Representation When Constituent Opinion and District Conditions Collide

Abstract

When constituent opinion and district conditions point in two different directions, which factor is most influential for representatives who face important legislative roll calls? To address this question, we combine four types of data for the period from 2000 to 2012: key congressional roll call votes, district-level survey data, objective measures of district conditions, and other district demographics. We show (1) that material conditions in a district have an effect on legislative behavior independent of constituents’ opinions; (2) that opinions are not always a better predictor of lawmaker decisions, compared to conditions; and (3) that whether lawmakers tend to reflect constituent opinions or district conditions is a function of the demographic makeup of their districts.

Keywords

representation roll call votes Congress vote choice public opinion

Introduction

In 2007, the new Democratic majority in Congress passed H.R. 976, a reauthorization and $35 million expansion of the Children’s Health Insurance Program (CHIP)—a program that provides health coverage for children in financial need. Congress passed the legislation but failed to override President Bush’s veto. Forty-four Republicans and two Democrats voted against the majority of their party in the attempted override vote.

Many of the party defectors represented what we might term “conflicted districts”—where conditions in the district point to one vote choice and constituents’ opinions point to another. Take, for example, Rick Renzi (AZ-1), a Republican among that party’s defectors. Almost 70 percent of Renzi’s district, according to polls taken at the time, were opposed to CHIP expansion, but as one of the poorer districts in the country, many of its residents stood to directly benefit from it. Renzi defied district opinion and the majority of his party colleagues and voted in favor of the legislation. Alternatively, Republican Greg Walden (OR-2) represented a district that was similarly opposed to CHIP, but was somewhat better off economically than Renzi’s district. Walden voted against the bill. Taken together, these lawmakers were responsive to the divergent material condition of their districts, as their party affiliations were the same and their constituents’ opinions were nearly so.

In the other type of conflicted district, Democrat Carolyn McCarthy (NY-4), represented an affluent district where few constituents would benefit directly from the CHIP bill, but where her constituents were very supportive of the program. Republican Jim Saxton (NJ-3) represented a similarly affluent district but one that was less supportive of CHIP. McCarthy voted yes, Saxton no, reflecting the difference in their constituents’ opinions.¹

Figure 1 plots the CHIP roll call votes for the subset of congressional lawmakers representing conflicted districts. Specifically, we plot the votes of members of Congress (MCs) who represented districts that were, according to public-opinion polls, either more opposed to CHIP than the average district and simultaneously more likely than the average district to benefit from it due to the district’s poverty rate (upper-left quadrant), or hailed from districts that were more supportive of CHIP than average but had a lower-than-average poverty rate and thus were less likely to benefit from the bill’s passage (lower-right quadrant).

Figure 1.

Constituent opinion, district conditions, and roll call voting on the 2007 CHIP extension.

We ask, when district conditions and constituent opinion diverge, which of these does the legislator represent? In the examples illustrated above, some lawmakers seemed to side with public opinion in their district while others followed their district’s objective conditions—and frequently this action was taken in opposition to the overwhelming majority of their party. How widespread are these patterns and how might they be explained?

These are important questions in a representative democracy. Indeed, political philosophers such as Edmund Burke, John Stuart Mill, and Robert Dahl have weighed in on the normative version of this question—how should representatives represent? While extant studies have taught us a great deal about the nature of political representation, they have tended to study the impact of constituents’ concerns and their needs independently from one another. We advance our understanding of the nature of political representation by studying these factors jointly, which heretofore has not been done explicitly or in considerable depth.

Specifically, we analyze 21 high-salience roll call votes with a focus on districts where constituent opinion and objective conditions point legislators in different directions. We examine the U.S. House of Representatives and merge three types of data—roll call votes from the 106th to 112th Congresses (2000–2012) identified by Congressional Quarterly (CQ) as Key Votes, survey data from the 2000 and 2004 administrations of the National Annenberg Election Survey (NAES) and the 2006–2012 administrations of the Cooperative Congressional Election Study (CCES), and measures of congressional district conditions, which are taken from the U.S. Census and other official sources.

Our analysis demonstrates, first, that both constituent opinion and objective conditions predict lawmaker behavior when the two are not aligned, but constituent opinions tend to predominate. Because we know that patterns of representation often vary,² we disaggregate these conflicted districts according to whether they exhibited liberal opinion and conservative conditions or conservative opinion and liberal conditions. Doing so reveals an important dichotomy. In conflicted districts with more liberal opinions, opinion is the better predictor of roll call behavior. In conflicted districts with more conservative opinions, conditions are more predictive. We find evidence that this pattern is in part explained by differences in electoral dynamics across the two types of conflicted districts. More educated and affluent voters are more likely to be represented by legislators whose party affiliations match their constituents’ issue opinions. Less educated voters are less likely to be represented by lawmakers whose partisanship matches their constituents’ opinions. The result is that lawmakers who represent more educated districts are more likely to follow constituent opinion when making voting decisions, while representatives hailing from less affluent districts are more likely to follow objective economic and social indicators when making roll call choices.

Our findings have important implications for understanding the nature and landscape of political representation in contemporary America, the incentives of lawmakers, and the consequences of disparate voting behavior based on political sophistication. We expand on each of these implications in our conclusion.

Background

The degree to which the needs and concerns of citizens are reflected in the actions of legislators is a central democratic concern. Among the many activities of lawmakers that we might use to gauge citizen representation, the roll call vote holds special status. This is where, after all is said and done, the legislator must unequivocally commit to supporting a direction for public policy. So, while we appreciate that representation takes a number of other forms, our focus is the roll call vote.

Among the studies that focus on roll call voting, there are two theoretical orientations (Pitkin 1972). The first, championed by Robert Dahl (1956) and V. O. Key (1961), stresses the critical link between the opinions of constituents and the actions of their representatives. This so-called “delegate” model of representation contends that democracy itself hinges on lawmakers’ reflection of public opinion in their actions. As Key (1961, 7) puts it, “Unless mass views have some place in the shaping of policy, all the talk about democracy is nonsense.”³ The overwhelming majority of these delegate-oriented studies take a “responsiveness” approach to assessing the degree to which the actions of legislators are representative of opinion (Achen 1978; Ansolabehere, Snyder, and Stewart 2001; Erikson and Wright 2013; Griffin and Newman 2005). In this approach, necessitated because roll call votes and district opinion are almost never measured on the same scale (but see Ansolabehere and Jones 2010), voting behavior is modeled as a function of district preferences, with the resultant parameter estimate indicating the degree to which more conservative districts are represented by more conservative-voting lawmakers. The delegate approach to the study of democratic representation has come to dominate the empirical literature (e.g., Achen 1978; Ansolabehere, Snyder, and Stewart 2001; Bartels 1991; Erikson 1978; Erikson and Wright 1980; Erikson, Wright, and McIver 1989; Gerber and Lewis 2004; Griffin 2006; Miller and Stokes 1963). Seldom if ever in these studies is lawmaker voting behavior modeled using both constituent preferences and a separate measure of constituents’ needs, or the conditions in the district.

