability and spatial voting, what do they imply for representation in American politics? Our model enables us to examine the relationship between legislator vote shares and legislator positions (cf. Canes-Wrone, Brady, and Cogan, 2002; Wilkins, 2012). It may be the case, for instance, that some small subset of the public votes according to the spatial theory of voting. If so, this nonetheless will have very small incentive effects for legislators. In order to test this, we simulate vote shares for each legislator from the sample of their actual electorate in our 25 dataset using a model derived from the models presented in Table 4.16 We produce predicted vote shares for each district at different values of incumbent DW-NOMINATE scores. Figure 3 show the results of our exercise in predicting vote shares. For each district, we calculate the change in vote share that would result from a one standard deviation move toward the center by the legislator. The left panel shows the kernel density plot of predicted changes in vote share for all districts represented by Democrats, and the right panel shows this density for all districts represented by Republicans. The mean and mode of these densities are very close to 0. In every case, Democrats are projected to change their voteshares by less than 3%, and in the large plurality less than 3%. These changes are only a little more likely to be increases than decreases. Likewise, in every case Republicans would increase their voteshares by less than 4%, and overwhelmingly these changes are less than 2%. Once again, negative changes are common. The cumulative effect of our model of individual-level voting is that legislator positions appear to have very small associations with their vote shares. Far more important drivers of vote shares are the legislator’s party and the partisan and ideological composition of the district. Conclusion The Founding Fathers thought that frequent elections were the key mechanism for ensuring that the “will of the people” is carried out. This electoral connection provides the foundation for the study of congressional behavior and lawmaking and for theories of representation more broadly. A number of recent studies have provided an empirical foundation for the assumption that voters hold their representatives accountable at the ballot box for their roll call voting behavior (Ansolabehere and Jones, 2010; Jessee, 2009; Jones, 2011; Shor and Rogowski, 2010). Yet these studies are puzzling considering the polarization and lack of 16 We used a combined, fully interactive version of the model where each model in Table 4 is embedded in one specification. This is substantively the same as using each of these models separately. Note that due to our large sample of voters’ ideal points, we have an average of roughly 350 people in every congressional district. 26 50 60 40 30 40 20 20 10 0 0 −0.02 −0.01 0.00 0.01 Democrats' Vote Share Gain 0.02 −0.04 −0.02 0.00 0.02 Republicans' Vote Share Gain 0.04 Figure 3: Relationship between Representatives’ Ideal Point and Expected Vote Share in the U.S. House: 2006-2012 – This graph shows the distribution of potential vote gains from legislators that moderate their position by one standard deviation. The left column responsiveness in the contemporary Congress. In this study, we provide an individual-level explanation for the lack of support for the spatial voting model’s core predictions in the modern House of Representatives. We find that voters’ policy preferences are highly predictive of which party they will support: liberal voters almost always support Democrats and conservative voters almost always support Republicans. Unlike previous survey-based studies, however, we find that there is a negligible association between legislators’ roll call voting behavior and the probability that most citizens will support them on election-day. Indeed, most low knowledge voters do not take legislators’ voting position into account at all. We find that only high knowledge voters who are moderate or cross-pressured take legislators’ positions into account in the ballot box, and that even 27 then the effect is small. This precludes a major role for all proximity-based accounts of voter choice based on candidate positions, including both the classic spatial model as well as newer variants of proximity voting models such as directional voting. It also provides few incentives for legislators to take ideologically moderate positions in Congress. We posit an explanation for our results grounded in the original applications of spatial voting theory. While voters have difficulty differentiating liberal Democrats from moderate Democrats, and conservative Republicans from moderate Republicans, they are quite capable of distinguishing between the parties. They may be able to roughly observe the proximity of their own desired policies to the policies supported by each party, and vote accordingly. This party-centered notion of spatial voting is supported by our individual level evidence. Voter ideology does appear to matter, it is just the individual incumbent’s ideal point which has very little predictive power. Proximity to the candidate plays a tiny role in voter decisions, if any, but proximity to the parties may well explain the importance of voter ideology. While our findings clearly contradict candidate-centered accounts of spatial voting in congressional elections (e.g., Ansolabehere and Jones, 2010; Nyhan et al., 2012; Shor and Rogowski, 2010),17 they do not fit neatly into the traditional framework of party identification. Voters’ policy positions appear to be a strong driver of party loyalty in U.S. House elections. This finding contrasts with the traditional view of party identification as primarily affective. Our findings are consistent with work that attempts to incorporate spatial voting in the context of party reputations (e.g., Sniderman and Stiglitz, 2012). This perspective suggests that legislators have strong electoral incentives to work to maintain the brand of their party (Aldrich, 1995; Cox and McCubbins, 2005). Indeed, our results suggest that voters punish parties that grow out-of-step with their constituents (Kim and LeVeck, 2013). The electoral connection in Congress may be alive and well, but at the level of parties rather than individual legislators. Unlike the candidate-centered view, in this view the incentives of individual legislators will often conflict with the incentives of parties. 17 It is worth noting however, that while the median voter theorem is strictly a phenomena of two-candidate elections, Downs (1957) formulated his spatial voting theory in terms of parties, not individual candidates. 