Issue Voting in U.S. House Elections

Issue Voting in U.S. House Elections | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Issue Voting in U.S. House Elections Richard R. Lau, Amy S. Funck This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8544888/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Representative democracy is based on the idea that citizens choose the candidates who most closely share their policy views and concerns. There is surprisingly little evidence for this idea from elections for the House of Representatives, however, primarily due to a lack of appropriate data. We overcome the data problems by employing the large samples from CCES surveys, and coding candidates’ policy positions from the campaign webpages of 394 major party candidates between 2010 and 2020. We find vote choice in House elections is strongly a function of agreement with the candidates on the issues, controlling for many alternative explanations that prior research has found to be important. We test six hypotheses about when issue voting should be more or less important. We know of no prior research on congressional elections that has comparably detailed evidence that clearly demonstrates that congressional elections provide citizens with considerable levels of substantive representation. Issue voting House elections Congressional elections democratic representation information processing cognitive limitations Figures Figure 1 Figure 2 Figure 3 Introduction Representative democracy is based on the idea that citizens choose the parties or candidates who most closely share their own policy views and concerns, and then monitor the behavior of winning candidates once they are in office to make sure they do what they said they would do. Such substantive representation is not the only type of representation – see for example Mansbridge, 2003 – but it is the easiest to justify normatively, and it is what most people think democratic representation is about. In the U.S., almost all research on issue voting involves presidential elections (e.g., Highton, 2010 ; Jesse, 2009 ; Petrocik, 1996 ), but the principle applies just as strongly (and even more directly) to legislative elections where the candidates being selected actually vote on the country’s laws. Although congressional voting has long been a major focus of political science research, we have surprising little direct evidence about the decisions of individual voters in House elections, and almost no evidence on the extent to which differential policy agreement with competing candidates affects the vote choice. Thus, the degree of substantive democratic representation provided by legislative elections in the U.S. is more a matter of faith than hard empirical evidence. The problem is one of data availability. On the one hand, almost no high-quality representative survey is conducted within a single – or even a few – legislative districts. Most election surveys focus on the presidential election and aim to be representative of the country as a whole. In the U.S., a nationally representative survey with 1000 respondents includes, on average, a little more than two respondents from each congressional district. It takes a survey with an extraordinarily large initial sample size before this first problem can be addressed. Even if the sample size problem can be overcome, a second and equally tricky problem immediately arises. To measure issue voting, you need estimates of the policy preferences of a set of voters in the district and estimates of the policy preferences on the same issues (measured on approximately the same scales) of the multiple candidates running for election in that district. The more elections you want to study, and the more issues you want to consider, the greater this second problem becomes. We overcome the first problem by utilizing the very large samples from the 2010 through 2020 Cooperative Congressional Election Study (CCES) surveys. To address the second issue, we have gathered data on the policy positions of the Democratic and Republican candidates running in 197 competitive House elections across those six recent election years from the campaign websites of those 394 candidates. The conventional wisdom on House voting assumes that the typical citizen knows so little about their incumbent representative – much less the candidate who is challenging her in the upcoming election – that relative policy agreement could not possibly play a large role in the vote decision. Controlling on a set of variables that have been used in prior research on congressional voting, however, we find that vote choice in House elections is strongly a function of relative agreement with the candidates on the issues. We know of no prior research on congressional elections that has comparable detailed data from so many different House elections. In the following section we briefly review the literature on congressional elections, focusing on prior research that has attempted to explore the role of policy agreement in congressional voting. After specifying a set of hypotheses that are derived from that literature, we describe the CCES surveys that we relied upon for individual-level data on the policy positions and vote decisions of voters in House elections between 2010 and 2020, and the procedures we developed to gather and code the policy positions the competing candidates took on those same issues. We then present the results of multilevel hierarchical nonlinear models of the vote choice in recent U.S. House of Representatives elections that are aimed at testing those hypotheses. Discussion focuses on how policy agreement can play such an important role in congressional elections when citizens seem to know so little about the candidates running in those elections, and the implications of our findings for understanding the true nature of representation in modern large-scale democracies. Literature Review Researchers studying voting in congressional elections have two primary types of extant data readily available to them: aggregate district- and state-level election returns, and individual-level responses from nationally representative surveys. There are several important facts about the vote decision in congressional elections that are readily revealed by aggregate-level election returns. First, turnout in off-year elections is always less than turnout in presidential election years – usually around 15 percent less. Even in presidential election years, 3 or 4 percent of citizens who vote for president will not bother to cast a vote for the House of Representatives. Clearly, the public is far more interested in the presidential election than in the typical House election. Second, there are several “nationalized” factors that increasingly influence congressional elections (see for example, Abramowitz and Webster, 2016). To begin, winning presidential candidates often have “coattails” that advantage congressional candidates from their party, increasing their vote totals by 3 or 4 percent. In addition, congressional elections are in part a referendum on the sitting president. When the president is popular, House candidates from his party usually have a significant advantage over candidates from the out-party. When the president is unpopular, House candidates from his party suffer. These three factors combine such that House candidates from the president’s party typically suffer a significant penalty during off-year elections. Although this number bounces around a fair amount from election to election, on average the president’s party will lose about 25 seats in the House of Representatives during off-year elections. Finally, despite all these influences from national politics, incumbents are usually re-elected. Since 1946 over 90 percent of all House elections involved an incumbent seeking reelection, and of those, over 90 percent were successful. See Aldrich, Carson, Gomez, and Merolla (2022), or Jacobson and Carson ( 2020 ), for recent summaries of this data. None of these insights about House elections, however, are based on the policy positions of any of the candidates involved in those elections. Individual-level survey data on voting in House elections has provided somewhat fewer insights into citizens’ voting behavior, in large part because of data limitations. As mentioned above, the typical political survey has a national focus, with relatively few respondents from any particular congressional district, and relatively few questions about the House election. Large majorities of Americans support the House candidate from their political party – as is the case for every other partisan election in the country – with most defections from party explained by incumbency. But again, none of this says anything about issue voting per se, and thus the extent of substantive representation that is provided by House elections. Indeed, the conventional assumption in political science research on congressional voting – going back to the seminal research of Miller and Stokes ( 1963 ) – is that most people know little about the policy preferences of their incumbent representative, and considerably less about the policy views of any challenger. If this view is correct, it is hard to imagine that policy agreement would have much to do with vote choice in House elections. As summarized by Warsaw in a recent article, “electoral accountability” [at the local, congressional, and gubernatorial levels] “is generally weak and contingent on various institutional moderators” (Warsaw, 2019, p. 462). On the other hand, as first noted by Mann and Wolfinger ( 1980 ), while less than half of all Americans can recall the name of their representative, over 90 percent will recognize the incumbent’s name from a short list (as they might have on a ballot), with somewhat more than half recognizing the major party challenger. The quality of the challenging candidate also matters, with candidate quality usually defined by prior electoral experience. This knowledge is not coming out of thin air. In our view, voters are working with more information about the major party candidates running in congressional elections than is assumed by conventional wisdom, it just may not be readily available for recall. Do Issues Matter in House Elections? Given the dearth of appropriate individual-level survey data, several research teams have tried to address this problem indirectly with aggregate-level data, which are readily available and stretch over longer periods of time. Both Canes-Wrone, Brady, and Cogan (2002), and Ansolabehere and Kuriwaki (2022), find that House incumbents are held accountable for being ideologically out of step with their constituents. On the other hand, Ansolabehere, Snyder, and Stewart (2001), and Highton ( 2019 ), conclude that members of Congress are not much constrained by voters in the positions they take on particular issues. The conflicting results reported by this prior literature using aggregate election results are, we suspect, a function of the indirect nature of their underlying research designs and the number of assumptions that must be made to infer individual vote decisions from aggregate election outcomes. True democratic representation is served when voters hold their representatives accountable for their actions in Congress and vote them out of office if they disagree with too many of them. This is retrospective accountability (Fiorina, 1981 ), looking backwards at the prior actions/performance of an incumbent in office. But this is only half of the story, and the smaller half at that. How were those incumbents elected in the first place? And how will voters select their replacements, and/or decide whether any available alternative is superior to the status quo incumbent? Prospective issue voting occurs when decision makers compare the policy positions offered by one candidate or party to the policy positions of alternative candidates or parties – not the imagined or presumed opinions of some hypothetical opponent but the actual opinions of competing candidates. Mansbridge ( 2003 ) calls this promissory representation. To explore issue voting in congressional elections, we need information on the policy preferences of a set of voters and on the policy stands of a set of competing candidates.[1] We have found only three earlier analyses of actual House elections that have the type of evidence needed. Using data from a 1966 SRC (i.e., pre-ANES) survey and candidate position data gathered by NBC News, Wright (1978) created domestic policy conservatism scores by combining available policy items in each of these surveys, and then calculating each respondent’s relative closeness to the Democratic and Republican House candidates. Although the statistical tools available to Wright were somewhat primitive by today’s standards, he reports a statistically significant impact of issue proximity on vote choice, controlling on party identification and incumbency – albeit one that was considerably weaker than the effects of either party ID or incumbency. We worry about the lack of statistical controls in this observational data, and the fact that 622 survey respondents are distributed across 117 congressional districts – on average, 5 respondents per district. Shor and Rogowski ( 2018 ; see also Rogowski, 2016 ) have much larger surveys, and modern statistical techniques, to bring to bear on this issue. These authors used data from the 2008 Cooperative Campaign Analysis Project (CCAP) and the common content of the entire 2010 CCES survey to measure voter’s policy preferences, and data from the National Political Awareness Test (NPAT) to estimate candidates’ policy positions. The latter is a survey administered by the nonpartisan group Project Vote Smart and sent to announced-candidates for all federal and state offices in an effort to provide voter’s with easy, unbiased information on candidate’s policy positions.[2] Controlling for party identification and incumbency, the authors find strong evidence that their measure of relative issue closeness to the Democratic and Republican candidates had a statistically significant and substantively meaningful effect on the vote choice. This is far and away the strongest direct evidence that issue voting plays an important role in the vote choice for House of Representatives elections. In our view, it suggests a necessary revision of conventional wisdom about the nature of accountability provided by congressional elections. The best evidence we have clearly suggests that policy agreement could be an important part of the vote calculus of many Americans, even for elections well below the top of the ballot. We hope to extend and improve upon Shor and Rogowski’s research in several important ways. We utilize CCES survey data extending across eleven years and six federal elections – three off-year elections, and three presidential year elections. We have a very different method for measuring candidates’ issue stands, one that does not rely upon both major party candidates publicly completing a survey of their policy positions, and one that can be replicated in future elections. Following Buttice and Stone’s (2012) counsel, we include a measure of the relative quality of the competing candidates in our analysis. We have a considerably larger average number of respondents per congressional district (in our case, 133; Shor and Rogowski’s means were only 8 in their 2008 data, but a much more robust 83 in their 2010 election data). We also employ a multilevel statistical procedure as our main analytic tool, thus providing much more reliable estimates of the importance of aggregate-level factors such as incumbency and candidate quality. Before proceeding we want to be very clear about one important factor. Although we are primarily interested in the causal effect of issue agreement on the vote choice, we – as Wright (1978), and Shor and Rogowski ( 2018 ), before us – only have correlational data, and thus any causal inferences we would like to make are open to omitted variables bias. The multivariate analysis that follows controls on a host of plausible moderating and mediating variables, but of course not every conceivable one, and thus the causal inferences one can draw from our results are suggestive at best. There is a growing body of recent experimental research (particularly involving conjoint designs) that provides strong causal evidence of the effects of policy agreement on hypothetical vote decisions (e.g., Alvarez and Morrier, 2025; Costa, 2021 ; Griffin, Newman, and Nickerson, 2019 ; Henderson et al., 2022 ; Mummolo, Peterson, and Westwood, 2021), although as is always the case with experimental data, the external validity of the reported results is somewhat suspect. Hypotheses One primary hypothesis drives all analysis in this paper. H 1 : Controlling for a host of aggregate- and individual-level factors that past research has found to be important, greater agreement with one of the competing candidates on a set of policy issues – that is, relative issue agreement – will be associated with a higher likelihood of voting for that candidate in elections for the House of Representatives. We also test several hypotheses suggested by prior research about the contingent nature of issue voting in House elections. Most of these hypotheses assume that issue voting is cognitively difficult, and factors that reduce the difficulty, or motivate greater effort, can increase the importance of policy considerations in the vote choice. According to Shor and Rogowski ( 2018 ), for example, the importance of issue agreement in the vote decision will be greater when the candidates have jointly spent a great deal of money than when relatively little has been spent of the campaign. The logic behind this hypothesis is that campaign spending – presumably much of it about policy – makes it easier for voters to learn the issue stands of the competing candidates. A somewhat more nuanced hypothesis realizes that it takes two to tango, and awareness of the policy positions of one candidate but not the other does not encourage true issue voting. Either way: H 2 : The effect of issue agreement on the vote choice is conditional on the magnitude (and/or equality) of