In contrast, the so-called “trustee” model of representation contends that citizens are better served by elected officials who do not always follow citizens’ opinions. As Edmund Burke famously argued, “Your representative owes you, not his industry only, but his judgment; and he betrays, instead of serving you, if he sacrifices it to your opinion” (Burke [1774] 1981; see also Mill [1861] 1962). If elected officials are not guided by voters’ opinions, what does guide them? Here, we observe two strands of the trustee perspective.⁴ One strand argues that elected officials ought to form their decisions based on what the country as a whole needs (e.g., Doherty 2013; Mansbridge 2003) or based on their own values or principles (Barker and Carman 2012; Mansbridge 2003). In either case, the needs of the district do not much come into play. The other strand of the trustee perspective is more applicable to our aims here. Namely, it contends that elected officials should “do what’s right for the people in [the] district” (Barker and Carman 2012; Mansbridge 2003, 516), irrespective perhaps of constituents’ expressed desires. This is similar to what Mansbridge (2003) terms “anticipatory” representation—identifying what constituents might use to reward or punish their representative in the next election (see also Arnold 1990). Some early empirical studies in this vein focused on whether legislators themselves reported adopting a “role orientation” consistent with the trustee model (Eulau et al. 1959). In the next several decades, “‘Dozens’ (hundreds) of voting studies have since been done, demonstrating the significant correlation between the (economic) interests of the geographic constituency and the voting behavior of its elected representative” (Richardson and Munger 1990, 14; see also Wright, Erikson, and McIver 1985). However, once again, the study of the impact of constituents’ needs, in this case, was isolated from an analysis of the impact of their concerns. Eventually, this attention to the representation of constituents’ needs was virtually abandoned by scholars.⁵ One reason, we speculate, is that the availability of large-volume survey data has rapidly improved, most notably with the creation of the NAES in 2000 and the CCES in 2002.

We are the first to simultaneously study whether legislators are responsive to their constituents’ expressed opinions or to the conditions in their districts. We ask, when constituent opinion and district conditions diverge, which of these does the legislator represent? And what explains differences in lawmaker behavior in this regard?

Theoretical Expectations

We preface our theoretical expectations with a brief detour into conceptualization and terminology. As the focus of the study is the effect of constituent opinions and district conditions on roll call voting when these two influences contradict one another, the lion’s share of our attention will be directed toward “conflicted” lawmakers. Lawmakers representing districts where opinion and conditions point to the same roll call decision do not face the trade-off at the heart of our research question, namely how they vote when conditions and opinion point in opposite directions. We want to know which factor better predicts conflicted representatives’ votes on salient roll calls, and under what conditions.

This would not be illuminating if district conditions and opinions were always in alignment. However, as a spate of recent investigations have explored, individual citizens sometimes voice policy preferences or exhibit electoral behavior at odds with their interests (Bartels 2005; Gelman et al. 2007). This may be owing to multidimensional decision making, especially in the context of voting for a candidate. As well, the conflict may be attributable to “unenlightened self-interest,” or a failure to appreciate the consequences of policies for one’s own welfare (Bartels 2005). Voters may also simply not prioritize material interests over all other considerations. Gelman et al. (2007) ask a related but somewhat different question: why do wealthy states vote for Democrats but wealthy individuals are more likely to vote for Republicans? They find that the relationship between income and voting for Republicans is much stronger in poorer states than in wealthier states. As we show below, we also find that in less wealthy locations, material interest is a stronger predictor of the partisanship of who gets elected.

A critical next step in the analysis is conceptualizing “conflicted” districts. That is, in our view, districts exhibit either conservative or liberal opinion and either “conservative” or “liberal” conditions relating to each roll call, based on whether the value of each factor is higher or lower than some threshold. Districts are conceptualized as conflicted if they are much more liberal than average on one measure and much more conservative than average on the other, or vice versa.

We employ a shorthand to label the two types of conflicted districts. We use the term “limousine liberal” (LL) to denote districts where constituent opinion is more liberal than average while conditions were more conservative than average. Conversely, we use the term “clunker conservative” (CC) for districts where constituent opinion is more conservative than average but conditions are more liberal than average.⁶ Thus, whether a district is conflicted varies with each roll call vote, as we employ a different set of associated district characteristics (constituent opinion and conditions) in each instance.

Next, we turn to developing our theoretical expectations. Recall that our preliminary research question is whether and under what conditions legislators’ votes reflect constituent opinions or district conditions when the two factors point in different directions. There are certainly a host of reasons to expect that legislators will follow constituent opinion in these instances. First, legislators have access to polling, constituent communication, and other information about the opinions in their districts, so they likely have the information needed to be attentive to constituent opinion (Butler and Nickerson 2011; Miler 2010; Verba 1993). With this information at hand, legislators are (in the view of many) “single-minded seekers of re-election,” who benefit by taking popular positions on public policies (Mayhew 1974). Roll call voting is the most visible and recurring tool for legislators to take positions on issues that constituents favor. What is more, we know that legislators who stray from the opinions of their districts fare worse in the next election (Ansolabehere and Jones 2010; Canes-Wrone, Brady, and Cogan 2002; Nyhan et al. 2012), and sometimes much worse, such that they are “unsafe at any margin” (Jacobson 1987; Mann 1978). Moreover, voters sometimes vote retrospectively (see Healy and Malhotra 2013 for a review) and vote based on their issue opinions (e.g., Aldrich et al. 1989), which should yield “like-minded” representatives casting votes consonant with constituents’ opinions (Miller and Stokes 1963). It is no surprise, then, that a plethora of studies find that legislator roll call voting is responsive to the expressed opinions of constituents (Ansolabehere, Snyder, and Stewart 2001; Butler and Nickerson 2011; Erikson 1978;Miller and Stokes 1963).