28 References Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Party Politics in America. Vol. 15 Chicago, IL: Cambridge University Press. Aldrich, John H, and Richard D McKelvey. 1977. “A Method of Scaling with Applications to the 1968 and 1972 Presidential Elections.” The American Political Science Review pp. 111– 130. Alvarez, R Michael, and Paul Gronke. 1996. “Constituents and Legislators: Learning about the Persian Gulf War Resolution.” Legislative Studies Quarterly 21(1): 105–127. Ansolabehere, Stephen, and Philip Edward Jones. 2010. “Constituents’ Responses to Congressional Roll-Call Voting.” American Journal of Political Science 54(3): 583–597. Ansolabehere, Stephen, James M Snyder Jr, and Charles Stewart III. 2001. “Candidate Positioning in US House Elections.” American Journal of Political Science 45(1): 136– 159. Black, Duncan. 1948. “On the Rationale of Group Decision-Making.” The Journal of Political Economy 56(1): 23–34. Bonica, Adam. 2013. “Ideology and Interests in the Political Marketplace.” American Journal of Political Science 57(2): 294–311. Brady, David W, Morris P Fiorina, and Arjun S Wilkins. 2011. “The 2010 elections: Why Did Political Science Forecasts Go Awry?” PS: Political Science & Politics 44(02): 247–250. Calvert, Randall L. 1985. “Robustness of the multidimensional voting model: Candidate motivations, uncertainty, and convergence.” American Journal of Political Science pp. 69– 95. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. New York, NY: Wiley. 29 Canes-Wrone, Brandice, David W. Brady, and John F. Cogan. 2002. “Out of Step, Out of Office: Electoral Accountability and House Members’ Voting.” American Political Science Review 96(1): 127–140. Clinton, Joshua D. 2006. “Representation in Congress: Constituents and Roll Calls in the 106th House.” Journal of Politics 68(2): 397–409. Clinton, Joshua, Simon Jackman, and Douglas Rivers. 2004. “The Statistical Analysis of Roll Call Data.” American Political Science Review 98(2): 355–370. Condorcet, Marquis de. 1785. “Essay on the Application of Analysis to the Probability of Majority Decisions.” Paris: Imprimerie Royale . Conover, Pamela Johnston, and Stanley Feldman. 1989. “Candidate Perception in an Ambiguous World: Campaigns, Cues, and Inference Processes.” American Journal of Political Science 33(4): 912–940. Cormack, Lindsey. 2013. “Sins of Omission: Legislator (Mis)Representation in Constituent Communications.” Presented at the APSA Annual Meeting . Cox, Gary W, and Mathew D McCubbins. 2005. Setting the Agenda: Responsible Party Government in the US House of Representatives. Cambridge University Press. Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row. Enelow, James M, and Melvin J Hinich. 1984. The Spatial Theory of Voting: An Introduction. Cambridge University Press. Green, Donald, Bradley Palmquist, and Eric Schickler. 2004. Partisan Hearts and Minds: Political Parties and the Social Identities of Voters. Yale University Press. Griffin, John D. 2006. “Electoral Competition and Democratic Responsiveness: A Defense of the Marginality Hypothesis.” Journal of Politics 68(4): 911–921. 30 Grimmer, Justin. 2013. Representational Style in Congress: What Legislators Say and why it Matters. Cambridge University Press. Hall, Andrew B. 2014. “What Happens When Extremists Win Primaries?” American Political Science Review Forthcoming. Henderson, John Arthur. 2013. “Downs’ Revenge: Elections, Responsibility and the Rise of Congressional Polarization.”. Hotelling, Harold. 1929. “Stability in Competition.” The Economic Journal 39(153): pp. 41–57. Jessee, Stephen A. 2009. “Spatial Voting in the 2004 Presidential Election.” American Political Science Review 103(1): 59–81. Jessee, Stephen A. 2012. Ideology and Spatial Voting in American Elections. New York, NY: Cambridge University Press. Joesten, Danielle A, and Walter J Stone. 2014. “Reassessing Proximity Voting: Expertise, Party, and Choice in Congressional Elections.” The Journal of Politics 76(3): 740–753. Jones, Phillip Edward. 2011. “Which Buck Stops Here? Accountability for Policy Positions and Policy Outcomes in Congress.” Journal of Politics 73(3): 764–782. Kim, Henry A, and Brad L LeVeck. 2013. “Money, Reputation, and Incumbency in US House Elections, or Why Marginals Have Become More Expensive.”The American Political Science Review 107(3): 492–504. Lee, David S, Enrico Moretti, and Matthew J Butler. 2004. “Do Voters Affect or Elect Policies? Evidence from the US House.” The Quarterly Journal of Economics pp. 807– 859. Levitt, Steven D. 1996. “How do senators vote? Disentangling the role of voter preferences, party affiliation, and senator ideology.” The American Economic Review pp. 425–441. 31 Lewis-Beck, Michael S, William G Jacoby, Helmut Norpoth, and Herbert F Weisberg. 2008. The American Voter Revisited. University of Michigan Press. Lewis, Jeffrey, and Chris Tausanovitch. 2013. “Has Joint Scaling Solved the Achen Objection to Miller and Stokes?” Presented at the Vanderbilt Miller-Stokes Conference on Representation. Lipinski, Daniel. 2001. “The Effect of Messages Communicated by Members of Congress: The Impact of Publicizing Votes.” Legislative Studies Quarterly 26(1): 81–100. Macdonald, Stuart Elaine, Ola Listhaug, and George Rabinowitz. 1991. “Issues and Party Support in Multiparty Systems.” The American Political Science Review 85(4): 1107–1131. Mayhew, David. 1974. The Electoral Connection. New Haven, CT: New Haven: Yale University Press. Miller, Warren E, and Donald E Stokes. 1963. “Constituency Influence in Congress.” The American Political Science Review 57(1): 45–56. Nyhan, Brendan, Eric McGhee, John Sides, Seth Masket, and Steven Greene. 2012. “One Vote Out of Step? The Effects of Salient Roll Call Votes in the 2010 Election.” American Politics Research 40(5): 844–879. Poole, Keith T, and Howard Rosenthal. 2000. Congress: A Political-Economic History of Roll Call Voting. New York, NY: Oxford University Press. Popkin, Samuel L. 1991. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. University of Chicago Press. Rabinowitz, George, and Stuart Elaine Macdonald. 1989. “A Directional Theory of Issue Voting.” The American Political Science Review 83(1): 93–121. Ramey, Adam. 2014. “Vox Populi, Vox Dei? Estimating Ideal Points with Voter Perceptions.” Working Paper, Social Science Research Network . 32 Shor, Boris, and Jon C Rogowski. 2010. “Congressional Voting by Spatial Reasoning.” Presented at the Annual Meeting of the Midwest Political Science Association, Chicago. Simas, Elizabeth N. 2013. “Proximity voting in the 2010 US House elections.” Electoral Studies 32(4): 708–717. Sniderman, Paul M, and Edward H Stiglitz. 