campaign spending in the election. The cross-level interaction between campaign spending and relative issue agreement should be positive. Another factor that should make issue voting easier is when the competing candidates take ideologically distinct positions. According to Buttice and Stone (2012), when candidates hold dramatically different policy views, voters should not only be more aware of those differences and should also care more about the outcome of the election, which could motivate the effort required for issue voting. H 3 : The importance of policy agreement in the vote decision will be greater when the two candidates are clearly ideologically distinct than when they have relatively similar ideologies. Thus the cross-level interaction between the ideological distinctiveness of the two candidates and relative issue agreement should be positive. On the other hand, Buttice and Stone argue, the greater the quality (experience) difference between the candidates, the lower the need and motivation to seek out policy-based information, and thus the weaker the relationship between issue positions and the vote. H 4 : Policy agreement will be less important when the quality/experience difference between the candidates is greater. Therefore the cross-level interaction between the candidate quality differences and relative issue agreement should be negative. Another factor that should motivate greater cognitive effort is the expected competitiveness of the election. If the election has a foregone conclusion – which is the case for many (if not most) House elections – why go to the effort of studying the competing candidates to figure out which one best represents my values and concerns? But if both candidates have a realistic chance of winning, expending the cognitive effort to learn about the candidates becomes much more rational. Hence: H 5 : The importance of policy agreement in the vote decision will be greater when the election is expected to be competitive, but less when one candidate has little or no chance of winning. Thus the cross-level interaction between the expected competitiveness of the election and relative issue agreement should be positive. Finally, to the extent that party identification provides a valuable heuristic that can greatly simplify decision making, particularly relative to the effort required to compare candidates on their policy promises, we would expect issue agreement to have the strongest effect on the vote decision among (“pure”) independent-independents who have no partisan leanings to rely upon. Hence: H 6 : The effect of relative issue agreement should be strongest when identification with either of the two major parties is absent – that is to say, among political independents. Method Sample Selection To begin exploring these questions, we utilize the common content from Cooperative Congressional Election Study (CCES) data ( https://cces.gov.harvard.edu/ ), pre-post online survey panels administered by YouGov that have been conducted during every federal election since 2006. From our perspective, the unbeatable advantage of the CCES surveys is their large sample size, upwards to 60,000 or more respondents every election year, providing many statewide samples of 1,000 respondents or more each election year, and many congressional district samples with 100 respondents or more. A shortcoming of the CCES data is that while the various samples are designed to be representative of the nation as a whole and of each state, the data are not perfectly representative at the congressional district level. In an ideal world they would be, although no one, to our knowledge, has large-scale data like that. Our goal is to make general statements about the nature of the vote decisions in congressional elections that are at least modestly competitive. That takes random sampling of elections off the table from the get-go. We trust that any small errors in the representativeness of the data from specific districts will cancel out as we combine the data from nearly 200 different congressional elections. To provide some evidence for the representativeness of our selected elections, Fig. 1 displays a simple bivariate plot comparing the actual two-party proportion of votes received by the Republican candidate, to the proportion of reported votes for the Republican candidate from the CCES sample, in each of the selected districts. The CCES data clearly tracks the actual results. On average, the 95 percent confidence interval around each of these points is almost +/- 9 points; only a handful of observed points are farther than that from the regression line. We see no reason to attribute the dispersion of data points to anything other than random sampling error. Figure 1 Plot of Estimated to Actual House Election Results We selected six recent federal election years for study: 2010, 2012, 2014, 2016, 2018, and 2020. Three of those years included presidential elections; three were off-year elections. Within each election year, we divided states into four categories according to how many “higher level” statewide elections were on the ballot: states with neither a gubernatorial nor senate election on the ballot, states with a gubernatorial election but no senate election, states with a senate election but no governor’s election, and states with both a gubernatorial and a senate election on the ballot. In each election year, our goal was to select four states from each of those categories with at least 1,000 respondents in the sample, half of which had reasonably competitive senate or gubernatorial elections (final margin less than 10 points), half of which were less competitive (final margin 10–24 points). Within each of the chosen states, we then wanted to select two congressional districts with 100 or more respondents in the data, one of which had a final vote spread of within 10 percent, and one less competitive, defined as between 11–25 percent. More often than not, we were able to hit those selection targets, but it was impossible in a few cases because rarely are even 10 percent of House elections moderately competitive. After we identified all districts that met our criteria and organized them by type of election, we took a random sample from within each cell to generate our dataset. In the end, we have data from 197 separate congressional elections, evenly distributed across the six federal election years. The number of respondents per district varies between 70 and 218, with a mean just over 133. The total combined sample has data from 26,230 survey respondents, although this number is reduced to 20,501 once nonvoters, supporters of third party candidates, and a few respondents with other missing data are removed. Operationalizing Variables Except for relative agreement with candidate issue positions, all individual-level predictors come from the common core of the CCES surveys. The dependent variable is dichotomous reported vote for the Democratic or Republican House candidate, with vote for the Republican coded high. All nonvoters and respondents who reported voting for a third-party candidate are excluded from the analysis. Voter Policy Preferences To measure issue voting, we first selected 10 policy issues about which there were decent measures in all six CCES surveys. Five of these issues were primarily moral/social issues: abortion, environmental protection, gay rights, gun control, and immigration. Five of the issues were primarily economic issues: domestic spending, deficit reduction, tax policy, health care (ACA/Obamacare), and defense spending. For the 2020 election, we added policy questions about the best way to deal with recovering from the pandemic, and policing reform. Most of the policy items asked in the CCES surveys are dichotomous, focused on bills that had been recently considered in Congress, but there were usually several of them falling into the different policy areas. The particular questions addressing the ten issues were not always the same across the six election years, so we translated the available items into reasonably comparable 7-point Very Liberal to Very Conservative policy stances across the various election surveys. Coding details are provided in the online appendix. Candidate Policy Positions Our biggest methodological challenge in this project was gathering reasonably objective estimates of where the competing candidates stood on the same policies and political concerns that were used to measure the naïve preferences of the voters in our surveys. To do this, we searched candidate webpages from the Library of Congress’s internet archive Wayback Machine to locate policy-relevant statements about these same ten political issues.[3] We then stripped all identifying information from the file, shuffled the data, and had two expert judges independently read the policy statements from the candidates’ webpages, and attribute a position for them on the same 7-point Very Liberal to Very Conservative policy scales that were used to represent respondents’ policy positions. Averaging across the ten issues, the two coders attributed exactly the same position on the 7-point policy scales 69 percent of the time, and had near perfect agreement – differing by 1 point or less – 92 percent of the time. Weighted Cohen’s Kappa averaged .92 across the 10 issues. Whenever their initial independent judgments differed by more than 1 point, the coders met to review the information and try to reconcile their disagreements. The senior author served as the tiebreaker the few instances when the original judges could not reach an agreement. We then took the mean rating of the two expert judges as an objective reading of where the candidates actually stood on each of these ten policy issues. Assuming we were able to estimate policy positions from both a voter and a candidate, issue agreement was calculated via the “directional” procedure developed by Rabinowitz and MacDonald (1989). We use the directional model because it seemed more plausible in situations where voters are unlikely to have much information about the policy positions of the competing candidates other than which side of the ideological fence they are on. However, we replicate all analyses using a measure of issue agreement that is based on a Euclidean proximity model, which produces essentially identical. The section on robustness describes an alternative measure of policy agreement that makes fewer comparability assumptions and also produces very similar results. By calculating issue agreement in this manner, we assume that the very-liberal to very-conservative scales used to measure voters’ policy preferences and candidates’ policy preferences are comparable. To the extent that assumption in not accurate – and we suspect that extent is pretty great – the issue voting variable will be quite noisy. Measurement error should make hypothesis tests more conservative. As is common in studies of issue voting, we then created a summary measure of issue agreement by averaging together the ten issue-specific agreement scores for the Democratic candidate and subtracting it from a similar measure averaging across the ten issue-specific scores for the Republican candidate. Other Variables Turning to the aggregate level, measures of incumbency for both the Democratic and Republican candidates are simple dummy variables that equaled 0 for all non-incumbents but equaled 1 whenever an incumbent was seeking reelection. More refined measures of candidate quality are based on years of experience as a member of the House of Representatives and, counting less, experience in other elected offices. Thus the measures of incumbency, and candidate quality, are going to be highly correlated with each other – so much so that we cannot include them both in the same equation. In most of the analysis presented below we rely on the more nuanced candidate quality variables because they have more variance, but present some analyses with the simpler incumbent dummy variables to demonstrate the robustness of our findings. District ideology is measured by the proportion of the two-party vote obtained by the Republican candidate in the most recent prior presidential election. Off-year penalty equals 1 for all candidates from the incumbent president’s party during off-year elections, and was 0 otherwise. Measures of campaign spending are based on official FEC records and sum across all reported spending by or on behalf of a candidate. Spending imbalance is calculated as the (natural) log of total spending for the Republican candidate minus the log of total spending for the Democratic candidate. Total campaign spending is the log of total spending for the Republican candidate plus the log of total spending for the Democratic candidate. Spending equality is the absolute value of the difference between the log of total spending for the Republican candidate minus the log of total spending for the Democratic candidate, reversed so that greater equality of spending is scored high. The CCES surveys always asked respondents to rate the ideology of major party congressional candidates on a standard 7-point strong liberal to strong conservative scale. We take the mean rating of political experts in every congressional district (those in the top quartile of general political knowledge), as a semi-objective reading of the candidates’ ideologies. The absolute value of the difference between these mean ratings of the two candidates is our measure of candidate ideological distinctiveness. The publicly available Cook Report rates all House elections on a 7-point scale ranging from − 3 (Safe Democrat) to + 3 (Safe Republican). To single out the most competitive elections, we created a dummy variable that equaled 1 whenever Cook rated the election as a “tossup,” and was 0 otherwise. To ease interpretation, all variables have been recoded to have a 1-point range. Descriptive statistics for all variables used in the analysis are presented in Table A- 1 in the online appendix. Results To test our hypotheses, we specify a series of nonlinear multilevel models in which survey respondents are nested within congressional districts. We include fixed effects for election year. Analyses are conducted with the computer program HLM 8.2 (Raudenbush, Bryk, and Congdon, 2022). All tables present the results from the population-average model with robust standard errors. We estimate fixed effects for all individual-level predictors except issue agreement, which is allowed to vary randomly across elections. Thus, the degrees of freedom for all statistical tests involving estimates of issue voting are 196, not 20,095 as they generally are for all other individual-level predictors. To give readers a good sense of the analyses presented throughout this paper, the results of our most basic models of vote choice are shown in Table 1. The aggregate level variables have their predicted sign and are statistically significant. Model 1 includes simple dummy variables noting the party of incumbents seeking reelection, while Model 2 includes instead the more refined measures of candidate experience/quality. Following the norms of this literature, we include a measure of relative candidate campaign spending, although we readily acknowledge (as have most others before us) that candidate spending is endogenous to the likely outcome of the election (e.g., Green and Krasno, 1988; Jacobson and Kernell, 1983). Table A-7 (Model 13) in the appendix reports the results of an identical model except with the relative spending variable removed from the equation. Although the coefficients shift around a bit, the basic results are identical with and without spending in the equation. All the individual-level predictors have their expected signs, and most are reliably different from zero. All else equal, Blacks and union members are significantly more likely to vote Democratic. Party identification has a very strong effect, as does self-reported liberal- conservative identification, and approval of the president’s job performance. But controlling on all these standard predictors, our measure of issue agreement (as noisy as it may be) is very strongly related to the vote choice. Indeed, its coefficient is 60 percent larger than any other in the equation. Table 1 Vote Choice in U.S. House of Representatives Elections Aggregate District-Level Predictors Model 1 Coeff. S.E. Model 2 Coeff. S.E. Intercept 0.07 0.23 0.18 0.21 Democratic Incumbent -0.47*** 0.16 Republican Incumbent 0.51*** 0.15 Democratic Candidate Quality/Experience -0.82*** 0.16 Republican Candidate Quality/Experience 0.51** 0.18 District Ideology (conservative high) 1.05*** 0.36 0.93** 0.37 Off-year Penalty (pro-Republican high) 1.25*** 0.52 1.10*** 0.38 Spending Imbalance (favoring Republican high) 2.00* 0.88 1.90* 0.89 Individual-Level Predictors Female -0.09 0.06 -0.09 0.06 Black -0.79*** 0.16 -0.78*** 0.16 Family Income 0.37* 0.22 0.36 0.22 Union Household -0.46*** 0.10 -0.47*** 0.10 Party Identification (Strong Republican high) 3.38*** 0.15 3.37*** 0.15 Liberal - Conservative ID (Very Conservative high) 1.77*** 0.20 1.77*** 0.20 Issue Agreement (closer to Republican high) 5.69*** 0.34 5.70*** 0.34 Approval of President’s Job Perf. (Obama) -2.16*** 0.14 -2.15*** 0.14 Trump President -0.36 0.31 -0.51 0.32 Approval of President’s Job Perf. x Trump Pres. 5.39*** 0.24 5.35*** 0.24 * p < .05 ** p < .01 *** p < .001 Note All tables display the population-average model with robust standard errors. There are 197 different House elections included in the analysis, and individual-level data from 20,501 survey respondents. The dependent variable is coded 0 for a vote for the Democratic candidate, and 1 for a vote for the Republican candidate. Nonvoters and supporters of third-party candidates are excluded from the analysis. All predictors have been rescaled to have a 1-point range. The measure of Issue Agreement employs the directional procedure averaged across the ten policy issues. The analyses include fixed effects for election year, not shown. Because all of our hypotheses are directional, we display one-tailed significance tests. Survey data come from the common content of the 2010, 2012, 2014, 2016, 2018, and 2020 CCES surveys. Aggregate level candidate- and