However, there are reasons to anticipate that legislators may be attentive to district conditions in addition to, or instead of, constituent opinion. For instance, MCs may not have direct knowledge of constituent opinion on every issue, especially issues that were not the focus of a recent election campaign (Sulkin 2011). On these issues, lawmakers may rely on their impressions of the economic and social needs of their district. As well, much of what representatives know about their district’s preferences originates from constituents who are disproportionately members of “issue publics” or those who have intense preferences on an issue (Bishin 2009). It may be that an issue public within a district communicates a preference that is more closely associated with the conditions in the district than it is associated with the majority opinion of the district. Second, even if elected officials are well aware of constituent opinion, and want to maximize their reelection chances, somewhat paradoxically it is not always clear that following constituent opinion will achieve the desired end. For example, Canes-Wrone and Shotts (2007) demonstrate that when an official has private information that her constituents are misguided, and where prior to the next election the policy will be realized, the official maximizes her reelection chances by declining to follow constituent opinion. Finally, there is recent evidence that voters often evaluate candidates not so much based on the policy positions they espouse but, in part, based on whether they appear competent (Lenz 2013). If this is true, politicians who want to remain in office should attend to their perceptions of voters’ needs, to generate impressions of competence, rather than following voters’ articulated policy stances.

In sum, there are substantial reasons to expect that lawmakers will be attentive to the opinions of constituents, but there are also reasons to expect that lawmakers may pay attention to the objective conditions in their districts either in addition to, or instead of, citizens’ opinions. Thus, we are ex ante agnostic about whether we expect to find that constituent opinions or district conditions predominate when legislators cast roll calls.

Our second theoretical expectation is to observe disparate patterns of behavior in the two types of conflicted districts, LL and CC. To develop this prediction, we draw from a literature on citizens’ expectations of their elected representatives (Carman 2007; Griffin and Flavin 2010; Harden 2013). These studies find at the congressional (Griffin and Flavin 2010) and state (Harden 2013) level that whites, the wealthy, and the well-educated tend to prioritize the reflection of their policy opinions in lawmaker roll call voting over other aspects of representation such as constituent casework or securing public funding, recognizing that a legislator’s time to attend to constituents’ concerns is finite. In contrast, some racial and ethnic minority groups, the less wealthy, and the less educated tend to prioritize casework and bringing money home to the district over working to ascertain and reflect constituent opinion. The reason, Griffin and Flavin (2010, 521) contend, is that

more politically informed groups, such as those with higher incomes or educational attainment, might be expected to place greater weight on policy responsiveness since they have invested in obtaining policy-relevant knowledge, feel more competent to provide policy input to political leaders, and are better prepared to monitor elected officials’ implementation of their preferences. (see also Hutchings 2001; Miler 2010; Wolpert and Gimpel 1997).

That is, those who are better off financially may value private goods less and public goods more than do the disadvantaged (Dixit and Londregan 1996). Applying this logic to our context, LL districts are by definition more populated by the well-educated and the wealthy than are CC districts.⁷ Accordingly, our expectation is that LL voters will be more likely to expect that their policy opinions be reflected in the voting behavior of their representatives. Thus, LL districts should tend to be represented by legislators who follow constituent opinion when it conflicts with conditions, while in CC districts, we expect that legislators will tend to follow district conditions.

Developing our third expectation begins with the contention that legislators’ partisan affiliations often channel the relationship between voters and legislator roll call voting. That is, voters often select candidates with the help of partisan cues, combined with voters’ own strong identification with parties (Campbell et al. 1960). In turn, legislators’ party affiliations are perhaps the strongest single predictor of their roll call behavior (Aldrich 1995; Lee 2009; Theriault 2008). Thus, legislators’ party affiliations mediate the relationship between voters and roll call decisions (Bullock and Brady 1983). Next, we know that more affluent and educated citizens are more likely to identify with and support the party that agrees with their opinions (Kinder and Kalmoe 2017; Lau and Redlawsk 2006). Less affluent and educated voters’ partisanship is less congruent with their responses to opinion questions, perhaps because they are more likely to hold “non-attitudes” about more issues. Thus, while these voters are partisans, their partisanship is less congruent with their stated opinions. We posit that, instead of opinion, district conditions predict legislator partisanship in less affluent and educated districts. In sum, we expect that in LL districts, partisan affiliations mediate the link between district opinion and legislative voting, while in CC districts, partisan affiliations mediate the link between district conditions and legislative voting.

In sum, we proceed to our empirical tests with three expectations. First, we are agnostic about whether, and to what extent, constituent opinion and district conditions impact roll call voting, when these factors conflict and their effects are estimated together. Next, we anticipate that district demographics will affect whether a lawmaker’s votes are impacted by constituent opinion or district conditions. Finally, we expect that the effect of constituent opinion and district conditions on lawmaker roll call voting will be mediated by lawmaker party, and, thus, that the bases of mass voting will diverge in the two district types.

Data

To compare the effects of constituent opinion and district conditions on legislative behavior, we identified substantively significant roll call votes for which both types of measures were available. For purposes of this study, we focus on very high-profile legislative decisions, and thus draw our roll calls from the universe of CQ Key Votes from 2000 to 2012 (the 106th–112th Congresses).⁸ These votes are identified by the editors of CQ Press as “a matter of major controversy, a test of presidential or political power, a decision of potentially great impact on the nation and lives of Americans” (CQ Congress Collection 2014).⁹ We end our analysis in 2012 due to the decennial redistricting. From the pool of CQ Key Votes, a roll call had to meet two further criteria to be included in our analysis: availability of (1) a measure of district geocoded public opinion on the issue, and (2) a measure of the district’s material stake in the outcome of the vote. For an enumeration of all measures used in the analysis, see the online appendix.

We first had to be able to identify a closely related public-opinion survey item asked of a large enough sample to estimate public opinion at the congressional district level.¹⁰ Survey items had to either ask directly about the roll call in question or for the respondent’s opinion on the same issue. Potentially qualifying survey items were identified by the authors from the NAES and the CCES. These surveys are of sufficient size—averaging 125 respondents per district—to construct reliable district-level opinion measures (Griffin and Newman 2008). Roll calls were only retained if a sufficiently related survey item was available.¹¹

In virtually every instance, we require that the survey instrument had to have been conducted prior to the roll call vote in question, so responses could not be affected either by the outcome of the vote or the specific actions of a respondent’s representative. While it was usually the case that survey results were not released until well after the relevant roll call vote—and we are under no illusions that NAES or CCES findings received any significant media attention—we operate from the assumption that MCs are often aware of public sentiment in their constituencies regarding high-profile matters (Clausen, Holmberg, and deHaven-Smith 1983), although perceptions of opinion may be imperfect (Butler and Dynes 2016; Miler 2010). All survey responses were aggregated by congressional district to create a district-level measure of opinion. Prior studies have shown that even though the NAES is not designed to be representative of congressional districts, it reasonably approximates the demographics of these units (Griffin and Newman 2008; see also Carsey and Harden 2010). The CCES is explicitly designed to be representative of congressional districts (Ansolabehere and Persily 2008; Levendusky and Pope 2010), and CCES congressional district-level estimates of presidential support are highly correlated with actual presidential election results at the district level (Tausanovitch and Warshaw 2013).