2012. The Reputational Premium: A Theory of Party Identification and Policy Reasoning. Princeton University Press. Snyder, James M, and Michael M Ting. 2002. “An Informational Rationale for Political Parties.” American Journal of Political Science 46(1): 90–110. Tausanovitch, Chris, and Christopher Warshaw. 2013. “Measuring Constituent Policy Preferences in Congress, State Legislatures and Cities.” Journal of Politics 75(2): 330–342. Tomz, Michael, and Robert P Van Houweling. 2008. “Candidate Positioning and Voter Choice.” American Political Science Review 102(3): 303–318. Wilkins, Arjun S. 2012. “The Effect of Extreme Incumbent Roll-Call Voting Records on U.S. House Elections, 1900-2010.” Prepared for Presention at the Annual Meeting of the American Political Science Association, New Orleans. Wittman, Donald. 1983. “Candidate motivation: A synthesis of alternative theories.” American Political Science Review 77(01): 142–157. 33 Online Appendix A: Models of Spatial Voting In this section, we elucidate the mathematical relationship between the logistic regression equation estimated in the paper and the theory of spatial voting. Assume that voters have quadratic utility functions over the distance between their ideal policy and the ideal policy of the legislator they choose to vote for. Let xi be the position of voter i, rj be the position of the incumbent representative in contest j, and cj be the position of the challenger. For simplicity, we begin by assuming that cj takes on values as follows: cj = µR if rj < 0 (11) µD if rj ≥ 0 Where µR is the position of the mean Republican and µD is the position of the mean Democrat. In other words, for the purposes of this example, voters assume that the challenger takes the mean position of the non-incumbent party.18 Let yij be voter i’s vote in contest j, where y is 1 if i votes for the incumbent and 0 otherwise. Voter i’s utility for voting for the incumbent and challenger is as follows: uij (yij = 1) = −(xi − rj )2 + vj + ij uij (yij = 0) = −(xi − cj )2 (12) In this model, vj represents the valence attached to the incumbent, which we assume to be a function of the correspondence between the partisan identification of legislators and citizens. This allows us to capture the potential that voters may make judgments based on 18 We empirically examine this assumption using estimates of candidate positions derived from campaign contributions (Bonica, 2013). We find that there is no correlation between the positions of candidates and the positions of incumbents. Moreover, there is only a modest correlation between the positions of candidates and the median voter in their district. Instead, consistent with our model, candidates’ party identification is the most important predictor of their positions. 34 party rather than position. Finally, ij is an independent random shock to the utility for the incumbent. For simplicity, we assume that utility for the challenger is errorless. Assume that ij is standard logistically distributed. Then the probability of voter i voting for the incumbent in contest j is: P (yij = 1) = Λ((xi − rj )2 − (xi − cj )2 + vj ) = Λ(−2xi rj + rj2 + 2xi cj − c2j + vj ) (13) (14) Here Λ() is the cumulative standard logistic distribution function. Let β0∗ = −c2j , β1∗ = 2cj , β2∗ = 1 and β3∗ = −2. Then: P (yij = 1) = Λ(β0∗ + β1∗ xi + β2∗ rj2 + β3∗ xi rj + β4∗ vj ) (15) We do not assume that rj is in the same space as xi . Instead, we assume that these spaces are linear transformations of one another, such that δxi + γ resides in the space of rj . Then: P (yij = 1) = Λ(β0∗ + β1∗ δxi + β1∗ γ + β2∗ rj2 + β3∗ δxi rj + β3∗ γrj + β4∗ vj + β5∗ vj ) (16) We do not know a priori the values of δ and γ, so an empirical specification will be consistent with any value for these variables. Let β0 = β0∗ + β1∗ γ, β1 = β1∗ δ, β2 = β2∗ , β3 = β3∗ δ and β4 = β3∗ γ. Then we have : P (yij = 1) = Λ(β0 + β1 xi + β2 rj2 + β3 xi rj + β4 rj + β5∗ vj ) This simple logit equation includes equation 13 as a special case. 35 (17) In the above model, we motivated our empirical specification by using the canonical quadratic utility spatial voting model with fixed partisan challengers. However, there are many other possible utility functions and many other possible challenger position functions. Moreover, the findings in the paper can be made consistent with most types of spatial voting, but only if voters display very high levels of indifference or low levels of acuity. Below we flesh out the implications of some alternative models of voter decisionmaking and how they may or may not be consistent with our findings. None of these alternatives can explain away the basic finding that voter decisionmaking is only weakly contingent on incumbent positions. Challenger Positioning In the above model, we assumed that challengers adopt the following positions: µR if rj < 0 cj = µD if rj ≥ 0 Empirically, the relative positioning of challengers and incumbents is not well established. Moreover, voters may view the challenger differently than they view the incumbent. One possibility is that voters consider the challenger to be an unknown, voting retrospectively based on the position of the incumbent only. In this case, utility for the challenger would be fixed: uij (yij = 0) = c Where c is just a fixed utility for an “unknown quantity.” The resulting vote generating function is: P (yij = 1) = Λ((xi − rj )2 − c) = Λ(x2i − 2xi rj + rj2 − c) 36 In this case, the x2i do not cancel. For this reason, we included x2i in our empirical specification, and this formulation is encapsulated by the model we estimated. The “fixed-utility challenger” model assumes voters that are less sophisticated than the “average partisan challenger” model that we used above. Under fixed challenger utility, voters do not seem to be cognisant of the fact that challengers in general elections typically come from the opposite party, and may represent positions that are highly dissonant with voter preferences. Another alternative is to assume that voters are even smarter than the voters in the “average partisan challenger” model, and that they know the exact position of the challenger. In the absence of good information about challenger positions, we then need to specify a positioning function for the challenger. A simple and plausible choice is that challengers emerge as a random draw from the existing party. More specifically: N (µR , σR ) if rj < 0 cj = N (µD , σD ) if rj ≥ 0 In order to understand how this candidate generating function would affect our estimates, we first simulate data from our original data generating function for comparison, separated by legislator party. Rather than randomly generating candidates and voters, we use the legislators and voters in our actual data- this ensures our simulations have the same statistical power as our actual analysis.19 We z-score the positions of legislators and voters, and generate data from the following vote generating function: P (yij = 1) = Λ((xi − rj )2 − (xi − cj )2 ) = Λ(−2xi rj + rj2 + 2xi cj − c2j ) In the “random challenger” model, c is generated as above. We begin by simulating vote probabilities from the standard model. The results are shown in Figure 4. 