election-specific data were collected by the authors. If we simulate a fairly common Democratic electoral environment with a Democratic incumbent seeking reelection, and a respondent with a weak Democratic party identification and a liberal ideological identification, moving between total indifference between the candidates on the issues and completely agreeing with the Democratic candidate increases by probability of a Democratic vote by 9.3 percent; creating a comparable Republican electoral environment and moving from total indifference to completely agreeing with the Republican candidate on the issues increases the probability of a Republican vote by an equal amount. These figures are a realistic ballpark in which to view the importance of issue concerns over and above everything else in the equation. Creating even more extreme partisan environments (with strong party and ideological identification) reduces the estimated residual effect of issue agreement to 4.2 percent. Any way you look at it, however, Hypothesis 1 is strongly confirmed. There can be no doubt that when a House race is at least vaguely competitive – as all our selected races were – policy agreement is a vital part of many citizens’ vote calculus. The first two rows of Fig. 2 display the coefficient estimates and 95% confidence intervals for our measure of issue voting from these first two models. The remaining rows of Fig. 2 display the confidence intervals for various alternative specifications of issue voting (discussed more fully below, and reported in the online appendix). No matter how the crucial issue agreement variable is operationalized, nor how the fuller model is specified, the story remains the same: Issues matter in House elections, and they matter a lot . The remaining hypotheses explore possible heterogeneity in the importance of issue considerations in the vote choice. Hypothesis 2 predicts that the effects of issue agreement will be enhanced when campaign spending is large or at least relatively equal across the two candidates. The results are shown in Table A-2 in the appendix. The cross-level interaction Figure 2 Different Estimates of the Effects of Issue Voting in U.S. House Elections between issue agreement and total spending (Model 3) does not quite reach conventional levels of statistical significance (p < .13, one-tailed), but the cross-level interaction for equality of spending across the two candidates, as shown in Model 4, does. We thus have mixed support for H 2 . Effect sizes for the crucial variables from these two equations are displayed in the top two rows of Fig. 3. Figure 3 Contextual Heterogeneity in the Importance of Issue Voting in House Elections Hypothesis 3 predicts that issue voting will be greater when the competing candidates are more ideologically distinct, while Hypothesis 4 suggests that issue agreement will be less important when the quality difference between the two candidates is great. The two hypotheses are tested in models 5 and 6 in Table A-3 in the appendix. In both cases, the crucial variable is the cross-level interaction between ideological distinctiveness/candidate quality differential and issue agreement. In the case of ideological distinctiveness (Model 5), the interaction term should be positive; it is trivial in magnitude and negative.[4] There is clearly no support for H 3 . On the other hand, the effect of a large candidate quality differential on the magnitude of issue voting coefficient (see Model 6) is negative, as expected, and both statistically significant and substantively great, reducing the impact of policy agreement on the vote choice by more than 35 percent. We thus have strong support for H 4 . The crucial effect sizes are shown in the 3rd and 4th rows of Fig. 3. Hypothesis 5 predicts that issue voting will be amplified in the most competitive elections. Model 7 in Table A-4 of the Appendix provides the strongest test of this hypothesis, singling out elections that Cook had rated as “tossups.” As predicted, the cross-level interaction between this dummy variable and issue agreement is positive and reliably greater than 0, providing clear support for H 5 . The crucial effects are displayed in the 5th row of Fig. 3. Hypothesis 6 predicts that relative issue agreement will have its strongest effect among political independents who have no partisan affiliation or leaning that helps them make their vote choices. To test this hypothesis, we removed the standard 7-point party identification scale from the equation, and replaced it with two dummy variables, one singling out respondents with strong or not very strong identifications with, or who generally lean toward, the Democratic party; and the other singling out strong, not so strong, and leaning Republicans. We then created two individual-level interaction terms by multiplying each of these new party-specific dummy variables by our measure of relative issue agreement. With such a model specification, the “main effect” of issue agreement in Model 8 represents the effect of that variable among political independents, while the two interaction terms represent the change in the effect of issue agreement among Democrats and Republicans, respectively. H 6 predicts that the effect of issue agreement among independents should be strongly positive, while the two interaction terms should be significantly negative – exactly the pattern of results displayed in Model 8, providing strong support for our final hypothesis. The crucial effect sizes are shown in the last two rows of Fig. 3. Robustness We have conducted a series of supplementary analyses to demonstrate the robustness of our findings. Here we will focus on different operationalizations of the crucial indicator of issue agreement. As described above, to measure issue voting we gathered information from both voters and candidates on ten different political issues, and then employed the Rabinowitz and MacDonald (1989) directional procedure for estimating differential agreement with the candidates on each issue. Many researchers who study issue voting employ a Euclidean proximity model for estimating issue “distances.” We used the proximity procedure to develop alternative measures of issue voting and replicated all of the analysis reported above. Models 9 and 10 in Table A-5 report the results of two of these analyses, which closely replicate the basic results from Table 1. The effect sizes for these alternative specifications of issue voting are shown in the 3rd and 4th rows of Fig. 2. A small cadre of political science methodologists have been trying to devise the best ways to estimate both constituent and legislator policy preferences (e.g., Bailey, 2007; Clinton, Jackman, and Rivers, 2004; Gerber and Lewis, 2004; Jesse, 2009; Lax and Phillips, 2009; Shor and Rogowski, 2018; Tausanovitch and Warshaw, 2013; Warshaw and Rodden, 2012). Some of this research utilizes the CCES surveys, as we do, to estimate constituent opinion, and one can often rely on incumbent roll call votes to estimate their policy positions, but there is no comparable data on the votes that challengers would have made on those bills if they had been in Congress for the previous session. We have not solved this intractable problem in any real sense. At first glance, by relying on candidates’ campaign webpages we have comparable data from the competing candidates, but the types of information different candidates provide on their webpages varies widely – making comparisons across candidates quite difficult – and the survey data we have on respondents (even assuming the policy questions are identical across election years, which they are not) are a totally different kettle of fish. We attempt to overcome all of these difficulties by having different sets of experts use whatever information is available to them to place voters and candidates on comparable 7-point issue scales, but we have no illusions about how precisely comparable the resulting data actually are. Thus we can make “weak” comparisons at best, and any resulting measure of issue agreement must be very noisy. In one sense, that fact only underscores the strength of our basic finding. Despite weak measurement of the crucial variable, we have found very strong evidence for the importance of issue voting in congressional elections. We have also created an alternative measure of issue agreement that reduces the number of scaling assumptions we make by recoding our 7-point issue scales into simpler 3-point scales, and then averaging the different indicators together. This procedure assumes that all survey respondents and all candidates hold either liberal, moderate, or conservative opinions on each of the ten issues. Model 11 in Table A-6 reports one of our analyses employing this simpler alternative measure of issue agreement. The coefficient for issue agreement in Model 11 is about a third smaller than the comparable coefficient in Model 2 (see the 5th row of Fig. 2), but it is still many times larger than its standard error, and the basic story does not change at all about the importance of issue agreement in the House vote decision. Finally, Model 13 in Table A-7 replicates our base model from Table 1 except without the indicator of a spending imbalance between the two candidates. It has long been recognized in the congressional voting literature that campaign resources are often a function of, rather than a cause of, the likely outcome of an election. Model 13 illustrates that any model misspecification resulting from including such a campaign spending variable in the equation does not seriously distort any of our important findings. The coefficient associated with issue agreement in Model 13 barely differs from the one reported in Model 2 of Table 1. Discussion This paper introduces a data set that compares the policy positions of over 20,000 voters in 197 congressional districts to the positions on those same issues taken by the two major party candidates running for election in those districts. Making plausible assumptions about the distributions underlying these policy positions, we calculate broad estimates of policy agreement between voters and the two major congressional candidates competing for their votes. We believe no one else has comparable data on the policy positions of both voters and candidates from so many different congressional elections. Importantly, the estimates of candidate policy positions are reasonably objective, were the information available to voters in the months immediately preceding an election, and were retrieved primarily from candidates’ own campaign websites. Our data should lay to rest the question of whether congressional elections in the United States provide significant levels of substantive representation. There can be little doubt that they do. With the sharp turn towards identity groups, polarization, and tribalism in politics today, we find it critical and heartening that issues still matter in Congressional elections. Despite increasing use of partisan symbols, dog whistles, and chants of “Not My President,” in our data voters consider the issues at least as much as their own partisan identity in making their vote decision. We find candidates in House elections have more varied policy stances during campaigns than you might expect in a highly polarized environment, and that variation is rewarded by voters who prioritize substantive issue agreement. That said, we would never argue that our data gathering procedure and statistical analysis could not be improved. We want to briefly discuss several important limitations of our current analysis and mention a few avenues of future analysis. To begin, as described above we consciously selected the elections to include in our analysis. Most of the selection criteria (e.g., the presence of other elections on the ballot) are reasonably exogenous to the decision calculus employed by voters in the elections we did select, but one was not – the goal of studying only those congressional elections that were at least mildly competitive. Thus, a first limitation of our findings is that they only apply to “at least mildly competitive” congressional elections – which as any congressional scholar knows, is well less than a majority of all such elections. We simply have no idea whether voters in elections with greater than a 24-point final spread also go to the trouble of comparing the candidates’ policy stands to their own. We suspect they do not, but the many noncompetitive congressional elections are simply excluded from our sample, and we have no way of knowing for sure. The restricted sample could be the best explanation for why we find such clear evidence for the importance of issue voting that runs so strongly against conventional wisdom. The average voter knows little about the candidates running in Congressional elections because the typical congressional election is a foregone conclusion. Even if the congressional race were the only election on the ballot, it would make little sense for most citizens to put any effort into learning about the candidates running in their local congressional election. Give the average citizen even mild hope for a competitive election, however, and many of them are perfectly capable of using policy agreement as an important factor in their vote calculus. Another limitation of our analysis is we do not have a random selection of survey respondents within the selected congressional districts. Such random selection is simply beyond the scope of every large-scale political survey that we are aware of, all of which care primarily about being representative at the national level. Although there is now substantial research on the “nationalization” of local politics (Carson, Sievert, and Williamson, 2020; Hopkins, 2018), we also worry that our findings could suffer because we use national level policy issues to measure substantive representation sub-nationally. But the results were strong despite this caveat. We suspect, however, that survey questions about more local or regional issues – e.g., wildfires in the west, marine mammal deaths in the northeast, immigration in border states, or local factory closings just about anywhere – would lead to issue agreement having an even stronger effect on the vote choice. In conclusion, we hope that most readers of this paper will be convinced that when it comes to voting in House of Representatives elections, issues matter – and they matter a lot. At the same time, those same readers could be at a loss the explain how cognitively limited voters could achieve such a result. Although this takes us beyond the scope of the current paper, we have one possible answer, Lodge’s impression-driven model of candidate evaluation (Lodge, McGraw, and Stroh, 1989; Lodge, Steenbergen, and Brau, 1995). According to Lodge, whenever citizens confront new information about a political candidate, they automatically (and almost immediately) download any prior “running tally” evaluation they hold about that candidate, update it according to the affective nature of the new information they have come across, and then remember the updated evaluation while generally forgetting the details of the new information they have seen. Certainly that new information could include the candidate’s policy stands, which would agree or disagree with the citizen’s own policy stands, and consequently increase or decrease the running tally. In this manner a voter’s evaluation of the candidate’s running for some election on the ballot, and ultimately their choice among the competing candidates, could actually reflect issue agreement without voters being able to report many specific details about any of the candidates in that election. We wish to make one final point. We tested five additional hypotheses about specific circumstances in which issue voting should be more or less important, all of which were based on a view of humans as cognitively limited information processors who continually utilize all sorts of cognitive shortcuts and heuristics to help them make sense of their worlds (Fiske and Taylor, 1991 ; Lau and Redlawsk, 2006 ; Lau and Sears, 1986; Lodge, et al., 1988). At one point in the not too distant past, the usefulness of this cognitive perspective for political science was strenuously debated (see for example Kuklinski, Luskin, and Bolland, 1991, and the several responses that immediately followed). We found significant support for four of those five hypotheses, most of which would not even have been offered without this cognitive-limitations viewpoint. There should be little controversy today that an information processing view of human cognition is an invaluable tool for generating testable hypotheses about political behavior. Declarations Acknowledgments : Financial support for this research was provided by the first author’s university. Compliance with Ethical Standards : This project, which involves secondary data analysis only, was deemed exempt from full review and approved by the Human Research Protection Program at REDACTED. Data Availability : The authors will make all data and script files freely available after this paper has been accepted for publication. Disclosures : The authors affirm that they have no competing interests, financial or non-financial, that are directly or indirectly related to the work submitted for publication. Pre-Registration : No formal pre-registration of the hypotheses tested in this paper has been submitted. References Abramowitz, A. I., & Steven Webster (2016). The Rise of Negative Partisanship and the Nationalization of US Elections in the 21st Century. Electoral Studies , 41 (1), 12–22. Aldrich, J. H., Jamie, L., Carson, Brad, T., Gomez, & Merolla, J. L. (2022). Change and Continuity in the 2020 Elections . Roman & Littlefield. Alvarez, M. R., and Jacob Morrier (2025). Do Polarized Issues Carry More Weight in Voter Decision-Making? Insights from the 2022 Congressional Midterm Elections. American Politics Research , 53 (September), 481–498. Ansolabehere, S., and Shiro Kuriwaki (2022). Congressional Representation: Accountability from the Constituent’s Perspective. American Journal of Political Science , 66 (January), 123–139. Ansolabehere, S., SnyderJr., J. M., & Charles