Second, for a vote to be included in the analysis, a measure of district conditions that reflects an objective constituent interest in the outcome of the vote had to be available. These district conditions are usually economic in nature, such as the district poverty rate for a minimum wage vote or the percent of the district with health insurance for the Affordable Care Act. Most data on district conditions are drawn from the U.S. Census Bureau’s Economic and Decennial Censuses and the American Community Survey. In one instance, we used aggregated survey items that were designed to elicit objective facts about the household for our measure of district conditions (see Online Appendix A for an extended discussion).

After culling roll calls for which either a matched opinion or condition measure was not available, we were left with twenty-one key votes between 2000 and 2012. Not unexpectedly, all votes that met the criteria for inclusion are among the most high-profile congressional actions of this period. For our purposes, this ensures that we have a fairly good representation of issues that are likely to be in the public discussion for MCs and their constituents. Virtually all the votes we employ are on the final passage of legislation, so we can be more confident that legislators’ votes are a sincere expression of the position for which they want to be “on the record.”

We recode all variables so that higher values indicate greater conservatism. Thus, opinions such as opposition to abortion, opposition to same-sex marriage, support for lowering taxes, opposition to a minimum wage increase, and support for repealing the estate tax were considered conservative positions. District conditions were oriented such that higher values would produce a greater benefit from a conservative outcome. For example, the poverty rate and the percentage of people without health insurance were subtracted from 100 to create measures of “percent not poor” and “percent with health insurance.” Each opinion and conditions measure was mean-centered at zero and rescaled to standard deviation units so that the different roll calls could be analyzed in a pooled model. Districts are identified as conflicted if they possess liberal opinion and conservative conditions or vice versa, as long as at least one of the measures (either opinion or conditions) is at least a half of a standard deviation greater or lesser than the mean.¹² This ensures that we are not identifying districts as “conflicted” when in fact they are quite close to average on both measures.¹³

We recognize that it might seem odd to say that a district is conflicted when, for example, it is much less supportive of gay marriage than the national average level of support, but where the Census indicates that 2 percent of district households include a same-sex couple, which is quite a bit higher than the national average but still a modest proportion of the district population. However, because there are no districts where 60 percent, or 40 percent, or even 20 percent of households report being gay, the politically relevant demographic “space” on this issue divides districts where very few openly gay people live and districts where a small percentage of households report being same sex to the census. In addition, the latter type of district is one in which residents more regularly have openly gay relatives, encounter gay friends and neighbors, and have personal and economic ties to gay people. Still, to be sure that this decision was not driving our findings, we also experimented with other approaches for establishing conflicted-district thresholds, which did not affect the conclusions we reach below.¹⁴

Finally, all roll call votes are coded with 1 as the conservative position (as judged by the position of the President and the President’s party affiliation) and 0 as the liberal position. Consequently, positive parameter estimates in the following analyses indicate greater responsiveness to district characteristics and opinions in the expected direction for each vote.

Figure 2 reflects the geographic distribution of districts according to how frequently they are LL or CC, beginning in the 110th Congress due to redistricting. The most frequent LL districts (shown in black and charcoal gray) are concentrated in New England, the upper Midwest, and the Pacific Coast, and include many areas that tend to be affluent and politically liberal such as San Francisco, California; Boulder, Colorado; and Chapel Hill, North Carolina. The entire states of Connecticut and the one congressional district comprising Vermont fall into this category as well. The most frequent CC districts (shown in gray and light gray) are concentrated in the Rocky Mountain West and Deep South. These tend to be less affluent areas that are also politically conservative such as northern Louisiana, Oklahoma, much of New Mexico, and Idaho.

Figure 2.

Geographic distribution of conflicted districts (110th–112th Congress).

While Figure 2 is informative on the question of where LL and CC districts tend to be located, it does not clearly indicate whether there are districts that are LL on some votes and CC on other votes. As will become clear, if such “flip-flopping” were prevalent, it would undermine our interpretation of the findings. However, district flip-flopping is rather uncommon. For example, in the 110th to 112th Congresses (a period with stable district boundaries), 176 (of 435) districts are in both groups at least once across the 11 roll call votes, but during this period, less than 10 percent (only thirty-five) districts were in both groups at least twice.¹⁵

Finally, while it is tempting to conclude from Figure 2 that LL districts tend to be located in politically “blue” parts of the country and CC districts tend to be located in “red” parts of the country, this is only partially true. Only 60.3 percent of CC districts are represented by Republicans in Congress, and only 62.8 percent of LL districts are represented by Democrats.

Do Legislators Represent Opinions or Conditions?

We are first interested in whether constituent opinions or district conditions predominate when they are not aligned, and lawmakers face a salient vote. The first model in Table 1 employs constituency opinion and district conditions to predict conservative votes by lawmakers who represent districts that are conflicted in either an LL or CC configuration. Standard errors are clustered by MC, and a dummy variable is included for each roll call (unreported to conserve space).¹⁶ In addition to the parameter estimates, we report the discrete change in the predicted probability of a legislator taking a conservative position resulting from a half standard deviation change in the independent variable. What stands out in the results is that the effect of opinion is substantively much larger than that of conditions, though conditions are also independently associated with roll call votes. A lawmaker representing a district half a standard deviation more conservative than the mean has a 22.3 percentage point greater probability of casting a conservative vote than lawmakers representing a district with average opinion. A commensurate difference in conditions yields a predicted 8-percentage point change in the probability of casting a conservative vote. To ensure that these effects were not being driven by just a few of the 21 roll calls analyzed, we fit separate models for each roll call. Again, we find that opinions consistently outpace conditions as a predictor of roll call voting (see Online Appendix B).

Table 1.

Effect of Opinion and Conditions, All Conflicted Districts.

Conservative vote
	Coefficient (SE)	1/2 SD effect	Coefficient (SE)	1/2 SD Effect
Opinion	0.883**	0.223	0.671**	0.166
Opinion	(0.071)		(0.101)
Conditions	0.320**	0.080	0.266**	0.064
Conditions	(0.064)		(0.083)
Republican	—		5.644**	0.596
Republican	—		(0.231)
Intercept	0.960**		−0.574**
Intercept	(0.194)		(0.251)
n	3,215		3,215
Wald χ²	330.684**		978.606**
PRE	33.778		81.245

Values are logit coefficients with standard errors (clustered by lawmaker) in parentheses. To the right of each model are the changes in the predicted probability of a conservative vote resulting from a half standard deviation change in the independent variable. Predicted changes for “Republican” are the differences in the predicted probabilities of Democrats and Republicans. PRE = percent reduction in error.

p < .05.

