A Paradox in Polarization? Cross-pressured Representatives and the Missing Incentive to Moderate

Abstract

Spanish

Chinese

Members of the public are often left choosing between two extreme candidates who will not represent the moderate, aggregate, public effectively. Cross-pressured members of the U.S. Congress serve a constituency that votes for the opposite party at the national level. If there is any group of representatives that have an incentive to moderate their voting behavior, it is cross-pressured members. In this article, I show that cross-pressured members are more moderate than the average member of their party. This could provide constraints on rampant partisanship in the form of districts that are comfortable electing a representative of one party and voting for the president of the other. However, I show that these members are significantly less likely to be reelected. Thus a paradox exists in which cross-pressured members who moderate their voting behavior are no more likely to be rewarded for behaving the way citizens claim they want to represent.

Keywords

Political Parties Polarization Moderate Moderation United States Congress Representation Cross-pressured Congressmen Partisanship House of Representatives

The American public regularly condemns a political system rife with partisan gridlock. In a 2014 study, close to 50 percent of Americans believed that both political parties needed to compromise more (Pew Research Center 2014). Polarization is distasteful to a broad swath of the public (Klar and Krupnikov 2016). For many, the easiest place to see this hated polarization on display is in the halls of Congress (Levendusky 2009).

A consistent plurality of Americans labels themselves as moderates (Ansolabehere, Rodden, and Snyder 2006). Yet evidence shows that the ideological innocence found by Converse (1964) more than 50 years ago remains the same for most Americans today (Kinder and Kalmoe 2017). Less than 40 percent of Americans have anything to say—positive or negative—about candidates for the House of Representatives (Dalton 2013). At the same time, many Americans know little about the legislative process and do not like how it operates (Hibbing and Theiss-Morse 1995). As a result, one assumes the public would reward members of Congress (MCs) who moderate their behavior.

Congress is more polarized now than in any other modern period, and we see fewer moderates elected. MCs are more extreme than the voters they represent (Gelman et al. 2010), and this trend is only exacerbated by newer members (Bafumi and Herron 2010). Furthermore, some longtime members have become less moderate due to pressure from their district (Nichols 2013).

Growing polarization in the electorate impacts what MCs see and hear from their constituency (Jacobson 2008). Americans now live in echo chambers where it is easy to ignore messages that disagree with one’s predispositions (Levendusky 2013). In many ways, MCs have an electorate that wants both polarization and moderation. Yet Mayhew’s (2004) axiom that MCs are single-minded seekers of reelection remains true. Although national trends are an “act of God” (Mayhew 2004), they are becoming more common. MCs, then, want to do everything in their power to remain in office.

One way of doing this is by attempting to make elections local. A member will focus on casework and constituent service in their reelection campaign (Jacobson 2008). A focus on past performance in the job alleviates the impact of partisanship. Reminding citizens of the incumbent’s beneficial activities helps reelection because the average American pays little attention to politics. However, we know that MCs are influenced by constituents in many ways and respond to constituent ideology (Erikson and Wright 2005). We also know that salient issues impact electoral chances (Gilens 2005; Monroe 1998; Wright and Berkman 1986). Legislators listen to public opinion when it is electorally salient to do so (Hill and Hurley 1999; W. E. Miller and Stokes 1963). When public opinion shifts on an issue, Congress often reacts by changing policy (Bartels 1991; Page and Shapiro 1983; Peterson et al. 2003). MCs will change their behavior to better represent their constituents (Adams 1997; Brady and Schwartz 1995; Gohmann and Ohsfeldt 1994; Karol 2009). Thus, we know that MCs desire reelection and work with that goal in mind. Yet scant research looks at the behavior of members from moderate districts. If constituents sanction members on voting behavior in a moderate district, we would expect MCs to moderate for electoral benefit. Are these members different from a typical partisan because they represent a moderate district?

Recent research has studied those who claim to desire moderate behavior. Independents prioritize compromise more than fighting for policy gains, but leaning partisans punish members from their own party for doing so (Klar and Krupnikov 2016). The answer may be simple in that partisans loathe it when their party compromises, even if they like the idea in theory (Carson et al. 2010; Harbridge and Malhotra 2011; Harbridge, Malhotra, and Harrison 2014). In essence, we claim to abhor the very behavior we punish individuals for not doing (Bauer, Yong, and Krupnikov 2017; Klar and Krupnikov 2016). However, the present article does not look specifically at members who represent districts that most likely declare moderation as a value. Instead, this study looks at the effects of representing a moderate district on an MC’s behavior and chances of reelection. Specifically, it looks at members who are cross-pressured. A cross-pressured member is one whose district votes with the other party for president. Hence, a Democratic representative in a district that voted for McCain or Romney would be cross-pressured. The first goal of this study is thus to test whether these members do indeed moderate their behavior vis-à-vis other members in their party. Second, do these moderate members gain an electoral advantage for moderating?