19 In particular, it would be very useful if we could randomly assign a different incumbent ideal point for each voter, since this would give us much more variation in legislator ideal points than we actually have in our data 37 Democratic Incumbents Republican Incumbents 1.00 Projected Vote Share 0.75 0.50 0.25 0.00 −1.5 −1.2 −0.9 Incumbent Position −0.6 0.6 0.9 1.2 Incumbent Position 1.5 Figure 4: Simulated Vote For The Data Generating Process Used in the Paper: The y-axis is probability of voting for the incumbent, and the x-axis is the incumbent’s ideal point. Probabilities are generated by a simulated quadratic utility model. Each line represents the vote probability for a set of voters with a given ideology. Voter ideology ranges from -1.5 to 1.5, with voters at -1.5 in dark blue and voters at 1.5 in dark red. Each line in between represents one of 11 separate categories of ideology, colored by closeness to 1.5 or -1.5. Figure 5 shows a simulation based on the random-challenger model. The results are almost exactly identical to the results of the non-random challenger model. More importantly, when we flip this process around and estimate our model on the simulated votes, we recover almost exactly the correct estimates, thanks to the large size of our sample. In other words, the random-challenger alternative is encapsulated by the model we used in the body of the paper. So far we have assumed that challenger positions do not depend on the position adopted by the incumbent, but this may not be true. Consider an extreme case: cj = −rj 38 Democratic Incumbents Republican Incumbents 1.00 Projected Vote Share 0.75 0.50 0.25 0.00 −1.5 −1.2 −0.9 Incumbent Position −0.6 0.6 0.9 1.2 Incumbent Position 1.5 Figure 5: Simulated Vote For The Data Generating Process Used with Random Challengers: The y-axis is probability of voting for the incumbent, and the x-axis is the incumbent’s ideal point. Probabilities are generated by a simulated quadratic utility model. Each line represents the vote probability for a set of voters with a given ideology. Voter ideology ranges from -1.5 to 1.5, with voters at -1.5 in dark blue and voters at 1.5 in dark red. Each line in between represents one of 11 separate categories of ideology, colored by closeness to 1.5 or -1.5. Further assume that votes are generated by quadratic utility as follows: P (yij = 1) = Λ((xi − rj )2 − (xi − cj )2 ) = Λ((xi − rj )2 − (xi + rj )2 ) substituting −rj for cj = Λ(−4xi rj ) This possibility is encapsulated by our model, and contradicted by our results. In fact, it can be shown that there is only one challenger strategy that will induce indifference over rj for all x. Consider values of c that would satisfy: (x − r)2 − (x − c)2 = 0 39 Doing the algebra: (x − c)2 = (x − r)2 p x − c = (x − r)2 p c = x − (x − r)2 So c is either equal to r or equal to 2x − r. Since c cannot be equal to 2x − r for all x, the unique solution that does not depend on x is c = r. In other words, the only way that candidate strategy can induce spatial voters to vote independently of r for all possible voter positions is for the challenger to always adopt the position of the incumbent. This strategy is starkly contradicted by the lack of convergence in actual House elections. Alternate Utility Functions In the above model, we rely on quadratic utility as a theoretical motivation. However, there are many other possibilities. Probably the most popular alternative is the normal utility model. Normal utility has very different implications from quadratic utility for voter decisionmaking. If voter utility is normal, then very extreme voters may be closer to indifferent between candidates than the average partisan is. This is due to the flat tails of the normal function. A normal utility model can be specified as follows: 1 uij (yij = 1) = exp(− (xi − rj )2 ) + ij 2 Where ij is an independent random shock to the utility for the incumbent. Once again, if we assume that utility for the challenger is errorless: 1 uij (yij = 0) = exp(− (xi − cj )2 ) 2 40 Asuming ij is distributed standard normal, the probability of voting for the incumbent is equal to: 1 1 P (yij = 1) = Φ(exp(− (xi − rj )2 ) − exp(− (xi − cj )2 )) 2 2 Unlike the many possible assumptions about challenger location, not all data generating processes based on normal utility are encapsulated by the empirical specification we use in our parametric analysis in the main body of the paper. However, the kinds of voting curves generated by normal utility are starkly inconsistent with our non-parametric results. As an example that is not too far from what our empirical specification can estimate, take the data generating function just considered, with challengers specified as in beginning of this appendix. As above, we use the data from our paper to simulate vote probabilities, but z-score the ideology measures. The result is shown in Figure 6. Democratic Incumbents 1.00 Republican Incumbents Projected Vote Share 0.75 0.50 0.25 0.00 −1.5 −1.2 −0.9 Incumbent Position −0.6 −1.5 −1.2 −0.9 Incumbent Position −0.6 Figure 6: Simulated Vote For The Incumbent Under Normal Utility: The y-axis is probability of voting for the incumbent, and the x-axis is the incumbent’s ideal point. Probabilities are generated by a simulated normal utility model. Each line represents the vote probability for a set of voters with a given ideology. Voter ideology ranges from -1.5 to 1.5, with voters at -1.5 in dark blue and voters at 1.5 in dark red. Each line in between represents one of 11 separate categories of ideology, colored by closeness to 1.5 or -1.5. Figure 6 is similar to some of the quadratic utility specifications that are encapsulated in 41 our model, particularly for voters in the middle of the ideological spectrum. However, due to greater indifference at the extremes, extreme voters are “bunched” at the top and bottom of the graph in a way that is not consistent with quadratic utility, although not far off. In order to simulate a situation that is much more inconsistent with quadratic utility, consider the following “high acuity” normal