Substantive comparisons of our results with extant studies can be difficult to draw because many roll call studies model aggregate patterns of voting (e.g., Americans for Democratic Action [ADA] and Nominal Three-Step Estimation[NOMINATE] scores), and/or they do not report marginal effects. The effect sizes we observe are akin to those others have observed.¹⁷

Others may interpret these findings somewhat differently. Specifically, because the conditions in a district are likely to impact the opinions of its residents, opinions could be an intervening variable between conditions and legislator voting. In response, first note that when we model roll call voting using just district conditions (i.e., omitting opinion) in CC and then LL districts, we obtain substantively similar parameter estimates.¹⁸ Also, this concern would not be cause to argue that the observed effect of conditions relative to the effect of opinion was deflated relative to its true value. The reason is that conditions and opinion are by definition negatively correlated in these data—and in Table 1, we are only examining districts where conditions and opinion diverge.¹⁹ Thus, any endogeneity of constituent opinions with respect to district conditions has the effect not of attenuating but of amplifying the observed effect of conditions.

When lawmaker party affiliation is added to the model, the estimated effects of both opinion and conditions decline, indicating that both of these factors affect lawmaker roll call voting indirectly, via MC party affiliation, but also directly, as both remain statistically significant and substantively important. The marginal, direct effect of constituent opinion is 16.6 percentage points, and the marginal, direct effect of district conditions is 6.4 percentage points.

Divergent Patterns of Representation

Recall that our first set of analyses pooled two types of conflicted districts, those where opinions were more liberal than average and conditions more conservative (LL), and the opposite (CC). Our next inquiry is whether the patterns of representation differ across these two types of conflicted districts.

As a preliminary matter, we first verified our expectations developed above about the demographic makeup of LL and CC districts. Specifically, we compared LL and CC districts in their proportion of residents who hold at least a bachelor’s degree, the district’s median income, and the proportion of the residents who are non-Hispanic whites. Our expectations were largely confirmed for education and income. For fifteen of the twenty-one votes, LL districts were disproportionately likely to be populated by college graduates. For twelve of the eighteen votes in which income is included in the model (see note to Online Appendix Table C-1), LL districts were wealthier than CC districts.²⁰

Next, we test our expectation that in LL districts, legislators’ votes will more likely be associated with district opinion, while in CC districts, they will more likely be associated with district conditions. To do so, we disaggregate the data used in Table 1 into the two types of conflicted districts. Table 2 reports the results of models with the same specification as those reported in Table 1, now analyzing lawmakers representing LL and CC districts separately (for separate models on each roll call, see Online Appendix B). Focusing first on LL districts, it is particularly striking that only opinion, and not conditions, has a significant effect in these districts. A half standard deviation increase in opinion is associated with a 14.3 percent greater probability of casting a conservative vote. We can also observe this in disaggregated models, where opinion is positively associated with lawmaker behavior in thirteen of the twenty-one votes, while conditions predict behavior in just two votes (see Online Appendix Table B-2).

Table 2.

Effect of Opinion and Conditions, Two Types of Conflicted District.

	Conservative vote: LL districts				Conservative vote: CC districts
	Coefficient (SE)	1/2 SD effect	Coefficient (SE)	1/2 SD effect	Coefficient (SE)	1/2 SD effect	Coefficient (SE)	1/2 SD effect
Opinion	0.993**	0.143	0.387**	0.053	0.466**	0.058	0.394**	0.035
Opinion	(0.118)		(0.185)		(0.136)		(0.190)
Conditions	0.110	0.015	0.475**	0.060	0.830**	0.116	0.539**	0.054
Conditions	(0.131)		(0.162)		(0.149)		(0.144)
Republican	—		6.458**	0.617	—		5.202**	0.406
Republican	—		(0.355)		—		(0.287)
Intercept	1.551**		−0.985**		1.205**		−0.100
Intercept	(0.292)		(0.394)		(0.342)		(0.405)
n	1,704		1,704		1,511		1,511
Wald χ²	242.003**		455.518**		150.764**		509.450**
PRE	19.596		75.638		−3.899		73.413

Values are logit coefficients with standard errors (clustered by lawmaker) in parentheses. To the right of each model are the changes in the predicted probability of a conservative vote resulting from a half standard deviation change in the independent variable. Predicted changes for “Republican” are the differences in the predicted probability of Democrats and Republicans. LL = limousine liberal; CC = clunker conservative; PRE = percent reduction in error.

p < .05.

Contrast these findings for LL districts with those for CC districts, those that exhibit more conservative opinions and liberal conditions than average. The results are nearly a mirror image of what we see among LL districts. In CC districts, increasingly conservative district conditions are associated with a greater likelihood of a conservative vote by the MC (a half standard deviation increase from the mean is associated with an 11.6 percent greater probability of the lawmaker taking a conservative position). Although constituent opinion is also significant in these districts, the magnitude of the effect is much less than conditions in CC districts and less than the estimated effect of opinion in LL districts. Disaggregating the votes reinforces this conclusion—conditions are positively associated with lawmaker votes in thirteen of twenty-one roll calls while opinion predicts roll call voting in just four votes (see Online Appendix Table B-3). Analyzing the two varieties of conflicted districts separately shows that the manner of representation differs between the two types of constituencies.²¹

Before we discuss the models that incorporate MC party, it is important to recognize that these party affiliations are endogenous to constituent opinion and district conditions (see, for example, Bullock and Brady 1983). More liberal districts are more likely to identify with the Democratic Party and to elect Democratic candidates, and the reverse is true of conservative districts.²² In addition, districts that are less white, are more urban, and are lower income tend to elect Democrats. Thus, the party affiliations of lawmakers are properly conceived of as intervening variables in models that also include district characteristics to model roll call decisions.

The party-incorporating models reported in Table 2 lend some support to this notion. When the representative’s party is included in the model of LL districts, the estimated effect of opinion is reduced by nearly two-thirds, although there remains a direct effect. The change in the predicted probability of a conservative vote from a half standard deviation increase in opinion from the mean is 5.3 percentage points. The estimated effect of conditions actually increases when party is accounted for, to a 6 percentage point change in the predicted probability. The reason is that in LL districts, conditions are slightly negatively correlated with the representative’s party. It appears that conditions have a modest direct effect on roll call voting in LL districts, and that opinion has a large effect that is mostly mediated through the lawmaker’s party. In CC districts, the estimated, direct effect of conditions resulting from a marginal shift is halved from 11.6 percentage points to 5.4 percentage points, and the estimated effect of opinion is reduced to 3.5 percentage points. The effect of conditions is greater in CC districts and, unlike LL districts, appears to be substantially mediated by the party of the legislator that voters choose to represent them.