I find that cross-pressured MCs are more moderate than other members of their party. For example, Democrats who are cross-pressured are more conservative than a typical Democrat in the House. Republicans who are cross-pressured are more liberal than the typical Republican in the House. Cross-pressured members have more difficulty winning reelection, even as they moderate voting behavior. This results in a paradox where those MCs who are less partisan than their peers, representing a district that seemingly values moderation, are less likely than partisan members to win reelection.

The next section of this article outlines the theory and hypotheses. The following section discusses the data and methods. I then present findings from the statistical tests, before concluding.

Theory and Hypotheses

MCs respond to constituency opinion. Yet polarization changes one’s constituency (Abramowitz 2010). Fenno (1978) showed that MCs see several different constituencies rather than one aggregate constituency. Often, MCs tailor remarks for each specific constituency. MCs also represent the policy preferences of constituents (Ansolabehere and Jones 2010; Erikson and Wright 2005; Hill and Hurley 1999; W. E. Miller and Stokes 1963). Thus, it is logical to presume that MCs follow the preferences of the majority of their constituency.

Recent evidence accounting for the rising polarization argues this is not the case. Ansolabehere, Snyder, and Stewart (2001) find that district-level concerns exerted pressure on Congressional candidates in the past. Now, however, the pressure for ideological consistency comes from the national party. Voting behavior in Congress aligns more with the primary electorate than the general electorate (Brady, Han, and Pope 2007; Wright 1989). Bafumi and Herron (2010) show that new MCs are more extreme than previous members, leapfrogging the public. Even in competitive districts, candidates shirk the mean district ideology in favor of activists and interest groups (Powell 1982). In short, one could argue that parties do not care about policy representation. Instead, they care about winning an ideological war with the other party for the public’s support (Hussey and Zaller 2011).

Citizens punish representatives who vote against their interest (Ansolabehere and Jones 2010), but it is often difficult for the average person to know when that has taken place. Political knowledge, which has a low average and high variance (Converse 1990), helps citizens understand policy-specific knowledge (Gilens 2001). When citizens do have knowledge on issues, they use it to make decisions about that policy (Bullock 2011). This is especially true if they have confidence in their knowledge (Gerber et al. 2011). Even with low political knowledge, most Americans have preferences on important bills. And many constituents use these beliefs to keep legislators accountable (Ansolabehere and Jones 2010).

The interests of competing groups lead to disagreements due to a desire for close policy representation. One can focus on winning the support of knowledgeable citizens (Dalton 2013; Gilens 2005). Another related group of individuals an MC draws support from is the wealthy (Bartels 2008; Gilens 2012). Members also often focus on the desires of the primary constituency (Brady, Han, and Pope 2007; Brady and Schwartz 1995; Layman et al. 2010). The consensus view, then, is that these are the most important constituencies for a member of Congress and their electoral chances. This leads representatives to be extreme. As a result, the moderate representation an average American desires does not exist (Bafumi and Herron 2010; G. Miller and Schofield 2003, 2008).

But this is not the end of the story. Americans claim they want moderate representation (Harbridge, Malhotra, and Harrison 2014). Furthermore, they are weary of the gridlock and polarization in Congress. I focus on the group most likely to engage with the other side and behave in moderate ways. Cross-pressured representatives “live in the middle” of party activists, pulling them to ideological extremity, and a district that votes in incongruous ways. Cross-pressured members serve districts that vote for presidential candidates of the opposite party. These members ought to have the most incentive to moderate, as they represent a district whose national-level ideology does not match the MCs party. Insight from other work suggests that these individuals would be no different than other MCs (see, for example, Ansolabehere, Snyder, and Stewart 2001; Bafumi and Herron 2010). In short, these members see growing polarization. Yet they represent a district that votes for the other party’s presidential candidate.

There is evidence, then, that MCs do concern themselves with constituent preferences and that constituents hold members accountable. Some members may engage in particularistic service to reap electoral benefits (Ashworth and de Mesquita 2006). Others may only focus on what the party activists in the district desire (Brady, Han, and Pope 2007). We know that issues matter for elections (Ansolabehere and Jones 2010; Erikson and Wright 2005), and that MCs believe they are constantly vulnerable (King 1997).

But, we simply do not know how these particular members behave (Grofman et al. 2000). There is little evidence with which to know anything about how cross-pressured members differ or align with their copartisans. So we do not know if they are more moderate than their copartisans. We also do not know if they win reelection at the same rate as other members who are not caught in this position. If these members do moderate and win reelection, we can be confident that these districts do indeed reward moderation. However, if cross-pressured members moderate and lose reelection contests, there is little incentive to moderate.

As a result, two main questions drive this study. First, do cross-pressured members of the House moderate their legislative behavior? An answer in either direction is important for the study of Congressional behavior. If cross-pressured members do not moderate, it would most likely be indicative of polarization’s impact. It would represent the effects of partisanship on all members without prejudice. Cross-pressured members should be most likely to assert differences with their party’s platform. If this is true, one can deduce that district-level factors do impact legislative behavior.