utility vote generating process: 1 1 P (yij = 1) = Φ(exp(− δ(xi − rj )2 ) − exp(− δ(xi − cj )2 )) 2 2 Here δ is an arbitrary multiplier. δ greater than 1 produces greater utility sensitivity to different distances between x, r, and c. Hence the term “high acuity.” Figure 7 shows the effect of simulating this model when δ = 2. Democratic Incumbents Republican Incumbents 1.00 Projected Vote Share 0.75 0.50 0.25 0.00 −1.5 −1.2 −0.9 Incumbent Position −0.6 −1.5 −1.2 −0.9 Incumbent Position −0.6 Figure 7: Simulated Vote For The Incumbent Under “High Acuity” Normal Utility: The y-axis is probability of voting for the incumbent, and the x-axis is the incumbent’s ideal point. Probabilities are generated by a simulated normal utility model with a “high acuity” multiplier. Each line represents the vote probability for a set of voters with a given ideology. Voter ideology ranges from -1.5 to 1.5, with voters at -1.5 in dark blue and voters at 1.5 in dark red. Each line in between represents one of 11 separate categories of ideology, colored by closeness to 1.5 or -1.5. The striking feature of Figure 7 is that the most conservative and most liberal voters are actually more likely to cross party lines than more moderate voters. This is starkly inconsistent with our non-parametric results. More generally, neither version of normal 42 utility simulated here is at all close to the results we present in the body of the paper. Low Acuity or High Indifference There is one way we can think of in which spatial voting can be reconciled with the results presented in this paper. That is, if voters vote spatially, but with very low sensitivity to legislator positions. This low sensitivity could be part of the utility function: voters may merely be indifferent between many different possible positions. It could also be caused by large errors, or an inability to distinguish different positions taken by legislators from one another. For instance, one possibility is that voters vote on the basis of their own ideology, but also have a weak spatial voting component to their utility. Say for instance that votes are generated by the following process: 1 1 (xi − rj )2 − (xi − cj )2 ) 10 10 1 1 2 1 1 = Λ(xi Ia − xi (1 − Ia ) − xi rj + rj + xi cj − c2j ) 5 10 5 10 P (yij = 1) = Λ(xi Ia − xi (1 − Ia ) + Where Ia is an indicator for whether xi and rj have the same sign. The direct effect of xi is a much larger component of this process than the effect of utility differentials between challengers and incumbents. Figure 8 shows the result of simulating this data generating process. The graphs are not unlike the results in our paper, with big differences between voters at different positions, and very little slope in the lines. This should not come as too much of a surprise, since the parameters we have specified are similar to the parameters we estimate in the paper. The point is that spatial voting is occuring in this example, it just plays a relatively small role in utility. Therefore our results do not disprove the theory of spatial voting per se. 43 Democratic Incumbents Republican Incumbents 1.00 Projected Vote Share 0.75 0.50 0.25 0.00 −1.5 −1.2 −0.9 Incumbent Position −0.6 −1.5 −1.2 −0.9 Incumbent Position −0.6 Figure 8: Simulated Vote For The Incumbent Under Very Low Acuity Quadratic Utility: The y-axis is probability of voting for the incumbent, and the x-axis is the incumbent’s ideal point. Probabilities are generated by a simulated quadratic utility model with a large direct effect of ideology and a very small utility difference for spatial proximity. Each line represents the vote probability for a set of voters with a given ideology. Voter ideology ranges from -1.5 to 1.5, with voters at -1.5 in dark blue and voters at 1.5 in dark red. Each line in between represents one of 11 separate categories of ideology, colored by closeness to 1.5 or -1.5. 44 Online Appendix B: Electoral Accountability on Important Votes The evidence in the body of the paper indicates that there is a very modest association between the roll call positions of legislators and most citizens’ votes. However, it is possible that we are looking in the wrong place. Perhaps there is a greater association between legislators’ roll call positions and citizens’ decisions at the ballot box for roll call votes on important roll calls. For instance, a number of recent papers suggest that Democrats lost their majority in 2010 because of their votes on salient roll calls such as the Affordable Care Act and the American Clean Energy and Security Act (Brady, Fiorina, and Wilkins, 2011; Nyhan et al., 2012). We examine this hypotheses using data on “Key Votes” from the Congressional Quarterly to scale legislators’ ideal points on important votes using Simon Jackman’s IDEAL package in R. There are between 20-30 Key Votes in each Congress (Table 7). The overall correlation between these ideal points on important votes and DW-Nominate scores is approximately .9. However, there is a weaker correlation with legislators’ Nominate scores within each party. In any given year, the within-party correlation between legislators’ scaled roll call positions on important votes and DW-Nominate scores varies between 0.3 and 0.7. Congress 108 109 110 111 112 Key Votes 23 25 23 27 24 Table 7: Number of CQ Key Votes by Year Next, we examine whether the roll call positions of legislators on important votes affect whether survey respondents supported their incumbent legislator on election day. Similarly to our analysis in the previous section, we find only limited evidence for electoral accountability 45 for roll call positions. Table 8 shows the implied change in the probability that constituents will support a legislator that moderates their position by roughly one standard deviation on important roll call votes. The figure shows that there is a relatively modest relationship between the roll call positions of legislators on important votes and citizens’ vote choice. More importantly, only six of twelve coefficients are statistically significant in the direction predicted by spatial voting theory. Moreover, four coefficients are statistically significant in the opposite direction predicted by spatial voting theory. As a result, there is no evidence to support the hypothesis that voters hold legislators more accountable on important votes than on other votes. Democratic Incumbent Liberal voter Mod. voter Conserv. voter Candidate-Centered Theoretical Prediction (H1) Republican Incumbent Liberal voter Mod. voter Conserv. voter ambiguous + + + + ambiguous Democratic voter 2.56%** (.07) -0.22% (.37) -27.03%** (3.28) -19.91** (.29) -2.21%* (1.17) 21.97%** (2.10) Independent