In sum, constituent opinions predict lawmaker roll call voting better than do district conditions, but only in LL districts. Legislators’ party affiliations are a more important mediator of constituent opinion in LL than CC districts and a more important mediator of district conditions in CC than LL districts.²³ Next, we test our explanation for these disparate patterns of representation.

Constituent Opinion, District Conditions, and Lawmaker Party

To explain why party plays a different role in the two types of conflicted districts requires that we return to an expectation we developed above, namely, that the opinions of educated and affluent voters should be well reflected by the party of their representative, with a weaker connection between opinion and representative party for less educated and affluent voters. To test this explanation of the above findings, we model representatives’ party affiliations as a function of public opinion and conditions, and interactions of each of these factors with the percentage of the district possessing at least a four-year college degree.²⁴ We control for the percentage of the district self-identifying as white, the percentage living in an urban area, the incumbent’s share of the vote in the previous election, and fixed effects for each congressional term. The results of this model comport with our expectations (see Table 3).

Table 3.

Models of Legislator Partisanship.

	Republican MC
	Coefficient (SE)
Average Opinion	−0.155
Average Opinion	(0.277)
Average Conditions	0.949**
Average Conditions	(0.315)
% College Grad	0.017
% College Grad	(0.013)
Average Opinion × % College Grad	0.037**
Average Opinion × % College Grad	(0.010)
Average Conditions × % College Grad	−0.014
Average Conditions × % College Grad	(0.012)
% White	0.063**
% White	(0.008)
% Urban	0.002
% Urban	(0.007)
South	1.048**
South	(0.278)
Northeast	−0.509
Northeast	(0.275)
West	0.523
West	(0.313)
Intercept	−5.478**
Intercept	(1.025)
n	3,014
Wald χ²	310.818**
PRE	48.174

Values are logit coefficients with standard errors, clustered by district, in parentheses. MC = member of Congress; PRE = percent reduction in error.

p < .05.

We find that when educational attainment is low, conditions predict lawmaker party better than opinion. The significant interaction between Opinion and % College Grad shows that educational attainment amplifies the association between opinion and legislator party affiliation. In contrast, the insignificant interaction between district conditions and the percentage of college graduates initially seems to suggest that the effect of conditions is constant across districts with varying levels of aggregate education (but see below).

Figure 3 illustrates the magnitude of these effects and reveals the role that conditions play in affecting legislators’ party affiliations. The left-hand panels present the marginal effects of opinion (top) and conditions (bottom) on the party of the MC across the observed levels of college education. The upper-left panel shows that the effect of opinion on MC party is nearly three times greater in districts where more than 60 percent of residents over the age of 25 have a college degree than in districts where the college education rate is 20 percent. In highly educated districts, constituent opinion is a particularly strong predictor of MC party, much stronger than in less educated districts. In contrast, the effect of conditions on MC party declines as educational attainment improves. As educational attainment rises from low to moderate values, the importance of conditions on MC party reliably declines. The right-hand panels in Figure 3 display the predicted probability of electing a Republican for the most and least educated districts across 95 percent of the range of opinion (top) and conditions (bottom). For opinion, the difference made by education is stark. Among the most educated districts, those with conservative residents are virtually guaranteed a Republican representative, and those with liberal residents are equally likely to elect a Democrat. Among the least educated districts, opinion has no discernable impact on the party of the lawmaker. Liberal and conservative districts with low rates of college education are about equally likely to be represented by a Republican.

Figure 3.

Marginal effect of opinion and conditions, by district educational attainment.

These findings show that among more educated districts, opinion predicts lawmaker party affiliation very accurately. This results in voter opinion in more educated districts being better reflected in roll calls. In less educated districts, the effect of opinion is much reduced, and district conditions are a better predictor of a lawmaker’s party affiliation.

Conclusion

We first found that not only do constituents’ opinions predict legislator roll call voting but conditions predict voting as well when opinions and conditions diverge. Next, we found that two patterns of representation emerge in different types of conflicted constituencies. In districts where opinions are liberal and conditions are conservative (LL), public opinion is a very strong predictor of roll call decisions. In districts where opinions are conservative, but conditions point to an interest in liberal policies (CC), the effect of conditions on roll call voting is much stronger. Finally, we find that these divergent patterns are rooted in demographic differences across the two kinds of constituencies. LL districts tend to have higher concentrations of well-educated and wealthy voters, who tend to vote for the party that shares their opinions. CC districts consist disproportionately of less educated and less affluent voters, who are somewhat less likely to vote for the party that reflects their opinions and more likely to opt for the party that aligns with the conditions in the district.

This study yields a series of findings with important implications for our understanding of several spheres of American politics, as well as for the work of political practitioners. First, this research shows that it is not uncommon for constituent opinions and district conditions to contradict each other. Only sixteen districts were not conflicted on any of the roll calls studied between 2003 and 2012, and the overall correlation between conditions and opinion is rather low. Maybe instead of asking “what’s the matter with Kansas” (Frank 2004) or “Connecticut” (Gelman et al. 2007), we should ask “what’s the matter with almost everywhere?”

Second, our finding that LL districts primarily respond to opinion while CC districts chiefly respond to conditions should serve as a reminder that generalizations about how MCs represent their districts can mask important variation in the nature of representation. Future studies of representation should be mindful of this heterogeneity. Third, students of voting behavior will take note of our finding that in some settings, voters rely on their opinions when choosing the party of their representative, while in others, conditions are the deciding factor. These findings mirror individual-level differences between ideological and nonideological voters (Lau and Redlawsk 2006). This last finding also has implications for candidate electoral strategies.

One direction for future research will be to identify somewhat lower salience votes and to repeat our study. Previous research has found greater responsiveness to the public on more salient issues (e.g., Miller and Stokes 1963), so it is possible that the relative importance of opinion and conditions differs on less salient policy questions. Scholars should also investigate the electoral fortunes and institutional positions of representatives from each kind of constituency. For instance, conflicted districts may experience more turnover and may require different electoral strategies, and constituents in the two types of conflicted districts may report having a different relationship with and set of expectations of their representatives. The demographic traits of legislators who represent conflicted districts (or one type) may also differ from those who represent nonconflicted districts. For now, we have learned that representation in Congress is varied but in ways that mirror the expectations of its citizens.