Stewart, I. I. I. (2001). Candidate Positioning in U.S. House Elections. American Journal of Political Science , 45 (January), 136–159. Bailey, M. A. (2007). ‘Comparable Preference Estimates Across Time and Institutions for the Court, Congress, and Presidency’. American Journal of Political Science , 51 (July), 433–448. Buttice, M. K., and Walter J. Stone (2012). Candidates Matter: Policy and Quality Differences in Congressional Elections. Journal of Politics , 74 (July), 870–887. Canes-Wrone, B., & Brady, D. W., and John F. Cogan (2002). Out of Step, Out of Office: Electoral Accountability and House Members’ Voting. American Political Science Review , 96 (March), 127–140. Carson Jamie, L., Sievert, J., & Ryan, D. W. (2020). Nationalization and the Incumbency Advantage. Political Research Quarterly , 73 (1), 156–168. Cayton, A., and Ryan Dawkins (2022). Incongruent Voting or Symbolic Representation? Asymmetrical Representation in Congress, 2008–2014. Perspective on Politics , 20 (September), 916–930. Clinton, J. D., & Jackman, S., and Douglas Rivers (2004). ‘The Statistical Analysis of Roll Call Data’. American Political Science Review , 98 (May), 355–370. Converse, P. E. (1964). The Nature of Belief Systems in Mass Publics. In E. David, & Apter (Eds.), Ideology and Discontent (pp. 206–261). Free. Costa, M. (2021). Ideology, Not Affect: What Americans Want from Political Representation. American Journal of Political Science , 65 (April), 342–358. Downs, A. (1957). An Economic Theory of Democracy . Harper and Row. Fiorina, M. P. (1981). Retrospective Voting in American National Elections . Yale University Press. Fiske, S. T., & Taylor, S. E. (1991). Social Cognition (2nd. edition). New York: McGraw-Hill. Gerber, E. R., & Lewis, J. B. (2004). ‘Beyond the Median: Voter Preferences, District Heterogeneity, and Political Representation’. Journal of Political Economy , 112 (December), 1364–1383. Green, D. P., and Jonathan S. Krasno (1988). Salvation for the Spendthrift Incumbent: Re-estimating the Effects of Campaign Spending in House Elections. American Journal of Political Science , 32 (November), 884–907. Griffin, J. D., Newman, B., & Nickerson, D. W. (2019). A God of Vengeance and of Reward? Voters and Accountability. Legislative Studies Quarterly , 44 (1), 133–162. Henderson, J. A., Sheagley, G., Goggin, S. N., Dancey, L., & Theodoridis, A. G. (2022). Primary Divisions: How Voters Evaluate Policy and Group Differences in Intraparty Contests. Journal of Politics , 84 (July), 1760–1776. Highton, B. (2010). The Contextual Causes of Issue and Party Voting in American Presidential Elections. Political Behavior , 32 (December), 453–472. Highton, B. (2019). Issue Accountability in U.S. House Elections. Political Behavior , 41 (2), 349–367. Hollibaugh, G. E., Rothenberg, L. S., & Rulison, K. K. (2013). Does It Really Hurt To Be Out of Step? Political Research Quarterly , 66 (4), 856–867. Hopkins Daniel, J. (2018). The Increasingly United States: How and Why American Political Behavior Nationalized . University of Chicago Press. Jacobson, G. C., & Carson, J. L. (2020). The Politics of Congressional Elections , 10th edition. New York: Roman & Littlefield. Jacobson, G. C., and Samuel Kernell (1983). Strategy and Choice in Congressional Elections (2nd ed.). Yale University Press. Jesse, S. A. (2009). Spatial Voting in the 2004 Presidential Election. American Political Science Review , 103 (February), 59–82. Joesten, D. A., and Walter J. Stone (2014). Reassessing Proximity Voting: Expertise, Party, and Choice in Congressional Elections. Journal of Politics , 76 (July), 740–753. Kuklinski, J. H., & Luskin, R. C., and John Bolland (1991). Where is the Schema? Going Beyond the S Word in Political Psychology. American Political Science Review , 85 (December), 1341–1356. Lau, R. R., & Redlawsk, D. P. (2006). How Voters Decide: Information Processing During Election Campaigns . Cambridge University Press. Lau, R. R., David, O., & Sears (Eds.). (1986). Political Cognition: The 19th Annual Carnegie Symposium on Cognition . Hillside, NJ: Erlbaum. Lax, J., and Justin Phillips (2009). ‘How Should We Estimate Public Opinion in the States?’. American Journal of Political Science , 53 (January), 197–121. Lodge, M., & McGraw, K. M., and Patrick Stroh (1989). An Impression-Driven Model of Candidate Evaluation. American Political Science Review , 83 (June), 399–420. Lodge, M., & Steenbergen, M. R., and Shawn Brau (1995). The Responsive Voter: Campaign Information and the Dynamics of Candidate Evaluation. American Political Science Review , 89 (June), 309–326. Mann, T. E., & Wolfinger, R. E. (1980). Candidates and Parties in Congressional Elections. American Political Science Review , 74 (September), 617–632. Mansbridge, J. (2003). Rethinking Representation American Political Science Review , 97 (November): 515–528. Miller, W. E., & Stokes, D. E. (1963). Constituency Influence in Congress. American Political Science Review , 57 (March), 45–56. Mummolo, J., & Peterson, E., and Sean Westwood (2021). The Limits of Partisan Loyalty. Political Behavior , 43 (September), 949–972. Petrocik, J. R. (1996). Issue Ownership in Presidential Elections, with a 1980 Case Study. American Journal of Political Science , 40 (August), 825–850. Rabinowitz, G., Stuart Elaine, & MacDonald (1989). A Directional Theory of Issue Voting. American Political Science Review , 83 (March), 93–121. Raudenbush, S., Anthony Bryk, and, & Congdon, R. (2022). HLM Hierarchical Linear and Nonlinear Modeling . Scientific Software International. Rogowski, J. C. (2016). Voter Decision-Making with Polarized Choices. British Journal of Political Science , 48 (1), 1–22. Sears, D. O. (1975). Political Socialization. In F. I. Greenstein, & N. W. Polsby (Eds.), Handbook of Political Science, Volume 2, Micropolitical Theory . Addison-Wesley. Shor, B., & Rogowski, J. C. (2018). Ideology and the US Congressional Vote. Political Science Research and Methods , 6 (April), 323–341. Tausanovitch, C., and Christopher Warshaw (2013). Measuring Constituent Policy Preferences in Congress, State Legislatures, and Cities. Journal of Politics , 75 (April), 330–342. Warshaw, C. (2019). Local Elections and Representation in the United States. Annual Review of Political Science , 22 , 461–479. Warshaw, C., and Jonathan Rodden (2012). ‘How Should We Measure District-Level Public Opinion on Individual Issues?’. Journal of Politics , 74 (January), 203–219. Wright, Gerald, J. (1978). Candidates’ Policy Positions and Voting in U.S. Congressional Elections. Legislative Studies Quarterly , 3 (August), 445–464. Footnotes Another set of researchers have gathered data on the general ideological positions of competing Congressional candidates as a substitute for evidence about most specific policy stands. For example, Buttice and Stone (2012; see also Joesten and Stone, 2014) estimate the ideologies and personal qualities of a set of major party candidates running in 155 congressional elections in 2006. This paper is an important step in the right direction in that the authors gather data about the competing candidates in a set of congressional elections (see also Hollibaugh, Rothenberg, and Rulison, 2013 ), but their measure of ideology based on experts’ global judgement of ideology is at best an imperfect instrument for a candidate’s actual policy positions, as ideology has considerable symbolic value as well as conveying policy information (e.g., Cayton and Dawkins, 2022). This test was completed by 72 percent of all major party candidates in 1996. Unfortunately, response rates to this survey, ironically renamed the Political Courage Test, has dropped from 72 percent in 1996 to 48 percent in 2008 and even further to 20 percent by 2016, according to Wikipedia. If these Wikipedia numbers are correct, the 2008 and 2010 congressional elections analyzed by Shor and Rogowski are probably the last with sufficient data from both major party House candidates. We asked Sherman to set the Wayback Machine to October 1 of each election year, and then accessed candidates’ webpage from https://www.loc.gov/collections/united-states-elections-web-archive/ . The appendix provides more detail on the coding of candidates’ policy stands. We employed our estimate of the very liberal to very conservative ideology of the competing candidates because that is the measure employed by Buttice and Stone (2012) in their research. If instead we utilize a measure of the candidate’s ideological difference based on the mean of each candidate’s policy positions across the ten policy areas, the interaction term becomes considerably larger (-5.43) but again in the wrong direction. Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8544888","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":581392502,"identity":"a6ff6f24-8ac5-45a7-b3da-c9f728b79087","order_by":0,"name":"Richard R. Lau","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFklEQVRIie3QPUvDQBjA8ScEkuUk6yOU+BVOMkRw8KtccLglBd1DuBCwS3Cu36KbCA4XHujkbsZAoVOHdpEIIkZrSoe0WQXvnwwHdz/uBcBk+ovZ7VcD+O1IaQDcnxOHSTsVAFg7YqmjBLYkUruFQySc2AsuEpKPbpaVTXIB4VS+1pvnFDw35n1kRE4gxJzGT0WpiM0RRlV8m02XBKfFqpegzWwtHBrPqkiRpRAQ4yhnWgOv+nfZkk+SvCXle0c+dApXR4iI7kh8E33yQ2SZg7aB4yHiBDy6l+fdXRiylfVQaGL4srzpJR4tcP12eRa6OW2aJPXRlfW60anvTa5nva/c1Z2BtT//HQy0d2y3HlxtMplM/6ovIF9is9TSqEMAAAAASUVORK5CYII=","orcid":"","institution":"Rutgers, The State University of New Jersey","correspondingAuthor":true,"prefix":"","firstName":"Richard","middleName":"R.","lastName":"Lau","suffix":""},{"id":581392503,"identity":"3dfbb165-fa9c-4218-b407-b3016d082485","order_by":1,"name":"Amy S. Funck","email":"","orcid":"","institution":"University of Pennsylvania","correspondingAuthor":false,"prefix":"","firstName":"Amy","middleName":"S.","lastName":"Funck","suffix":""}],"badges":[],"createdAt":"2026-01-07 19:53:39","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8544888/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8544888/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101454556,"identity":"c6acbe03-0856-4d11-8352-0308e9a7d786","added_by":"auto","created_at":"2026-01-29 22:03:09","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":141589,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePlot of Estimated to Actual House Election Results\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8544888/v1/93c5d519c79a292498b52113.png"},{"id":101454558,"identity":"0cdd024c-a92d-435c-a8b8-c29ba34547e8","added_by":"auto","created_at":"2026-01-29 22:03:09","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":145056,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDifferent Estimates of the Effects of Issue Voting in U.S. House Elections\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8544888/v1/38846cc9df01cd7fcfe46ef7.png"},{"id":101454555,"identity":"7df0a0f3-a291-4778-979a-c56b4a6501ef","added_by":"auto","created_at":"2026-01-29 22:03:08","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":138813,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eContextual Heterogeneity in the Importance of Issue Voting in House Elections\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8544888/v1/f84b181181d2fe7c7db39a97.png"},{"id":106093403,"identity":"f92b6589-268e-4277-8df5-77fa4b62659c","added_by":"auto","created_at":"2026-04-03 11:37:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1256820,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8544888/v1/88f50f54-7fce-44df-bc4b-397ddf527764.pdf"},{"id":101454557,"identity":"185c5cf1-c478-439b-a17b-16e3741b914e","added_by":"auto","created_at":"2026-01-29 22:03:09","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":44929,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-8544888/v1/34814977369bbf31f7a9b70d.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Issue Voting in U.S. House Elections","fulltext":[{"header":"Introduction","content":"\u003cp\u003eRepresentative democracy is based on the idea that citizens choose the parties or candidates who most closely share their own policy views and concerns, and then monitor the behavior of winning candidates once they are in office to make sure they do what they said they would do. Such substantive representation is not the only type of representation \u0026ndash; see for example Mansbridge, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2003\u003c/span\u003e \u0026ndash; but it is the easiest to justify normatively, and it is what most people think democratic representation is about. In the U.S., almost all research on issue voting involves presidential elections (e.g., Highton, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Jesse, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Petrocik, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e1996\u003c/span\u003e), but the principle applies just as strongly (and even more directly) to legislative elections where the candidates being selected actually vote on the country\u0026rsquo;s laws. Although congressional voting has long been a major focus of political science research, we have surprising little direct evidence about the decisions of individual voters in House elections, and almost no evidence on the extent to which differential policy agreement with competing candidates affects the vote choice. Thus, the degree of substantive democratic representation provided by legislative elections in the U.S. is more a matter of faith than hard empirical evidence.\u003c/p\u003e \u003cp\u003eThe problem is one of data availability. On the one hand, almost no high-quality representative survey is conducted within a single \u0026ndash; or even a few \u0026ndash; legislative districts. Most election surveys focus on the presidential election and aim to be representative of the country as a whole. In the U.S., a nationally representative survey with 1000 respondents includes, on average, a little more than two respondents from each congressional district. It takes a survey with an extraordinarily large initial sample size before this first problem can be addressed. Even if the sample size problem can be overcome, a second and equally tricky problem immediately arises. To measure issue voting, you need estimates of the policy preferences of a set of voters in the district \u003cem\u003eand\u003c/em\u003e estimates of the policy preferences on the same issues (measured on approximately the same scales) of the multiple candidates running for election in that district. The more elections you want to study, and the more issues you want to consider, the greater this second problem becomes.\u003c/p\u003e \u003cp\u003eWe overcome the first problem by utilizing the very large samples from the 2010 through 2020 Cooperative Congressional Election Study (CCES) surveys. To address the second issue, we have gathered data on the policy positions of the Democratic and Republican candidates running in 197 competitive House elections across those six recent election years from the campaign websites of those 394 candidates. The conventional wisdom on House voting assumes that the typical citizen knows so little about their incumbent representative \u0026ndash; much less the candidate who is challenging her in the upcoming election \u0026ndash; that relative policy agreement could not possibly play a large role in the vote decision. Controlling on a set of variables that have been used in prior research on congressional voting, however, we find that vote choice in House elections is strongly a function of relative agreement with the candidates on the issues. We know of no prior research on congressional elections that has comparable detailed data from so many different House elections.\u003c/p\u003e \u003cp\u003eIn the following section we briefly review the literature on congressional elections, focusing on prior research that has attempted to explore the role of policy agreement in congressional voting. After specifying a set of hypotheses that are derived from that literature, we describe the CCES surveys that we relied upon for individual-level data on the policy positions and vote decisions of voters in House elections between 2010 and 2020, and the procedures we developed to gather and code the policy positions the competing candidates took on those same issues. We then present the results of multilevel hierarchical nonlinear models of the vote choice in recent U.S. House of Representatives elections that are aimed at testing those hypotheses. Discussion focuses on how policy agreement can play such an important role in congressional elections when citizens seem to know so little about the candidates running in those elections, and the implications of our findings for understanding the true nature of representation in modern large-scale democracies.\u003c/p\u003e"},{"header":"Literature Review","content":"\u003cp\u003eResearchers studying voting in congressional elections have two primary types of extant data readily available to them: aggregate district- and state-level election returns, and individual-level responses from nationally representative surveys. There are several important facts about the vote decision in congressional elections that are readily revealed by aggregate-level election returns. First, turnout in off-year elections is always less than turnout in presidential election years \u0026ndash; usually around 15 percent less. Even in presidential election years, 3 or 4 percent of citizens who vote for president will not bother to cast a vote for the House of Representatives. Clearly, the public is far more interested in the presidential election than in the typical House election. Second, there are several \u0026ldquo;nationalized\u0026rdquo; factors that increasingly influence congressional elections (see for example, Abramowitz and Webster, 2016). To begin, winning presidential candidates often have \u0026ldquo;coattails\u0026rdquo; that advantage congressional candidates from their party, increasing their vote totals by 3 or 4 percent. In addition, congressional elections are in part a referendum on the sitting president. When the president is popular, House candidates from his party usually have a significant advantage over candidates from the out-party. When the president is unpopular, House candidates from his party suffer. These three factors combine such that House candidates from the president\u0026rsquo;s party typically suffer a significant penalty during off-year elections. Although this number bounces around a fair amount from election to election, on average the president\u0026rsquo;s party will lose about 25 seats in the House of Representatives during off-year elections. Finally, despite all these influences from national politics, incumbents are usually re-elected. Since 1946 over 90 percent of all House elections involved an incumbent seeking reelection, and of those, over 90 percent were successful. See Aldrich, Carson, Gomez, and Merolla (2022), or Jacobson and Carson (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), for recent summaries of this data. None of these insights about House elections, however, are based on the policy positions of any of the candidates involved in those elections.