voter -.37% (.24) 2.51%** (1.80) 1.97%** (0.41) -16.71%** (1.28) -1.31% (2.01) 5.52%** (0.28) Republican voter -3.25%* (1.77) 2.64%* (1.45) 1.32%** (0.14) 9.09%** (3.00) 1.92%** (0.54) -0.23%** (0.04) Table 8: Effect of incumbent moderating their ideal point by one standard deviation on important votes on the probability that citizens vote for the incumbent in congressional elections. Effects that are significant at the .05 level are shown with two stars and effects that are significant at the .1 level are shown with one star. 46 Online Appendix C: Electoral Accountability in Competitive Districts Another possibility is that competitive elections are required to facilitate democratic accountability (Griffin, 2006). This hypothesis implies that we should see strong evidence of spatial voting in electorally competitive districts. To examine whether voters hold legislators accountable for their roll call positions in electorally competitive districts, we first narrow the range of districts in our analysis using data from Bonica (2013). We define competitive districts as those where the challenger raised at least $500,000 in campaign contributions. Note, however, that our findings are not sensitive to our definition of competitive districts. We find similar results using other definitions of competitive districts. Democratic Incumbent Liberal voter Mod. voter Conserv. voter Candidate-Centered Theoretical Prediction (H1) Republican Incumbent Liberal voter Mod. voter Conserv. voter ambiguous + + + + ambiguous Democratic voter 0.02% (0.13) -1.15% (0.89) -6.46% (5.44) 0.55** (0.22) 2.47%** (1.15) 2.48% (6.10) Independent voter 2.65%* (1.55) 9.27%** (4.52) 0.83% (0.59) 1.02% (0.95) 0.51% (3.57) -0.68% (0.70) Republican voter 13.69%** (6.21) 9.13%** (2.01) 0.68%** (0.14) -11.01%* (5.89) -3.11%** (1.29) 0.14% (0.14) Table 9: Effect of incumbents in competitive elections moderating their ideal point by one standard deviation on the probability that citizens vote for the incumbent in congressional elections. Effects that are significant at the .05 level are shown with two stars and effects that are significant at the .1 level are shown with one star. Do the roll call positions of legislators in competitive districts affect whether survey respondents supported their incumbent legislator on election day? Similarly to our analysis in the previous sections, we find only limited evidence of electoral accountability for roll call positions. Table 9 shows the implied change in the probability that constituents will support a legislator that moderates their position by roughly one standard deviation. Similar to our 47 main results, the figure shows that there is only a modest relationship between the roll call positions of legislators in competitive races and citizens’ vote choice. 48 Online Appendix D: Survey Questions Table 10 shows the complete list of survey questions that we used to jointly scale respondents from the 2006-2013 Cooperative Congressional Election Studies. The questions are a mix of items from the common content and modules that we created. Note that we only used respondents from the 2006, 2008, 2010, and 2012 CCES surveys in this study because the odd numbered year respondents do not indicate congressional vote choice. Table 10: Survey Question Text Variable Survey Question Text v2072 v2092 v2103 v3019 v3022 v3024 v3027 v3060 v2102 v3063 v2101 v3069 v3072 v3075 v3066 v3078 q34 cc06 v2072 cc06 v2092 cc06 v2103 cc06 v3019 cc06 v3022 cc06 v3024 cc06 v3027 cc06 v3060 cc06 v3063 cc06 v3075 cc46 cc06 v3078 cc34 cc38 cc12x 5 cc310 cc311 cc312 cc313 cc316b cc316c cc316e cc316a cc316f cc316g cc316d cc316h cc316i cc417 cc422 CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES Raise minimum wage to $7.25 Should we take action on climate change? Amendment banning gay marriage When should abortions be allowed? Climate change is real Social security privatization Affirmative action for discriminatory companies Ban late-term abortion Expand funding for stem cell research Expand funding for stem cell research Path to citizenship or strict enforcement Citizenship opportunity for illegal immigrants Favor/oppose raising minimum wage Extend capital gains tax cuts Withdrawing troops from Iraq Free trade agreement with Central America Support state voter ID laws Raise minimum wage to $7.25 Should we take action on climate change? Amendment banning gay marriage When should abortions be allowed? Protect environment over jobs/economy Social security privatization Affirmative action for discriminatory companies Ban late-term abortion Expand funding for stem cell research Extend capital gains tax cuts Withdrawing troops from Iraq Free trade agreement with Central America Expand SCHIP - health care for children Surveillance of foreigners in US Build a wall between US and Mexico When should abortions be allowed? Protect environment over jobs/economy Social security privatization Affirmative action for discriminatory companies Raise minimum wage to $7.25 Expand funding for stem cell research Fund health insurance for children Withdrawing troops from Iraq Support/oppose amendment banning gay marriage Federal assistance for housing crisis Eavesdrop overseas without court order Extend NAFTA to Peru & Columbia U.S. government bank bailout Government guaranteed health insurance Carbon tax to reduce emissions 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2007 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 49 Survey Text 10 Continued from previous page Variable Survey Question Text cc419 6 cc09 51 cc09 54 cc09 53 cc09 55 cc09 59a cc09 59b cc09 59c cc09 59d cc09 59e cc09 59f cc09 59g sta302 1 sta302 2 sta302 3 sta302 4 sta302 5 sta302 6 sta302 7 sta302 8 sta303 1 sta303 2 sta303 3 sta303 4 sta303 5 sta303 6 sta303 7 sta304a sta304c sta304d sta304e sta304f sta305 1 sta305 2 sta305 3 sta305 4 sta305 5 sta306 1 sta306 2 sta306 3 sta306 4 307a 307b 307c 307d sta312 sta314 sta315 sta317 sta319 sta320 sta321 sta322 sta360a sta360b sta360c sta360d sta360e sta360f sta360g sta360h sta360i sta360j sta361a sta361b CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES Require photo ID to vote Take action against global warming Amendment banning gay marriage When should abortions be allowed? Affirmative action for discriminatory companies Lilly Ledbetter Fair Pay Act Hate Crimes Act - include LGBT American Recovery & Reinvestment Act Expand SCHIP - health care for children Renewable energy funding, carbon caps Require health insurance Appoint Sotomayor to Supreme Court Increase funding for job training programs Reduce government regulation Employers should offer childcare Increase