Supplemental Material

Cayton_Replication_Data – Supplemental material for Representation When Constituent Opinion and District Conditions Collide

Supplemental material, Cayton_Replication_Data for Representation When Constituent Opinion and District Conditions Collide by E. Scott Adler, Adam F. Cayton and John D. Griffin in Political Research Quarterly

Supplemental Material

Online_Appendix_online_supp – Supplemental material for Representation When Constituent Opinion and District Conditions Collide

Supplemental material, Online_Appendix_online_supp for Representation When Constituent Opinion and District Conditions Collide by E. Scott Adler, Adam F. Cayton and John D. Griffin in Political Research Quarterly

Footnotes

Acknowledgements

The authors would like to thank Gabrielle Schneiderman for her outstanding research assistance, as well as Paul Quirk, Jason Roberts, Jennifer Victor, and three anonymous reviewers for helpful comments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Supplemental Material

Replication data for this article are available with the manuscript on the Political Research Quarterly (PRQ) website.

References

Achen

Christopher H.

1978. “Measuring Representation.” American Journal of Political Science 22(3): 475–510.

Aldrich

John H.

1995. Why Parties? The Origin and Transformation of Political Parties in America. Chicago: University of Chicago Press.

Aldrich

John H.

Sullivan

John L.

Borgida

Eugene

. 1989. “Foreign Affairs and Issue Voting: Do Presidential Candidates ‘Waltz before a Blind Audience?’” American Political Science Review 83(1): 123–41.

Ansolabehere

Stephen

Jones

Philip Edward

. 2010. “Constituents’ Responses to Congressional Roll Call Voting.” American Journal of Political Science 54(3): 583–97.

Ansolabehere

Stephen

Persily

Nathaniel

. 2008. “Vote Fraud in the Eye of the Beholder: The Role of Public Opinion in the Challenger to Voter Identification Requirements.” Harvard Law Review 121: 1737–74.

Ansolabehere

Stephen

Snyder

James M.

Jr. Stewart

Charles

III . 2001. “Candidate Positioning in U.S. House Elections.” American Journal of Political Science 45(1): 136–59.

Arnold

R. Douglas

. 1990. The Logic of Congressional Action. New Haven and London: Yale University Press.

Barker

David C.

Carman

Christopher J.

2012. Representing Red and Blue: How the Culture Wars Change the Way Citizens Speak and Politicians Listen. Oxford: Oxford University Press.

Bartels

Larry M.

1991. “Constituency Opinion and Congressional Policy Making: The Reagan Defense Buildup.” American Political Science Review 85(2): 457–74.

10.

Bartels

Larry M.

2005. “Homer Gets a Tax Cut: Inequality and Public Policy in the American Mind.” Perspectives on Politics 3(1): 15–31.

11.

Bishin

Benjamin G.

2009. Tyranny of the Minority the Subconstituency Politics Theory of Representation. Philadelphia: Temple University Press.

12.

Bullock

Charles S.

Brady

David W.

1983. “Party, Constituency, and Roll-Call Voting in the U.S. Senate.” Legislative Studies Quarterly 8(1): 29–43.

13.

Burke

Edmund

. (1774) 1981. The Writings and Speeches of Edmund Burke: Volume II: Party, Parliament and the American Crisis, 1766–1774. Edited by Langford

Paul

. Oxford: Oxford University Press.

14.

Butler

Daniel M.

Dynes

Adam

. 2016. American Journal of Political Science 60(4): 975–89.

15.

Butler

Daniel M.

Nickerson

David W.

2011. “Can Learning Constituency Opinion Affect How Legislators Vote? Results from a Field Experiment.” Quarterly Journal of Political Science 6:55–83.

16.

Campbell

Angus

Converse

Philip E.

Miller

Warren E.

Stokes

Donald E.

1960. The American Voter. Chicago: University of Chicago Press.

17.

Canes-Wrone

Brandice

Brady

David W.

Cogan

John F.

2002. “Out of Step, Out of Office: Electoral Accountability and House Members’ Voting.” American Political Science Review 96 (1): 127–40.

18.

Canes-Wrone

Brandice

Shotts

Kenneth W.

2007. “When Do Elections Encourage Ideological Rigidity?” American Political Science Review 101(2): 273–88.

19.

Carman

Chris J.

2007. “Assessing Preferences for Political Representation in the U.S.” Journal of Elections, Public Opinion, and Parties 17(1): 1–19.

20.

Carsey

Thomas

Harden

Jeffrey

. 2010. “New Measures of Partisanship, Ideology, and Policy Mood in the American States.” State Politics & Policy Quarterly 10(2): 136–56.

21.

Cayton

Adam F.

2017. “Consistency vs. Responsiveness: Do Members of Congress Change Positions on Specific Issues in Response to Their Districts?” Political Research Quarterly 70(1): 3–18.

22.

Clausen

Aage R.

Holmberg

Soren

deHaven-Smith

Lance

. 1983. “Contextual Factors in the Accuracy of Leader Perceptions of Constituents’ Views.” Journal of Politics 45(2): 449–72.

23.

CQ: Congress Collection. 2014. Help: Advanced Key Vote Analysis. Washington, DC: Congressional Quarterly Press.

24.

Dahl

Robert

. 1956. A Preface to Democratic Theory. New Haven: Yale University Press.

25.

Davidson

Roger H.

1970. “Public Prescriptions for the Job of Congressman.” Midwest Journal of Political Science 14(4): 648–66.

26.

Dixit

Avinash

Londregan

John

. 1996. “The Determinants of Success of Special Interests in Redistributive Politics.” Journal of Politics 58(4): 1132–55.

27.

Doherty

David

. 2013. “To Whom Do People Think Representatives Should Respond: Their District or the Country?” Public Opinion Quarterly 77(1): 237–55.

28.

Erikson

Robert S.

1978. “Constituency Opinion and Congressional Behavior: A Reexamination of the Miller-Stokes Representation Data.” American Journal of Political Science 22(3): 511–35.

29.

Erikson

Robert S.

Wright

Gerald C.

Jr.

1980. “Policy Representation of Constituency Interests.” Political Behavior 2(1): 91–106.

30.

Erikson

Robert S.

Wright

Gerald C.

2013. “Voters, Candidates, and Issues in Congressional Elections.” In Congress Reconsidered, vol. 10., edited by Lawrence C. Dodd and Bruce I. Oppenheimer, 91–116. Washington, DC: Congressional Quarterly.

31.

Erikson

Robert S.

Wright

Gerald C.

Jr. McIver

John P.

1989. “Political Parties, Public Opinion, and State Policy in the United States.” American Political Science Review 83:729–50.

32.

Eulau

Heinz

Wahlke

John C.

Buchanan

William

Ferguson

Leroy C.

1959. “The Role of the Representative: Some Empirical Observations on the Theory of Edmund Burke.” American Political Science Review 53(3): 742–56.

33.

Fox

Justin

Shotts

Kenneth W.