\u003c/p\u003e \u003cp\u003eIndividual-level survey data on voting in House elections has provided somewhat fewer insights into citizens\u0026rsquo; voting behavior, in large part because of data limitations. As mentioned above, the typical political survey has a national focus, with relatively few respondents from any particular congressional district, and relatively few questions about the House election. Large majorities of Americans support the House candidate from their political party \u0026ndash; as is the case for every other partisan election in the country \u0026ndash; with most defections from party explained by incumbency. But again, none of this says anything about issue voting per se, and thus the extent of substantive representation that is provided by House elections. Indeed, the conventional assumption in political science research on congressional voting \u0026ndash; going back to the seminal research of Miller and Stokes (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e1963\u003c/span\u003e) \u0026ndash; is that most people know little about the policy preferences of their incumbent representative, and considerably less about the policy views of any challenger. If this view is correct, it is hard to imagine that policy agreement would have much to do with vote choice in House elections. As summarized by Warsaw in a recent article, \u0026ldquo;electoral accountability\u0026rdquo; [at the local, congressional, and gubernatorial levels] \u0026ldquo;is generally weak and contingent on various institutional moderators\u0026rdquo; (Warsaw, 2019, p. 462).\u003c/p\u003e \u003cp\u003eOn the other hand, as first noted by Mann and Wolfinger (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e1980\u003c/span\u003e), while less than half of all Americans can recall the name of their representative, over 90 percent will \u003cem\u003erecognize\u003c/em\u003e the incumbent\u0026rsquo;s name from a short list (as they might have on a ballot), with somewhat more than half recognizing the major party challenger. The quality of the challenging candidate also matters, with candidate quality usually defined by prior electoral experience. This knowledge is not coming out of thin air. In our view, voters are working with more information about the major party candidates running in congressional elections than is assumed by conventional wisdom, it just may not be readily available for recall.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eDo Issues Matter in House Elections?\u003c/h2\u003e \u003cp\u003eGiven the dearth of appropriate individual-level survey data, several research teams have tried to address this problem indirectly with aggregate-level data, which are readily available and stretch over longer periods of time. Both Canes-Wrone, Brady, and Cogan (2002), and Ansolabehere and Kuriwaki (2022), find that House incumbents are held accountable for being ideologically out of step with their constituents. On the other hand, Ansolabehere, Snyder, and Stewart (2001), and Highton (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), conclude that members of Congress are not much constrained by voters in the positions they take on particular issues. The conflicting results reported by this prior literature using aggregate election results are, we suspect, a function of the indirect nature of their underlying research designs and the number of assumptions that must be made to infer individual vote decisions from aggregate election outcomes. True democratic representation is served when voters hold their representatives accountable for their actions in Congress and vote them out of office if they disagree with too many of them. This is retrospective accountability (Fiorina, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1981\u003c/span\u003e), looking backwards at the prior actions/performance of an incumbent in office. But this is only half of the story, and the smaller half at that. How were those incumbents elected in the first place? And how will voters select their replacements, and/or decide whether any available alternative is superior to the status quo incumbent? Prospective \u003cem\u003eissue voting\u003c/em\u003e occurs when decision makers compare the policy positions offered by one candidate or party to the policy positions of alternative candidates or parties \u0026ndash; not the imagined or presumed opinions of some hypothetical opponent but the actual opinions of competing candidates. Mansbridge (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) calls this promissory representation. To explore issue voting in congressional elections, we need information on the policy preferences of a set of voters \u003cem\u003eand\u003c/em\u003e on the policy stands of a set of competing candidates.[1]\u003ca class=\"FNLink\" href=\"#Fn1\" id=\"#FNLinkFn1\"\u003e\u003c/a\u003e\u003c/p\u003e \u003cp\u003eWe have found only three earlier analyses of actual House elections that have the type of evidence needed. Using data from a 1966 SRC (i.e., pre-ANES) survey and candidate position data gathered by NBC News, Wright (1978) created domestic policy conservatism scores by combining available policy items in each of these surveys, and then calculating each respondent\u0026rsquo;s relative closeness to the Democratic and Republican House candidates. Although the statistical tools available to Wright were somewhat primitive by today\u0026rsquo;s standards, he reports a statistically significant impact of issue proximity on vote choice, controlling on party identification and incumbency \u0026ndash; albeit one that was considerably weaker than the effects of either party ID or incumbency. We worry about the lack of statistical controls in this observational data, and the fact that 622 survey respondents are distributed across 117 congressional districts \u0026ndash; on average, 5 respondents per district.\u003c/p\u003e \u003cp\u003eShor and Rogowski (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; see also Rogowski, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) have much larger surveys, and modern statistical techniques, to bring to bear on this issue. These authors used data from the 2008 Cooperative Campaign Analysis Project (CCAP) and the common content of the entire 2010 CCES survey to measure voter\u0026rsquo;s policy preferences, and data from the National Political Awareness Test (NPAT) to estimate candidates\u0026rsquo; policy positions. The latter is a survey administered by the nonpartisan group Project Vote Smart and sent to announced-candidates for all federal and state offices in an effort to provide voter\u0026rsquo;s with easy, unbiased information on candidate\u0026rsquo;s policy positions.[2]\u003ca class=\"FNLink\" href=\"#Fn2\" id=\"#FNLinkFn2\"\u003e\u003c/a\u003e Controlling for party identification and incumbency, the authors find strong evidence that their measure of relative issue closeness to the Democratic and Republican candidates had a statistically significant and substantively meaningful effect on the vote choice. This is far and away the strongest direct evidence that issue voting plays an important role in the vote choice for House of Representatives elections. In our view, it suggests a necessary revision of conventional wisdom about the nature of accountability provided by congressional elections. The best evidence we have clearly suggests that policy agreement could be an important part of the vote calculus of many Americans, even for elections well below the top of the ballot.\u003c/p\u003e \u003cp\u003eWe hope to extend and improve upon Shor and Rogowski\u0026rsquo;s research in several important ways. We utilize CCES survey data extending across eleven years and six federal elections \u0026ndash; three off-year elections, and three presidential year elections. We have a very different method for measuring candidates\u0026rsquo; issue stands, one that does not rely upon both major party candidates publicly completing a survey of their policy positions, and one that can be replicated in future elections. Following Buttice and Stone\u0026rsquo;s (2012) counsel, we include a measure of the relative quality of the competing candidates in our analysis. We have a considerably larger average number of respondents per congressional district (in our case, 133; Shor and Rogowski\u0026rsquo;s means were only 8 in their 2008 data, but a much more robust 83 in their 2010 election data). We also employ a multilevel statistical procedure as our main analytic tool, thus providing much more reliable estimates of the importance of aggregate-level factors such as incumbency and candidate quality.\u003c/p\u003e \u003cp\u003eBefore proceeding we want to be very clear about one important factor. Although we are primarily interested in the \u003cem\u003ecausal\u003c/em\u003e effect of issue agreement on the vote choice, we \u0026ndash; as Wright (1978), and Shor and Rogowski (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), before us \u0026ndash; only have correlational data, and thus any causal inferences we would like to make are open to omitted variables bias. The multivariate analysis that follows controls on a host of plausible moderating and mediating variables, but of course not every conceivable one, and thus the causal inferences one can draw from our results are suggestive at best. There is a growing body of recent experimental research (particularly involving conjoint designs) that provides strong causal evidence of the effects of policy agreement on \u003cem\u003ehypothetical\u003c/em\u003e vote decisions (e.g., Alvarez and Morrier, 2025; Costa, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Griffin, Newman, and Nickerson, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Henderson et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Mummolo, Peterson, and Westwood, 2021), although as is always the case with experimental data, the external validity of the reported results is somewhat suspect.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eHypotheses\u003c/h3\u003e\n\u003cp\u003eOne primary hypothesis drives all analysis in this paper.\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e \u003csub\u003e \u003cem\u003e1\u003c/em\u003e \u003c/sub\u003e: \u003cem\u003eControlling for a host of aggregate- and individual-level factors that past research has found to be important, greater agreement with one of the competing candidates on a set of policy issues \u0026ndash; that is, relative issue agreement \u0026ndash; will be associated with a higher likelihood of voting for that candidate in elections for the House of Representatives.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eWe also test several hypotheses suggested by prior research about the contingent nature of issue voting in House elections. Most of these hypotheses assume that issue voting is cognitively difficult, and factors that reduce the difficulty, or motivate greater effort, can increase the importance of policy considerations in the vote choice. According to Shor and Rogowski (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), for example, the importance of issue agreement in the vote decision will be greater when the candidates have jointly spent a great deal of money than when relatively little has been spent of the campaign. The logic behind this hypothesis is that campaign spending \u0026ndash; presumably much of it about policy \u0026ndash; makes it easier for voters to learn the issue stands of the competing candidates. A somewhat more nuanced hypothesis realizes that it takes two to tango, and awareness of the policy positions of one candidate but not the other does not encourage true issue voting. Either way:\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e \u003csub\u003e \u003cem\u003e2\u003c/em\u003e \u003c/sub\u003e: \u003cem\u003eThe effect of issue agreement on the vote choice is conditional on the magnitude (and/or equality) of campaign spending in the election. The cross-level interaction between campaign spending and relative issue agreement should be positive.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eAnother factor that should make issue voting easier is when the competing candidates take ideologically distinct positions. According to Buttice and Stone (2012), when candidates hold dramatically different policy views, voters should not only be more aware of those differences and should also \u003cem\u003ecare\u003c/em\u003e more about the outcome of the election, which could motivate the effort required for issue voting.\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e \u003csub\u003e \u003cem\u003e3\u003c/em\u003e \u003c/sub\u003e: \u003cem\u003eThe importance of policy agreement in the vote decision will be greater when the two candidates are clearly ideologically distinct than when they have relatively similar ideologies. Thus the cross-level interaction between the ideological distinctiveness of the two candidates and relative issue agreement should be positive.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eOn the other hand, Buttice and Stone argue, the greater the quality (experience) difference between the candidates, the lower the need and motivation to seek out policy-based information, and thus the weaker the relationship between issue positions and the vote.\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e \u003csub\u003e \u003cem\u003e4\u003c/em\u003e \u003c/sub\u003e: \u003cem\u003ePolicy agreement will be less important when the quality/experience difference between the candidates is greater. Therefore the cross-level interaction between the candidate quality differences and relative issue agreement should be negative.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eAnother factor that should motivate greater cognitive effort is the expected competitiveness of the election. If the election has a foregone conclusion \u0026ndash; which is the case for many (if not most) House elections \u0026ndash; why go to the effort of studying the competing candidates to figure out which one best represents my values and concerns? But if both candidates have a realistic chance of winning, expending the cognitive effort to learn about the candidates becomes much more rational. Hence:\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e \u003csub\u003e \u003cem\u003e5\u003c/em\u003e \u003c/sub\u003e: \u003cem\u003eThe importance of policy agreement in the vote decision will be greater when the election is expected to be competitive, but less when one candidate has little or no chance of winning. Thus the cross-level interaction between the expected competitiveness of the election and relative issue agreement should be positive.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eFinally, to the extent that party identification provides a valuable heuristic that can greatly simplify decision making, particularly relative to the effort required to compare candidates on their policy promises, we would expect issue agreement to have the strongest effect on the vote decision among (\u0026ldquo;pure\u0026rdquo;) independent-independents who have no partisan leanings to rely upon. Hence:\u003c/p\u003e \u003cp\u003e \u003cem\u003eH\u003c/em\u003e \u003csub\u003e \u003cem\u003e6\u003c/em\u003e \u003c/sub\u003e: \u003cem\u003eThe effect of relative issue agreement should be strongest when identification with either of the two major parties is absent \u0026ndash; that is to say, among political independents.\u003c/em\u003e\u003c/p\u003e"},{"header":"Method","content":"\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eSample Selection\u003c/h2\u003e \u003cp\u003eTo begin exploring these questions, we utilize the common content from Cooperative Congressional Election Study (CCES) data (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://cces.gov.harvard.edu/\u003c/span\u003e\u003cspan address=\"https://cces.gov.harvard.edu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), pre-post online survey panels administered by YouGov that have been conducted during every federal election since 2006. From our perspective, the unbeatable advantage of the CCES surveys is their large sample size, upwards to 60,000 or more respondents every election year, providing many statewide samples of 1,000 respondents or more each election year, and many congressional district samples with 100 respondents or more. A shortcoming of the CCES data is that while the various samples are designed to be representative of the nation as a whole and of each state, the data are not perfectly representative at the congressional district level. In an ideal world they would be, although no one, to our knowledge, has large-scale data like that. Our goal is to make general statements about the nature of the vote decisions in congressional elections that are at least modestly competitive. That takes random sampling of elections off the table from the get-go. We trust that any small errors in the representativeness of the data from specific districts will cancel out as we combine the data from nearly 200 different congressional elections.\u003c/p\u003e \u003cp\u003eTo provide some evidence for the representativeness of our selected elections, Fig.