minimum wage Support workers right to unionize Eliminate federal unemployment programs Include sexual orientation in anti-discrimination laws Include gender in anti-discrimination laws Universal healthcare Expand tax-free medical savings accounts Allow importation of prescription drugs Expand Medicare prescription drug coverage Tax credits to offset insurance costs Expand child healthcare programs Providing healthcare is not responsibility of government Allow same-sex marriage Funding for stem cell research (existing) Funding for stem cell research (new embryos) Affirmative action for federal contractors Continue federal affirmative action programs Private social security accounts Increase payroll tax to ensure social security viability Decrease benefits to retirees to ensure social security viability Increase social security benefits with cost of living Raise the retirement age to ensure social security viability Require welfare recipients to work Federal block grants for welfare Housing assistance for welfare recipients Abolish federal welfare programs Public health insurance option Monetary limits in malpractice lawsuits Require balanced federal budget Government funds to stimulate economy Free trade agreement with Central America Expand funding for stem cell research Citizenship opportunity for illegal immigrants Affirmative action for discriminatory companies Path to citizenship or strict enforcement Increase minimum wage Extend capital gains tax cuts Amendment banning gay marriage Eliminate the minimum wage Government guarantee standard of living No taxes for low-income families Prohibit incomes above $1 million Eliminate food subsidies for children Tax rate the same for rich and poor No government assistance for low-income Government should provide universal jobs Rich should pay higher tax rate than poor Minimum wage should be $15/hour Ban some high-fat foods from restaurants Government standards for prescription drugs 2008 2009 2009 2009 2009 2009 2009 2009 2009 2009 2009 2009 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module 50 Survey Text 10 Continued from previous page Variable Survey sta361c sta361d sta361e sta361f sta361g sta361h sta361j sta361k sta361l sta361m sta361n sta362a sta362b sta362c sta362d sta362e sta362f sta362g sta362h sta362i sta362j sta362k sta370a sta370b sta370c sta370d sta370e sta370f sta380a sta380b sta380c sta380d sta380e sta380f sta380g sta380h sta381a sta381b sta401a sta401b sta401c sta401d sta401e sta401f sta401g sta402 sta403a sta403b sta403c sta403d sta403e sta405a sta405b sta406a sta406b sta411c sta411d sta411e sta411f sta412 sta413 sta430a sta430b sta430c sta430d CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 Question Text Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module All public buildings accessible to handicapped Government-enforced nutrition standards No limits on pollution from businesses Government-enforced advertising standards All motorcyclists required to wear helmets Ban sale of energy-inefficient appliances Privatize the Post Office Military burden shifted to private contractors Government takeover of bad companies Require power plants to reduce emissions Require residential carbon monoxide detectors Hold BP executives liable for oil spill Require public schools to teach creationism Limit ATM fees to $1 Eliminate Environmental Protection Agency Deport all illegal immigrants Grant all illegal immigrants citizenship End subsidies for green energy Government-funded high-speed railroad Felons should have right to vote Prohibit construction of 9-11 site mosque Ban late-term abortion procedures Require business-provided health insurance Require all people buy health insurance Limit damages in malpractice lawsuits Medical experts decide which tests insured Patients pay more for “ineffective”” treatments Public insurance entity for low-cost insurance Government funds to insure all children Right of patients to sue HMO Make it harder to obtain abortion Allow the death penalty for some crimes Require license to purchase handgun Allow gays to serve in military Federal law to allow school prayer Flat tax law for all Americans Eliminate regulations for businesses Protect environment/natural resources Government help insure all children Government help employers pay for insurance Eliminate the estate tax Social Security privitization Easier for labor unions to organize Federal funding for stem cell research Extend federal ban on assault weapons Same-sex marriage in your state Increase the minimum wage Government reduce income inequality Government reduction of federal taxes Government vouchers for private school Amendment banning gay marriage Increase federal funding to public school Government-funded universal health care Should the government restrict immigration? Should the government restrict gun sales? Health insurance for low-income children Assist homeowners facing foreclosure Extend NAFTA to Peru & Columbia U.S. government bank bailout Carbon tax to reduce emissions Guaranteed universal health insurance Housing vouchers for homeless Maintain welfare-to-work requirements Provide food stamps to legal immigrants Continue Medicaid for welfare-to-work 51 Survey Text 10 Continued from previous page Variable Survey sta430e sta430f sta430g sta430h sta430i sta430j sta430k sta450 sta451 sta460a sta460b sta460c cc324 cc325 cc326 cc327 cc332a cc332b cc332c cc332d cc332e cc332f cc332g cc332h cc332i cc332j cc321 cc341a cc341b cc341c cc341d cc341e cc341f cc341g cc341h cc354 cc353 cc352 cc351 1 cc351 2 cc351 3 hsu301 hsu302 hsu303 hsu304 hsu305 hsu306 hsu310 hsu311 hsu312 hsu313 hsu314 hsu320 hsu321 hsu322 hsu323 hsu324 hsu325 hsu326 hsu327 hsu328 hsu329 hsu330 hsu360 hsu361 CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2010 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 Question Text Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Federal poverty aid through religious orgs. Additional funding for state Medicaid Tax credits for businesses with childcare Federal aid for states with more immigrants Prohibit state laws denying immigrations services Increase quota for skilled immigrants Collect fingerprint data from visa applicants Federal income tax level Support same-sex marriage Path to citizenship for immigrants Increase border security with Mexico Drivers licenses for undocumented immigrants When should abortions be allowed? Protect environment over jobs/economy Amendment banning gay marriage Affirmative action for discriminatory companies American Recovery & Reinvestment Act Expand SCHIP - health care for children Renewable energy funding, carbon caps Require health insurance Appoint Kagan to Supreme Court Financial Reform Bill End Don’t Ask, Don’t Tell Overseas surveillance of foreigners Federal funding for stem cell research U.S. government bank bailout Belief in climate change American Recovery & Reinvestment Act Expand SCHIP - health care for children Renewable energy funding, carbon caps Require health insurance End Don’t Ask, Don’t Tell Overseas surveillance of foreigners Federal funding for stem cell research U.S. government bank bailout Affirmative action for discriminatory companies Amendment banning gay marriage When should abortions be allowed? Citizenship opportunity for illegal immigrants Increase patrols of U.S.