2009. “Delegates or Trustees? A Theory of Political Accountability.” Journal of Politics 71(4): 1225–37.

34.

Frank

Thomas

. 2004. What’s the Matter with Kansas? How Conservatives Won the Heart of America. New York: Macmillan.

35.

Gelman

Andrew E.

Shor

Boris

Bafumi

Joseph

Park

David

. 2007. “Rich State, Poor State, Red State, Blue State: What’s the Matter with Connecticut?” Quarterly Journal of Political Science 2(4): 345–67.

36.

Gerber

Elisabeth R.

Lewis

Jeffrey B.

2004. “Beyond the Median: Voter Preferences, District Heterogeneity, and Political Representation.” Journal of Political Economy 112(6): 1364–83.

37.

Griffin

John D.

2006. “Electoral Competition and Democratic Responsiveness: A Defense of the Marginality Hypothesis.” Journal of Politics 68(4): 911–21.

38.

Griffin

John D.

Flavin

Patrick

. 2010. “How Citizens and Their Legislators Prioritize Spheres of Representation.” Political Research Quarterly 64(3): 520–33.

39.

Griffin

John D.

Newman

Brian

. 2005. “Are Voters Better Represented?” Journal of Politics 67(4): 1206–27.

40.

Griffin

John D.

Newman

Brian

. 2008. Minority Report: Evaluating Political Equality in America. Chicago: University of Chicago Press.

41.

Harden

Jeffrey J.

2013. “Multidimensional Responsiveness: The Determinants of Legislators’ Representational Priorities.” Legislative Studies Quarterly 38(2): 155–84.

42.

Healy

Andrew

Malhotra

Neil

. 2013. “Retrospective Voting Reconsidered.” Annual Review of Political Science 16:285–306.

43.

Holian

David B.

Krebs

Timothy B.

Walsh

Michael H.

1997. “Constituency Opinion, Ross Perot, and Roll-Call Behavior in the U.S. House: The Case of the NAFTA.” Legislative Studies Quarterly 22(3): 369–92.

44.

Hutchings

Vincent L.

2001. “Political Context, Issue Salience, and Selective Attentiveness: Constituent Knowledge of the Clarence Thomas Confirmation Vote.” Journal of Politics 63(3): 846–68.

45.

Imai

Kosuke

Keele

Luke

Tingley

Dustin

. 2010. “A General Approach to Causal Mediation Analysis.” Psychological Methods 15(4): 309–34.

46.

Jacobson

Gary C.

1987. “The Marginals Never Vanished: Incumbency and Competition in Elections to the U.S. House of Representatives, 1952–82.” American Journal of Political Science 31(1): 126–41.

47.

Jacoby

William G.

2010. “Policy Attitudes, Ideology and Voting Behavior in the 2008 Election.” Electoral Studies 29(4): 557–68.

48.

Key

V. O.

1961. Public Opinion and American Democracy. New York: John Wiley.

49.

Kinder

Donald R.

Kalmoe

Nathan P.

2017. Neither Liberal nor Conservative: Ideological Innocence in the American Public. Chicago: University of Chicago Press.

50.

Lau

Richard R.

Redlawsk

David R.

2006. How Voters Decide: Information Processing during Election Campaigns. Cambridge: Cambridge University Press.

51.

Lazarus

Jeffrey

. 2010. “Giving the People What They Want? The Distribution of Earmarks in the U.S. House of Representatives.” American Journal of Political Science 54(2): 338–53.

52.

Lee

Frances E.

2009. Beyond Ideology: Politics, Principles, and Partisanship in the U.S. Senate. Chicago and London: The University of Chicago Press.

53.

Lenz

Gabriel

. 2013. Follow the Leader? How Voters Respond to Politicians’ Policies and Performance. Chicago: University of Chicago Press.

54.

Levendusky

Matthew S.

Pope

Jeremy C.

2010. “Measuring Aggregate-Level Ideological Heterogeneity.” Legislative Studies Quarterly 35(2): 259–82.

55.

Mann

Thomas E.

1978. Unsafe at Any Margin: Interpreting Congressional Elections. Washington, DC: AEI Press.

56.

Mansbridge

Jane

. 2003. “Rethinking Representation.” American Political Science Review 97(4): 515–28.

57.

Mayhew

David

. 1974. Congress: The Electoral Connection. New Haven: Yale University Press.

58.

Miler

Kristina C.

2010. Constituency Representation in Congress: The View from Capitol Hill. Cambridge: Cambridge University Press.

59.

Mill

John Stuart

. [1861] 1962. Essays on Politics and Culture. New York: Doubleday.

60.

Miller

Warren E.

Stokes

Donald E.

1963. “Constituency Influence in Congress.” American Political Science Review 57(1): 45–56.

61.

Nyhan, Brendan, Eric McGhee, John Sides, Seth Masket, and Steven Greene . 2012. “One Vote out of Step? The Effects of Salient Roll Call Votes in the 2010 Election.” American Politics Research 40 (5): 844–79.

62.

Pitkin

Hanna

. 1972. The Concept of Representation. Berkeley: University of California Press.

63.

Richardson

Lilliard E.

Jr. Munger

Michael C.

1990. “Shirking, Representation, and Congressional Behavior: Voting on the 1983 Amendments to the Social Security Act.” Public Choice 67(1): 11–33.

64.

Sigelman

Lee

Sigelman

Carol K.

Walkosz

Barbara J.

1992. “The Public and the Paradox of Leadership: An Experimental Analysis.” American Journal of Political Science 36(2): 366–85.

65.

Sulkin

Tracy

. 2011. The Legislative Legacy of Congressional Campaigns. New York: Cambridge University Press.

66.

Tausanovitch

Chris

Warshaw

Christopher

. 2013. “Measuring Constituent Policy Preferences in Congress, State Legislatures, and Cities.” The Journal of Politics 75(2): 330–42.

67.

Theriault

Sean M.

2008. Party Polarization in Congress. Cambridge and New York: Cambridge University Press.

68.

Verba

Sidney

. 1993. “The Citizen as Respondent: Sample Surveys and American Democracy Presidential Address, American Political Science Association, 1995.” American Political Science Review 90(1): 1–7.

69.

Wolpert

Robin M.

Gimpel

James G.

1997. “Information, Recall, and Accountability: The Electorate’s Response to the Clarence Thomas Nomination.” Legislative Studies Quarterly 22(4): 535–50.

70.

Wright

Gerald C.

Erikson

Robert S.

McIver

John P.

1985. “Measuring State Partisanship and Ideology with Survey Data.” Journal of Politics 47(2): 453–60.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.32 MB

0.00 MB

0.19 MB