\u0026nbsp;1 displays a simple bivariate plot comparing the actual two-party proportion of votes received by the Republican candidate, to the proportion of reported votes for the Republican candidate from the CCES sample, in each of the selected districts. The CCES data clearly tracks the actual results. On average, the 95 percent confidence interval around each of these points is almost +/- 9 points; only a handful of observed points are farther than that from the regression line. We see no reason to attribute the dispersion of data points to anything other than random sampling error.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure 1\u003c/b\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePlot of Estimated to Actual House Election Results\u003c/h3\u003e\n\u003cp\u003e \u003c/p\u003e \u003cp\u003eWe selected six recent federal election years for study: 2010, 2012, 2014, 2016, 2018, and 2020. Three of those years included presidential elections; three were off-year elections. Within each election year, we divided states into four categories according to how many \u0026ldquo;higher level\u0026rdquo; statewide elections were on the ballot: states with neither a gubernatorial nor senate election on the ballot, states with a gubernatorial election but no senate election, states with a senate election but no governor\u0026rsquo;s election, and states with both a gubernatorial and a senate election on the ballot. In each election year, our goal was to select four states from each of those categories with at least 1,000 respondents in the sample, half of which had reasonably competitive senate or gubernatorial elections (final margin less than 10 points), half of which were less competitive (final margin 10\u0026ndash;24 points). Within each of the chosen states, we then wanted to select two congressional districts with 100 or more respondents in the data, one of which had a final vote spread of within 10 percent, and one less competitive, defined as between 11\u0026ndash;25 percent. More often than not, we were able to hit those selection targets, but it was impossible in a few cases because rarely are even 10 percent of House elections moderately competitive. After we identified all districts that met our criteria and organized them by type of election, we took a random sample from within each cell to generate our dataset.\u003c/p\u003e \u003cp\u003eIn the end, we have data from 197 separate congressional elections, evenly distributed across the six federal election years. The number of respondents per district varies between 70 and 218, with a mean just over 133. The total combined sample has data from 26,230 survey respondents, although this number is reduced to 20,501 once nonvoters, supporters of third party candidates, and a few respondents with other missing data are removed.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eOperationalizing Variables\u003c/h2\u003e \u003cp\u003eExcept for relative agreement with candidate issue positions, all individual-level predictors come from the common core of the CCES surveys. The dependent variable is dichotomous reported vote for the Democratic or Republican House candidate, with vote for the Republican coded high. All nonvoters and respondents who reported voting for a third-party candidate are excluded from the analysis.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eVoter Policy Preferences\u003c/h3\u003e\n\u003cp\u003eTo measure issue voting, we first selected 10 policy issues about which there were decent measures in all six CCES surveys. Five of these issues were primarily moral/social issues: abortion, environmental protection, gay rights, gun control, and immigration. Five of the issues were primarily economic issues: domestic spending, deficit reduction, tax policy, health care (ACA/Obamacare), and defense spending. For the 2020 election, we added policy questions about the best way to deal with recovering from the pandemic, and policing reform. Most of the policy items asked in the CCES surveys are dichotomous, focused on bills that had been recently considered in Congress, but there were usually several of them falling into the different policy areas. The particular questions addressing the ten issues were not always the same across the six election years, so we translated the available items into reasonably comparable 7-point Very Liberal to Very Conservative policy stances across the various election surveys. Coding details are provided in the online appendix.\u003c/p\u003e\n\u003ch3\u003eCandidate Policy Positions\u003c/h3\u003e\n\u003cp\u003eOur biggest methodological challenge in this project was gathering reasonably objective estimates of where the competing candidates stood on the same policies and political concerns that were used to measure the na\u0026iuml;ve preferences of the voters in our surveys. To do this, we searched candidate webpages from the Library of Congress\u0026rsquo;s internet archive Wayback Machine to locate policy-relevant statements about these same ten political issues.[3]\u003ca class=\"FNLink\" href=\"#Fn3\" id=\"#FNLinkFn3\"\u003e\u003c/a\u003e We then stripped all identifying information from the file, shuffled the data, and had two expert judges independently read the policy statements from the candidates\u0026rsquo; webpages, and attribute a position for them on the same 7-point Very Liberal to Very Conservative policy scales that were used to represent respondents\u0026rsquo; policy positions. Averaging across the ten issues, the two coders attributed exactly the same position on the 7-point policy scales 69 percent of the time, and had near perfect agreement \u0026ndash; differing by 1 point or less \u0026ndash; 92 percent of the time. Weighted Cohen\u0026rsquo;s Kappa averaged .92 across the 10 issues. Whenever their initial independent judgments differed by more than 1 point, the coders met to review the information and try to reconcile their disagreements. The senior author served as the tiebreaker the few instances when the original judges could not reach an agreement. We then took the mean rating of the two expert judges as an objective reading of where the candidates actually stood on each of these ten policy issues.\u003c/p\u003e \u003cp\u003eAssuming we were able to estimate policy positions from both a voter and a candidate, issue agreement was calculated via the \u0026ldquo;directional\u0026rdquo; procedure developed by Rabinowitz and MacDonald (1989). We use the directional model because it seemed more plausible in situations where voters are unlikely to have much information about the policy positions of the competing candidates other than which side of the ideological fence they are on. However, we replicate all analyses using a measure of issue agreement that is based on a Euclidean proximity model, which produces essentially identical. The section on robustness describes an alternative measure of policy agreement that makes fewer comparability assumptions and also produces very similar results.\u003c/p\u003e \u003cp\u003eBy calculating issue agreement in this manner, we assume that the very-liberal to very-conservative scales used to measure voters\u0026rsquo; policy preferences and candidates\u0026rsquo; policy preferences are comparable. To the extent that assumption in not accurate \u0026ndash; and we suspect that extent is pretty great \u0026ndash; the issue voting variable will be quite noisy. Measurement error should make hypothesis tests more conservative. As is common in studies of issue voting, we then created a summary measure of issue agreement by averaging together the ten issue-specific agreement scores for the Democratic candidate and subtracting it from a similar measure averaging across the ten issue-specific scores for the Republican candidate.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eOther Variables\u003c/h2\u003e \u003cp\u003eTurning to the aggregate level, measures of incumbency for both the Democratic and Republican candidates are simple dummy variables that equaled 0 for all non-incumbents but equaled 1 whenever an incumbent was seeking reelection. More refined measures of candidate quality are based on years of experience as a member of the House of Representatives and, counting less, experience in other elected offices. Thus the measures of incumbency, and candidate quality, are going to be highly correlated with each other \u0026ndash; so much so that we cannot include them both in the same equation. In most of the analysis presented below we rely on the more nuanced candidate quality variables because they have more variance, but present some analyses with the simpler incumbent dummy variables to demonstrate the robustness of our findings.\u003c/p\u003e \u003cp\u003eDistrict ideology is measured by the proportion of the two-party vote obtained by the Republican candidate in the most recent prior presidential election. Off-year penalty equals 1 for all candidates from the incumbent president\u0026rsquo;s party during off-year elections, and was 0 otherwise. Measures of campaign spending are based on official FEC records and sum across all reported spending by or on behalf of a candidate. Spending imbalance is calculated as the (natural) log of total spending for the Republican candidate minus the log of total spending for the Democratic candidate. Total campaign spending is the log of total spending for the Republican candidate plus the log of total spending for the Democratic candidate. Spending equality is the absolute value of the difference between the log of total spending for the Republican candidate minus the log of total spending for the Democratic candidate, reversed so that greater equality of spending is scored high.\u003c/p\u003e \u003cp\u003eThe CCES surveys always asked respondents to rate the ideology of major party congressional candidates on a standard 7-point strong liberal to strong conservative scale. We take the mean rating of political experts in every congressional district (those in the top quartile of general political knowledge), as a semi-objective reading of the candidates\u0026rsquo; ideologies. The absolute value of the difference between these mean ratings of the two candidates is our measure of candidate ideological distinctiveness. The publicly available Cook Report rates all House elections on a 7-point scale ranging from \u0026minus;\u0026thinsp;3 (Safe Democrat) to +\u0026thinsp;3 (Safe Republican). To single out the most competitive elections, we created a dummy variable that equaled 1 whenever Cook rated the election as a \u0026ldquo;tossup,\u0026rdquo; and was 0 otherwise.\u003c/p\u003e \u003cp\u003eTo ease interpretation, all variables have been recoded to have a 1-point range. Descriptive statistics for all variables used in the analysis are presented in Table A-\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e in the online appendix.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eTo test our hypotheses, we specify a series of nonlinear multilevel models in which survey respondents are nested within congressional districts. We include fixed effects for election year. Analyses are conducted with the computer program HLM 8.2 (Raudenbush, Bryk, and Congdon, 2022). All tables present the results from the population-average model with robust standard errors. We estimate fixed effects for all individual-level predictors except issue agreement, which is allowed to vary randomly across elections. Thus, the degrees of freedom for all statistical tests involving estimates of issue voting are 196, not 20,095 as they generally are for all other individual-level predictors.\u003c/p\u003e\n\u003cp\u003eTo give readers a good sense of the analyses presented throughout this paper, the results of our most basic models of vote choice are shown in Table\u0026nbsp;1. The aggregate level variables have their predicted sign and are statistically significant. Model 1 includes simple dummy variables noting the party of incumbents seeking reelection, while Model 2 includes instead the more refined measures of candidate experience/quality. Following the norms of this literature, we include a measure of relative candidate campaign spending, although we readily acknowledge (as have most others before us) that candidate spending is endogenous to the likely outcome of the election (e.g., Green and Krasno, 1988; Jacobson and Kernell, 1983). Table A-7 (Model 13) in the appendix reports the results of an identical model except with the relative spending variable removed from the equation. Although the coefficients shift around a bit, the basic results are identical with and without spending in the equation.\u003c/p\u003e\n\u003cp\u003eAll the individual-level predictors have their expected signs, and most are reliably different from zero. All else equal, Blacks and union members are significantly more likely to vote Democratic. Party identification has a very strong effect, as does self-reported liberal- conservative identification, and approval of the president’s job performance. But controlling on all these standard predictors, our measure of issue agreement (as noisy as it may be) is very strongly related to the vote choice. Indeed, its coefficient is 60 percent larger than any other in the equation.\u003c/p\u003e\n\u003cdiv\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eVote Choice in U.S. House of Representatives Elections\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAggregate District-Level Predictors\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eModel 1\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eCoeff. S.E.\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eModel 2\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eCoeff. S.E.\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntercept\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDemocratic Incumbent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.47***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRepublican Incumbent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.51***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDemocratic Candidate Quality/Experience\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.82***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRepublican Candidate Quality/Experience\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.51**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDistrict Ideology (conservative high)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.05***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOff-year Penalty (pro-Republican high)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.25***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.10***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpending Imbalance (favoring Republican high)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.90*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e\u003cem\u003eIndividual-Level Predictors\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.79***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.78***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFamily Income\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.37*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.22\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUnion Household\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.46***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.47***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eParty Identification (Strong Republican high)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.38***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.37***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLiberal - Conservative ID (Very Conservative high)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.77***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.77***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.20\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eIssue Agreement (closer to Republican high)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e5.69***\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.34\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e5.70***\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.34\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eApproval of President’s Job Perf. (Obama)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-2.16***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-2.15***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTrump President\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.32\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eApproval of President’s Job Perf. x Trump Pres.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.39***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5.35***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e*\u003cem\u003ep\u003c/em\u003e \u0026lt; .05 **\u003cem\u003ep\u003c/em\u003e \u0026lt; .01 ***\u003cem\u003ep\u003c/em\u003e \u0026lt; .001\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNote\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll tables display the population-average model with robust standard errors. There are 197 different House elections included in the analysis, and individual-level data from 20,501 survey respondents. The dependent variable is coded 0 for a vote for the Democratic candidate, and 1 for a vote for the Republican candidate. Nonvoters and supporters of third-party candidates are excluded from the analysis. All predictors have been rescaled to have a 1-point range. The measure of Issue Agreement employs the directional procedure averaged across the ten policy issues. The analyses include fixed effects for election year, not shown. Because all of our hypotheses are directional, we display one-tailed significance tests. Survey data come from the common content of the 2010, 2012, 2014, 2016, 2018, and 2020 CCES surveys. Aggregate level candidate- and election-specific data were collected by the authors.