-Mexico border Allow police to question suspected immigrants Guaranteed universal health insurance Protect right of workers to unionize Government reduce income inequality Reduce regulation of private sector Raise minimum wage to $10 Allow corporations unlimited campaign contributions Allow same-sex marriage Allow LGBT to legally form civil unions Ban or limit contraceptive use Ban sex between persons of same gender Require 24-hour waiting period for abortion Raise taxes a few hundred dollars Raise taxes on rich ($250,000+/year) Reduce tax break for homeowners Make retirees pay for Medicare Increase capital gains taxes Increase taxes on corporations Reduce Medicaid benefits for low-income Eliminate student loan subsidies Reduce federal worker pensions Make deep cuts in defense spending Increase retirement age to 68 Fine businesses that hire illegal immigrants Allow states to deport illegal immigrants 52 Survey Text 10 Continued from previous page Variable Survey hsu362 hsu363 hsu364 hsu365 hsu367 hsu370 hsu371 hsu372 hsu373 hsu374 hsu375 hsu376 hsu377 hsu378 hsu379 hsu380 hsu381 cc350 ucm301 ucm302 ucm303 ucm304 ucm305 ucm306 ucm307 ucm321 ucm322 ucm323 ucm324 ucm325 ucm326 ucm327 ucm328 ucm329 ucm330 ucm331 ucm332 ucm333 ucm370 ucm371 ucm372 ucm373 ucm374 ucm375 ucm376 ucm377 ucm378 ucm379 ucm380 ucm381 ucm382 ucm401 ucm402 ucm403 ucm404 ucm405 ucm406 ucm407 ucm408 cc321 cc324 cc325 cc327 cc326 cc332a CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2011 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 Question Text Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Module Allow police to ask for immigration documents Deport all illegal immigrants Remove fence on border with Mexico Same treatment of Mexican & Canadian immigrants Allow states to admit immigrants Federal government should protect environment Require power plants to reduce emissions Eliminate the Environmental Protection Agency Require 10% electricity renewable statewide Require 25% electricity renewable statewide Government should protect endangered species States should set pollution limits States should keep waterways clean Support coal plant within 25 miles of home Support wind power plant within 25 miles of home Support oil/gas drilling within 25 miles of home Power plants near home should be regulated Should we take action on climate change? Guaranteed universal health insurance Protect worker right to unionize Government reduce income inequality Reduce regulation of private sector Raise the minimum wage to $10 Allow corporations unlimited campaign contributions Allow drilling in Alaskan Wildlife Refuge City should provide health benefits to same-sex partners Reduce greenhouse gas emissions in city Subsidize mass transit for low-income in city Subsidies for residential solar energy in city Ban smoking in local bars/restaurants in city Require local residents to recycle in city Reduce pension for government employees in city Tax breaks to incentivize businesses to move in city Limit how much landlords can raise rent in city Offer subsidized housing to homeless in city Eliminate tenure for school teachers in city Close city parks to save money Close city libraries to save money Require parental permission for teen abortion Require 24-hour waiting period for abortion Require photo ID to vote Legalize casino gambling in states State law capping property taxes Take away union right to bargain Allow LGBT to legally form civil unions Allow same-sex marriage In-state tuition for illegal immigrant graduates If your state opted out of Medicaid expansion Allow death penalty for convicted murderers Require waiting period for gun purchases Raise the minimum wage to $8 Set limits on CO2 emissions Require 10% electricity renewable statewide Require 25% electricity renewable statewide State gasoline tax less than $0.25/gallon Renewable energy tax on electricity bill Require more efficient use of electricity Set limits on CO2 emissions State should prepare for climate change Should we take action on climate change? When should abortions be allowed? Protect environment over jobs/economy Affirmative action for discriminatory companies Amendment banning gay marriage House Budget plan - cut Medicare/Medicaid 53 Survey Text 10 Continued from previous page Variable Survey Question Text cc332b cc332c cc332d cc332e cc332f cc332g cc332h cc332i cc332j cc322 1 cc322 2 cc322 3 cc322 4 cc329 cc332a cc332b cc332c cc332d cc332e cc332f cc332g cc332h cc327 cc328 cc330 cc325 cc13 320a cc13 320b cc13 320c cc13 320d cc13 320e cc326 1 cc326 2 cc326 3 cc326 4 CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES CCES Simpson-Bowles plan - 15% cuts Middle Class Tax Cut Act Tax Hike Prevent Act Religious exemption for birth control coverage Free trade agreement with Korea Repeal Affordable Care Act Approve Keystone XL pipeline Support ACA - required health insurance Allow Gays in the Military Citizenship opportunity for illegal immigrants Increase patrols of U.S.-Mexico border Allow police to question suspected immigrants Fine businesses that hire illegal immigrants Allow same-sex marriage Prohibit abortions after 22nd week Simpson-Bowles plan - 15% cuts Repeal Affordable Care Act Approve Keystone XL pipeline Allow internet sales to be taxed Violence Against Women Act Block NSA collection of phone records Decentralize education decision-making When should abortions be allowed? Protect environment over jobs/economy Affirmative action for discriminatory companies Should we take action on climate change? Background check for all gun sales Prohibit publication of names of gun owners Ban high-capacity gun magazines Ban assault rifles Easier to apply for concealed-carry permit Citizenship opportunity for illegal immigrants Increase patrols of U.S.-Mexico border Allow police to question suspected immigrants Fine businesses that hire illegal immigrants 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2012 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 2013 54
© Copyright 2026 Paperzz