\u003c/p\u003e\n\u003cp\u003eIf we simulate a fairly common Democratic electoral environment with a Democratic incumbent seeking reelection, and a respondent with a weak Democratic party identification and a liberal ideological identification, moving between total indifference between the candidates on the issues and completely agreeing with the Democratic candidate increases by probability of a Democratic vote by 9.3 percent; creating a comparable Republican electoral environment and moving from total indifference to completely agreeing with the Republican candidate on the issues increases the probability of a Republican vote by an equal amount. These figures are a realistic ballpark in which to view the importance of issue concerns over and above everything else in the equation. Creating even more extreme partisan environments (with strong party and ideological identification) reduces the estimated residual effect of issue agreement to 4.2 percent. Any way you look at it, however, Hypothesis 1 is strongly confirmed. There can be no doubt that when a House race is at least vaguely competitive – as all our selected races were – policy agreement is a vital part of many citizens’ vote calculus.\u003c/p\u003e\n\u003cp\u003eThe first two rows of Fig.\u0026nbsp;2 display the coefficient estimates and 95% confidence intervals for our measure of issue voting from these first two models. The remaining rows of Fig.\u0026nbsp;2 display the confidence intervals for various alternative specifications of issue voting (discussed more fully below, and reported in the online appendix). No matter how the crucial issue agreement variable is operationalized, nor how the fuller model is specified, the story remains the same: Issues \u003cem\u003ematter\u003c/em\u003e in House elections, and they matter \u003cem\u003ea lot\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThe remaining hypotheses explore possible heterogeneity in the importance of issue considerations in the vote choice. Hypothesis 2 predicts that the effects of issue agreement will be enhanced when campaign spending is large or at least relatively equal across the two candidates. The results are shown in Table A-2 in the appendix. The cross-level interaction\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFigure 2\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv id=\"Sec13\"\u003e\n \u003ch2\u003eDifferent Estimates of the Effects of Issue Voting in U.S. House Elections\u003c/h2\u003e\n \u003cp\u003ebetween issue agreement and total spending (Model 3) does not quite reach conventional levels of statistical significance (p \u0026lt; .13, one-tailed), but the cross-level interaction for equality of spending across the two candidates, as shown in Model 4, does. We thus have mixed support for H\u003csub\u003e2\u003c/sub\u003e. Effect sizes for the crucial variables from these two equations are displayed in the top two rows of Fig.\u0026nbsp;3.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eFigure 3\u003c/strong\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\"\u003e\n \u003ch2\u003eContextual Heterogeneity in the Importance of Issue Voting in House Elections\u003c/h2\u003e\n \u003cp\u003e\u003cstrong\u003eHypothesis 3\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003epredicts that issue voting will be greater when the competing candidates are more ideologically distinct, while Hypothesis 4 suggests that issue agreement will be less important when the quality difference between the two candidates is great. The two hypotheses are tested in models 5 and 6 in Table A-3 in the appendix. In both cases, the crucial variable is the cross-level interaction between ideological distinctiveness/candidate quality differential and issue agreement. In the case of ideological distinctiveness (Model 5), the interaction term should be positive; it is trivial in magnitude and negative.[4]\u003ca href=\"#Fn4\" id=\"#FNLinkFn4\"\u003e\u003c/a\u003e There is clearly no support for H\u003csub\u003e3\u003c/sub\u003e.\u003c/p\u003e\n \u003cp\u003eOn the other hand, the effect of a large candidate quality differential on the magnitude of issue voting coefficient (see Model 6) is negative, as expected, and both statistically significant and substantively great, reducing the impact of policy agreement on the vote choice by more than 35 percent. We thus have strong support for H\u003csub\u003e4\u003c/sub\u003e. The crucial effect sizes are shown in the 3rd and 4th rows of Fig.\u0026nbsp;3.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eHypothesis 5\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003epredicts that issue voting will be amplified in the most competitive elections. Model 7 in Table A-4 of the Appendix provides the strongest test of this hypothesis, singling out elections that Cook had rated as “tossups.” As predicted, the cross-level interaction between this dummy variable and issue agreement is positive and reliably greater than 0, providing clear support for H\u003csub\u003e5\u003c/sub\u003e. The crucial effects are displayed in the 5th row of Fig.\u0026nbsp;3.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eHypothesis 6\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003epredicts that relative issue agreement will have its strongest effect among political independents who have no partisan affiliation or leaning that helps them make their vote choices. To test this hypothesis, we removed the standard 7-point party identification scale from the equation, and replaced it with two dummy variables, one singling out respondents with strong or not very strong identifications with, or who generally lean toward, the Democratic party; and the other singling out strong, not so strong, and leaning Republicans. We then created two individual-level interaction terms by multiplying each of these new party-specific dummy variables by our measure of relative issue agreement. With such a model specification, the “main effect” of issue agreement in Model 8 represents the effect of that variable among political independents, while the two interaction terms represent the change in the effect of issue agreement among Democrats and Republicans, respectively. H\u003csub\u003e6\u003c/sub\u003e predicts that the effect of issue agreement among independents should be strongly positive, while the two interaction terms should be significantly negative – exactly the pattern of results displayed in Model 8, providing strong support for our final hypothesis. The crucial effect sizes are shown in the last two rows of Fig.\u0026nbsp;3.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\"\u003e\n \u003ch2\u003eRobustness\u003c/h2\u003e\n \u003cp\u003eWe have conducted a series of supplementary analyses to demonstrate the robustness of our findings. Here we will focus on different operationalizations of the crucial indicator of issue agreement. As described above, to measure issue voting we gathered information from both voters and candidates on ten different political issues, and then employed the Rabinowitz and MacDonald (1989) directional procedure for estimating differential agreement with the candidates on each issue. Many researchers who study issue voting employ a Euclidean proximity model for estimating issue “distances.” We used the proximity procedure to develop alternative measures of issue voting and replicated all of the analysis reported above. Models 9 and 10 in Table A-5 report the results of two of these analyses, which closely replicate the basic results from Table\u0026nbsp;1. The effect sizes for these alternative specifications of issue voting are shown in the 3rd and 4th rows of Fig.\u0026nbsp;2.\u003c/p\u003e\n \u003cp\u003eA small cadre of political science methodologists have been trying to devise the best ways to estimate both constituent and legislator policy preferences (e.g., Bailey, 2007; Clinton, Jackman, and Rivers, 2004; Gerber and Lewis, 2004; Jesse, 2009; Lax and Phillips, 2009; Shor and Rogowski, 2018; Tausanovitch and Warshaw, 2013; Warshaw and Rodden, 2012). Some of this research utilizes the CCES surveys, as we do, to estimate constituent opinion, and one can often rely on incumbent roll call votes to estimate their policy positions, but there is no comparable data on the votes that challengers would have made on those bills if they had been in Congress for the previous session. We have not solved this intractable problem in any real sense.\u003c/p\u003e\n \u003cp\u003eAt first glance, by relying on candidates’ campaign webpages we have comparable data from the competing candidates, but the types of information different candidates provide on their webpages varies widely – making comparisons across candidates quite difficult – and the survey data we have on respondents (even assuming the policy questions are identical across election years, which they are not) are a totally different kettle of fish. We attempt to overcome all of these difficulties by having different sets of experts use whatever information is available to them to place voters and candidates on comparable 7-point issue scales, but we have no illusions about how precisely comparable the resulting data actually are. Thus we can make “weak” comparisons at best, and any resulting measure of issue agreement \u003cem\u003emust\u003c/em\u003e be very noisy. In one sense, that fact only underscores the strength of our basic finding. \u003cem\u003eDespite\u003c/em\u003e weak measurement of the crucial variable, we have found very strong evidence for the importance of issue voting in congressional elections.\u003c/p\u003e\n \u003cp\u003eWe have also created an alternative measure of issue agreement that reduces the number of scaling assumptions we make by recoding our 7-point issue scales into simpler 3-point scales, and then averaging the different indicators together. This procedure assumes that all survey respondents and all candidates hold either liberal, moderate, or conservative opinions on each of the ten issues. Model 11 in Table A-6 reports one of our analyses employing this simpler alternative measure of issue agreement. The coefficient for issue agreement in Model 11 is about a third smaller than the comparable coefficient in Model 2 (see the 5th row of Fig.\u0026nbsp;2), but it is still many times larger than its standard error, and the basic story does not change at all about the importance of issue agreement in the House vote decision.\u003c/p\u003e\n \u003cp\u003eFinally, Model 13 in Table A-7 replicates our base model from Table\u0026nbsp;1 except without the indicator of a spending imbalance between the two candidates. It has long been recognized in the congressional voting literature that campaign resources are often a function of, rather than a cause of, the likely outcome of an election. Model 13 illustrates that any model misspecification resulting from including such a campaign spending variable in the equation does not seriously distort any of our important findings. The coefficient associated with issue agreement in Model 13 barely differs from the one reported in Model 2 of Table\u0026nbsp;1.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis paper introduces a data set that compares the policy positions of over 20,000 voters in 197 congressional districts to the positions on those same issues taken by the two major party candidates running for election in those districts. Making plausible assumptions about the distributions underlying these policy positions, we calculate broad estimates of policy agreement between voters and the two major congressional candidates competing for their votes. We believe no one else has comparable data on the policy positions of both voters and candidates from so many different congressional elections. Importantly, the estimates of candidate policy positions are reasonably objective, were the information available to voters in the months immediately preceding an election, and were retrieved primarily from candidates\u0026rsquo; own campaign websites. Our data should lay to rest the question of whether congressional elections in the United States provide significant levels of substantive representation. There can be little doubt that they do.\u003c/p\u003e \u003cp\u003eWith the sharp turn towards identity groups, polarization, and tribalism in politics today, we find it critical and heartening that \u003cem\u003eissues still matter\u003c/em\u003e in Congressional elections. Despite increasing use of partisan symbols, dog whistles, and chants of \u0026ldquo;Not My President,\u0026rdquo; in our data voters consider the issues at least as much as their own partisan identity in making their vote decision. We find candidates in House elections have more varied policy stances during campaigns than you might expect in a highly polarized environment, and that variation is rewarded by voters who prioritize substantive issue agreement.\u003c/p\u003e \u003cp\u003eThat said, we would never argue that our data gathering procedure and statistical analysis could not be improved. We want to briefly discuss several important limitations of our current analysis and mention a few avenues of future analysis. To begin, as described above we consciously selected the elections to include in our analysis. Most of the selection criteria (e.g., the presence of other elections on the ballot) are reasonably exogenous to the decision calculus employed by voters in the elections we did select, but one was not \u0026ndash; the goal of studying only those congressional elections that were at least mildly competitive. Thus, a first limitation of our findings is that they only apply to \u0026ldquo;at least mildly competitive\u0026rdquo; congressional elections \u0026ndash; which as any congressional scholar knows, is well less than a majority of all such elections. We simply have no idea whether voters in elections with greater than a 24-point final spread also go to the trouble of comparing the candidates\u0026rsquo; policy stands to their own. We suspect they do not, but the many noncompetitive congressional elections are simply excluded from our sample, and we have no way of knowing for sure. The restricted sample could be the best explanation for why we find such clear evidence for the importance of issue voting that runs so strongly against conventional wisdom. The average voter knows little about the candidates running in Congressional elections because the typical congressional election is a foregone conclusion. Even if the congressional race were the only election on the ballot, it would make little sense for most citizens to put any effort into learning about the candidates running in their local congressional election. Give the average citizen even mild hope for a competitive election, however, and many of them are perfectly capable of using policy agreement as an important factor in their vote calculus.\u003c/p\u003e \u003cp\u003eAnother limitation of our analysis is we do not have a random selection of survey respondents within the selected congressional districts. Such random selection is simply beyond the scope of every large-scale political survey that we are aware of, all of which care primarily about being representative at the national level.\u003c/p\u003e \u003cp\u003eAlthough there is now substantial research on the \u0026ldquo;nationalization\u0026rdquo; of local politics (Carson, Sievert, and Williamson, 2020; Hopkins, 2018), we also worry that our findings could suffer because we use national level policy issues to measure substantive representation sub-nationally. But the results were strong despite this caveat. We suspect, however, that survey questions about more local or regional issues \u0026ndash; e.g., wildfires in the west, marine mammal deaths in the northeast, immigration in border states, or local factory closings just about anywhere \u0026ndash; would lead to issue agreement having an even stronger effect on the vote choice.\u003c/p\u003e \u003cp\u003eIn conclusion, we hope that most readers of this paper will be convinced that when it comes to voting in House of Representatives elections, issues matter \u0026ndash; and they matter a lot. At the same time, those same readers could be at a loss the explain how cognitively limited voters could achieve such a result. Although this takes us beyond the scope of the current paper, we have one possible answer, Lodge\u0026rsquo;s impression-driven model of candidate evaluation (Lodge, McGraw, and Stroh, 1989; Lodge, Steenbergen, and Brau, 1995). According to Lodge, whenever citizens confront new information about a political candidate, they automatically (and almost immediately) download any prior \u0026ldquo;running tally\u0026rdquo; evaluation they hold about that candidate, update it according to the affective nature of the new information they have come across, and then remember the updated evaluation while generally forgetting the details of the new information they have seen. Certainly that new information could include the candidate\u0026rsquo;s policy stands, which would agree or disagree with the citizen\u0026rsquo;s own policy stands, and consequently increase or decrease the running tally. In this manner a voter\u0026rsquo;s evaluation of the candidate\u0026rsquo;s running for some election on the ballot, and ultimately their choice among the competing candidates, could actually reflect issue agreement without voters being able to report many specific details about any of the candidates in that election.\u003c/p\u003e \u003cp\u003eWe wish to make one final point. We tested five additional hypotheses about specific circumstances in which issue voting should be more or less important, all of which were based on a view of humans as \u003cem\u003ecognitively limited information processors\u003c/em\u003e who continually utilize all sorts of cognitive shortcuts and heuristics to help them make sense of their worlds (Fiske and Taylor, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e1991\u003c/span\u003e; Lau and Redlawsk, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Lau and Sears, 1986; Lodge, et al., 1988). At one point in the not too distant past, the usefulness of this cognitive perspective for political science was strenuously debated (see for example Kuklinski, Luskin, and Bolland, 1991, and the several responses that immediately followed). We found significant support for four of those five hypotheses, most of which would not even have been offered without this cognitive-limitations viewpoint. There should be little controversy today that an information processing view of human cognition is an invaluable tool for generating testable hypotheses about political behavior.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e: Financial support for this research was provided by the first author\u0026rsquo;s university.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eCompliance with Ethical Standards\u003c/strong\u003e: This project, which involves secondary data analysis only, was deemed exempt from full review and approved by the Human Research Protection Program at REDACTED.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e: The authors will make all data and script files freely available after this paper has been accepted for publication.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eDisclosures\u003c/strong\u003e: The authors affirm that they have no competing interests, financial or non-financial, that are directly or indirectly related to the work submitted for publication.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003ePre-Registration\u003c/strong\u003e: No formal pre-registration of the hypotheses tested in this paper has been submitted. \u003c/p\u003e\n"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbramowitz, A. I., \u0026amp; Steven Webster (2016). The Rise of Negative Partisanship and the Nationalization of US Elections in the 21st Century. \u003cem\u003eElectoral Studies\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(1), 12\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAldrich, J. H., Jamie, L., Carson, Brad, T., Gomez, \u0026amp; Merolla, J. L. (2022). \u003cem\u003eChange and Continuity in the 2020 Elections\u003c/em\u003e. Roman \u0026amp; Littlefield.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlvarez, M. R., and Jacob Morrier (2025). Do Polarized Issues Carry More Weight in Voter Decision-Making? Insights from the 2022 Congressional Midterm Elections. \u003cem\u003eAmerican Politics Research\u003c/em\u003e, \u003cem\u003e53\u003c/em\u003e(September), 481\u0026ndash;498.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnsolabehere, S., and Shiro Kuriwaki (2022). Congressional Representation: Accountability from the Constituent\u0026rsquo;s Perspective. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e66\u003c/em\u003e(January), 123\u0026ndash;139.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnsolabehere, S., SnyderJr., J. M., \u0026amp; Charles Stewart, I. I. I. (2001). Candidate Positioning in U.S. House Elections. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(January), 136\u0026ndash;159.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBailey, M. A. (2007). \u0026lsquo;Comparable Preference Estimates Across Time and Institutions for the Court, Congress, and Presidency\u0026rsquo;. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e51\u003c/em\u003e(July), 433\u0026ndash;448.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eButtice, M. K., and Walter J. Stone (2012). Candidates Matter: Policy and Quality Differences in Congressional Elections. \u003cem\u003eJournal of Politics\u003c/em\u003e, \u003cem\u003e74\u003c/em\u003e(July), 870\u0026ndash;887.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCanes-Wrone, B., \u0026amp; Brady, D. W., and John F. Cogan (2002). Out of Step, Out of Office: Electoral Accountability and House Members\u0026rsquo; Voting. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e96\u003c/em\u003e(March), 127\u0026ndash;140.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarson Jamie, L., Sievert, J., \u0026amp; Ryan, D. W. (2020). Nationalization and the Incumbency Advantage. \u003cem\u003ePolitical Research Quarterly\u003c/em\u003e, \u003cem\u003e73\u003c/em\u003e(1), 156\u0026ndash;168.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCayton, A., and Ryan Dawkins (2022). Incongruent Voting or Symbolic Representation? Asymmetrical Representation in Congress, 2008\u0026ndash;2014. \u003cem\u003ePerspective on Politics\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(September), 916\u0026ndash;930.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClinton, J. D., \u0026amp; Jackman, S., and Douglas Rivers (2004). \u0026lsquo;The Statistical Analysis of Roll Call Data\u0026rsquo;. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e98\u003c/em\u003e(May), 355\u0026ndash;370.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConverse, P. E. (1964). The Nature of Belief Systems in Mass Publics. In E. David, \u0026amp; Apter (Eds.), \u003cem\u003eIdeology and Discontent\u003c/em\u003e (pp. 206\u0026ndash;261). Free.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCosta, M. (2021). Ideology, Not Affect: What Americans Want from Political Representation. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e65\u003c/em\u003e(April), 342\u0026ndash;358.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDowns, A. (1957). \u003cem\u003eAn Economic Theory of Democracy\u003c/em\u003e. Harper and Row.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFiorina, M. P. (1981). \u003cem\u003eRetrospective Voting in American National Elections\u003c/em\u003e. Yale University Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFiske, S. T., \u0026amp; Taylor, S. E. (1991). Social Cognition (2nd. edition). New York: McGraw-Hill.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGerber, E. R., \u0026amp; Lewis, J. B. (2004). \u0026lsquo;Beyond the Median: Voter Preferences, District Heterogeneity, and Political Representation\u0026rsquo;. \u003cem\u003eJournal of Political Economy\u003c/em\u003e, \u003cem\u003e112\u003c/em\u003e(December), 1364\u0026ndash;1383.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreen, D. P., and Jonathan S. Krasno (1988). Salvation for the Spendthrift Incumbent: Re-estimating the Effects of Campaign Spending in House Elections. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(November), 884\u0026ndash;907.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGriffin, J. D., Newman, B., \u0026amp; Nickerson, D. W. (2019). A God of Vengeance and of Reward? Voters and Accountability. \u003cem\u003eLegislative Studies Quarterly\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(1), 133\u0026ndash;162.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHenderson, J. A., Sheagley, G., Goggin, S. N., Dancey, L., \u0026amp; Theodoridis, A. G. (2022). Primary Divisions: How Voters Evaluate Policy and Group Differences in Intraparty Contests. \u003cem\u003eJournal of Politics\u003c/em\u003e, \u003cem\u003e84\u003c/em\u003e(July), 1760\u0026ndash;1776.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHighton, B. (2010). The Contextual Causes of Issue and Party Voting in American Presidential Elections. \u003cem\u003ePolitical Behavior\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(December), 453\u0026ndash;472.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHighton, B. (2019). Issue Accountability in U.S. House Elections. \u003cem\u003ePolitical Behavior\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(2), 349\u0026ndash;367.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHollibaugh, G. E., Rothenberg, L. S., \u0026amp; Rulison, K. K. (2013). Does It Really Hurt To Be Out of Step? \u003cem\u003ePolitical Research Quarterly\u003c/em\u003e, \u003cem\u003e66\u003c/em\u003e(4), 856\u0026ndash;867.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHopkins Daniel, J. (2018). \u003cem\u003eThe Increasingly United States: How and Why American Political Behavior Nationalized\u003c/em\u003e. University of Chicago Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobson, G. C., \u0026amp; Carson, J. L. (2020). \u003cem\u003eThe Politics of Congressional Elections\u003c/em\u003e, 10th edition. New York: Roman \u0026amp; Littlefield.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobson, G. C., and Samuel Kernell (1983). \u003cem\u003eStrategy and Choice in Congressional Elections\u003c/em\u003e (2nd ed.). Yale University Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJesse, S. A. (2009). Spatial Voting in the 2004 Presidential Election. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e103\u003c/em\u003e(February), 59\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJoesten, D. A., and Walter J. Stone (2014). Reassessing Proximity Voting: Expertise, Party, and Choice in Congressional Elections. \u003cem\u003eJournal of Politics\u003c/em\u003e, \u003cem\u003e76\u003c/em\u003e(July), 740\u0026ndash;753.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuklinski, J. H., \u0026amp; Luskin, R. C., and John Bolland (1991). Where is the Schema? Going Beyond the S Word in Political Psychology. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e85\u003c/em\u003e(December), 1341\u0026ndash;1356.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLau, R. R., \u0026amp; Redlawsk, D. P. (2006). \u003cem\u003eHow Voters Decide: Information Processing During Election Campaigns\u003c/em\u003e. Cambridge University Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLau, R. R., David, O., \u0026amp; Sears (Eds.). (1986). \u003cem\u003ePolitical Cognition: The 19th Annual Carnegie Symposium on Cognition\u003c/em\u003e. Hillside, NJ: Erlbaum.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLax, J., and Justin Phillips (2009). \u0026lsquo;How Should We Estimate Public Opinion in the States?\u0026rsquo;. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e53\u003c/em\u003e(January), 197\u0026ndash;121.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLodge, M., \u0026amp; McGraw, K. M., and Patrick Stroh (1989). An Impression-Driven Model of Candidate Evaluation. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e83\u003c/em\u003e(June), 399\u0026ndash;420.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLodge, M., \u0026amp; Steenbergen, M. R., and Shawn Brau (1995). The Responsive Voter: Campaign Information and the Dynamics of Candidate Evaluation. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e89\u003c/em\u003e(June), 309\u0026ndash;326.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMann, T. E., \u0026amp; Wolfinger, R. E. (1980). Candidates and Parties in Congressional Elections. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e74\u003c/em\u003e(September), 617\u0026ndash;632.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMansbridge, J. (2003). \u003cem\u003eRethinking Representation American Political Science Review\u003c/em\u003e, 97 (November): 515\u0026ndash;528.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiller, W. E., \u0026amp; Stokes, D. E. (1963). Constituency Influence in Congress. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e57\u003c/em\u003e(March), 45\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMummolo, J., \u0026amp; Peterson, E., and Sean Westwood (2021). The Limits of Partisan Loyalty. \u003cem\u003ePolitical Behavior\u003c/em\u003e, \u003cem\u003e43\u003c/em\u003e(September), 949\u0026ndash;972.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetrocik, J. R. (1996). Issue Ownership in Presidential Elections, with a 1980 Case Study. \u003cem\u003eAmerican Journal of Political Science\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e(August), 825\u0026ndash;850.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRabinowitz, G., Stuart Elaine, \u0026amp; MacDonald (1989). A Directional Theory of Issue Voting. \u003cem\u003eAmerican Political Science Review\u003c/em\u003e, \u003cem\u003e83\u003c/em\u003e(March), 93\u0026ndash;121.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaudenbush, S., Anthony Bryk, and, \u0026amp; Congdon, R. (2022). \u003cem\u003eHLM Hierarchical Linear and Nonlinear Modeling\u003c/em\u003e. Scientific Software International.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRogowski, J. C. (2016). Voter Decision-Making with Polarized Choices. \u003cem\u003eBritish Journal of Political Science\u003c/em\u003e, \u003cem\u003e48\u003c/em\u003e(1), 1\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSears, D. O. (1975). Political Socialization. In F. I. Greenstein, \u0026amp; N. W. Polsby (Eds.), \u003cem\u003eHandbook of Political Science, Volume 2, Micropolitical Theory\u003c/em\u003e. Addison-Wesley.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShor, B., \u0026amp; Rogowski, J. C. (2018). Ideology and the US Congressional Vote. \u003cem\u003ePolitical Science Research and Methods\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(April), 323\u0026ndash;341.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTausanovitch, C., and Christopher Warshaw (2013). Measuring Constituent Policy Preferences in Congress, State Legislatures, and Cities. \u003cem\u003eJournal of Politics\u003c/em\u003e, \u003cem\u003e75\u003c/em\u003e(April), 330\u0026ndash;342.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWarshaw, C. (2019). Local Elections and Representation in the United States. \u003cem\u003eAnnual Review of Political Science\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e, 461\u0026ndash;479.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWarshaw, C., and Jonathan Rodden (2012). \u0026lsquo;How Should We Measure District-Level Public Opinion on Individual Issues?\u0026rsquo;. \u003cem\u003eJournal of Politics\u003c/em\u003e, \u003cem\u003e74\u003c/em\u003e(January), 203\u0026ndash;219.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWright, Gerald, J. (1978). Candidates\u0026rsquo; Policy Positions and Voting in U.S. Congressional Elections. \u003cem\u003eLegislative Studies Quarterly\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(August), 445\u0026ndash;464.\u003c/span\u003e\u003c/li\u003e \u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e Another set of researchers have gathered data on the general ideological positions of competing Congressional candidates as a substitute for evidence about most specific policy stands. For example, Buttice and Stone (2012; see also Joesten and Stone, 2014) estimate the ideologies and personal qualities of a set of major party candidates running in 155 congressional elections in 2006. This paper is an important step in the right direction in that the authors gather data about the competing candidates in a set of congressional elections (see also Hollibaugh, Rothenberg, and Rulison, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), but their measure of ideology based on experts\u0026rsquo; global judgement of ideology is at best an imperfect instrument for a candidate\u0026rsquo;s actual policy positions, as ideology has considerable symbolic value as well as conveying policy information (e.g., Cayton and Dawkins, 2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e This test was completed by 72 percent of all major party candidates in 1996. Unfortunately, response rates to this survey, ironically renamed the Political Courage Test, has dropped from 72 percent in 1996 to 48 percent in 2008 and even further to 20 percent by 2016, according to Wikipedia. If these Wikipedia numbers are correct, the 2008 and 2010 congressional elections analyzed by Shor and Rogowski are probably the last with sufficient data from both major party House candidates.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e We asked Sherman to set the Wayback Machine to October 1 of each election year, and then accessed candidates\u0026rsquo; webpage from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.loc.gov/collections/united-states-elections-web-archive/\u003c/span\u003e\u003cspan address=\"https://www.loc.gov/collections/united-states-elections-web-archive/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The appendix provides more detail on the coding of candidates\u0026rsquo; policy stands.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e We employed our estimate of the very liberal to very conservative ideology of the competing candidates because that is the measure employed by Buttice and Stone (2012) in their research. If instead we utilize a measure of the candidate\u0026rsquo;s ideological difference based on the mean of each candidate\u0026rsquo;s policy positions across the ten policy areas, the interaction term becomes considerably larger (-5.43) but again in the wrong direction.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Issue voting, House elections, Congressional elections, democratic representation, information processing, cognitive limitations","lastPublishedDoi":"10.21203/rs.3.rs-8544888/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8544888/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRepresentative democracy is based on the idea that citizens choose the candidates who most closely share their policy views and concerns. There is surprisingly little evidence for this idea from elections for the House of Representatives, however, primarily due to a lack of appropriate data. We overcome the data problems by employing the large samples from CCES surveys, and coding candidates\u0026rsquo; policy positions from the campaign webpages of 394 major party candidates between 2010 and 2020. We find vote choice in House elections is strongly a function of agreement with the candidates on the issues, controlling for many alternative explanations that prior research has found to be important. We test six hypotheses about \u003cem\u003ewhen\u003c/em\u003e issue voting should be more or less important. We know of no prior research on congressional elections that has comparably detailed evidence that clearly demonstrates that congressional elections provide citizens with considerable levels of substantive representation.\u003c/p\u003e","manuscriptTitle":"Issue Voting in U.S. House Elections","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-29 22:03:04","doi":"10.21203/rs.3.rs-8544888/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b0f62245-0a9f-41e6-aa1f-4c229851494e","owner":[],"postedDate":"January 29th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T20:55:00+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-29 22:03:04","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8544888","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8544888","identity":"rs-8544888","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}