I argue here that cross-pressured members are more likely to moderate voting behavior compared with copartisans. As discussed above, legislative behavior impacts how constituents see candidates. These MCs understand they are representing a constituency that values the other party, at least at the presidential level. So, they will be more likely to engage in any activity that encourages their chance of reelection (Mayhew 2004). This leads to the first hypothesis:

Hypothesis 1: Cross-pressured members are likely to be significantly more moderate, in their ideal point estimates, than other members of the same party in their voting behavior. This effect is true for both Democrats and Republican.

Cross-pressured members may seek reelection at a lower rate than those who are not cross-pressured, and I expect the incumbency rate to be lower among this group. Descriptive analysis shows that the average number of terms completed is lower for cross-pressured members in four of the five sessions, and a t test confirms there is a significant difference in the average terms served between cross-pressured and noncross-pressured MCs (as Table 4 shows later in this article). I expect cross-pressured members to be significantly more likely to lose reelection bids than other members seeking reelection, even when controlling for members who are moderate. In other words, it is being cross-pressured and not being moderate that makes one less likely to be reelected. Thus, there is no incentive to moderate for individuals who represent districts that are cross-pressured because moderation does not increase reelection rates. This leads to my second hypothesis:

Hypothesis 2: Cross-pressured members will lose reelection bids more often than other members seeking reelection to the House of Representatives. This is true regardless of whether or not the cross-pressured member moderates voting behavior.

Data and Method

I focus on the 109th to 113th U.S. House in this article. Both parties controlled the House during this period and each party saw the rise of national issues sweeping them into power. The outcomes of the 2006 and 2010 elections were heavily influenced by national waves. Yet the 2008 and 2012 elections had the focus of a presidential election, and 2014 was a low turnout midterm election. As a result, there were a wide variety of sessions and electoral constraints alive during this time. One could make the case that each of these were elections where a rational MC would focus on national ideology (Ansolabehere, Snyder, and Stewart 2001).

The preceding discussion also leads to the expectation that politicians will be prospective. This means they generally know the constituency that will be voting in the next election (Fenno 1978; Hayes, Hibbing, and Sulkin 2010). Thus, for the purposes of this study, I am most interested in how representatives behave before they are cross-pressured. In other words, cross-pressured members know their reelection constituency votes for the other party at the national level. So, I look at behavior in the 109th U.S. House (2005–2006) in relation to the 2006 election. This pattern continues, and I look at members of the 113th U.S. House (2013–2014) leading into the 2014 election.

I measure cross-pressured members through the same method employed by the Cook Partisan Voting Index (PVI). The PVI averages the district-level vote for each party’s presidential candidate over the last two cycles. One then must determine the average district vote for each party. Next, the PVI compares how each district relates compared with the average district over the last two cycles. For instance, assume the Democratic presidential candidate averages 51 percent of the vote over the last two cycles. One would compare how a specific district voted over the same period. Thus, if the average for a district was 55 percent voting for the Democratic candidate, the PVI would be DEM+4.¹ The range of PVI is DEM+43 to GOP+32. The median district is GOP+2, and half of all districts exist between DEM+10 and GOP+10.

Table 1 presents a descriptive view of cross-pressured MCs. In our time of interest, 246 members were cross-pressured, which means they represent a district that favored the other party candidate more than the average district. This represents 11 percent of all members of the House throughout this time. Yet one must also mention the significant decrease in this population starting in the 112th House. Table 1 also indicates that most of the cross-pressured MCs are Democrats.

Table 1

Cross-pressured Membership in the U.S. House.

Variable	109th house	110th house	111th house	112th house	113th house	Total
Cross-pressured	56	62	77	32	20	246
Democrat cross-pressured	27	44	68	20	14	173
Republican cross-pressured	29	18	8	12	6	73

One could argue that cross-pressured members are no different than marginal members. Yet the evidence indicates these are different phenomena. Being a cross-pressured MC predicts legislative behavior better than does a marginal victory. This effect is the same even when looking at being marginal in the previous election. The appendix uses models that incorporate both variables in analysis to demonstrate this point.

Mayhew (1974) defines a marginal district in two specific ways. The strictest measure is when the Democratic two-party vote is between 45 percent and 55 percent for the legislative seat. The more relaxed variable is when the vote is between 40 percent and 60 percent of the two-party vote. There are more marginal representatives than there are cross-pressured MCs. In this series, 297 members won their election with less than 55 percent of the two-party vote, and 644 with less than 60 percent. This compares to 246 members who are cross-pressured.

Arguing that being cross-pressured and being marginal are the same concept falls short when looking at correlations (see Table 2). One notices a weak relationship, at best, between the two concepts. The strongest correlation is .28, which is between being cross-pressured and winning less than 60 percent of the vote. Using Mayhew’s measure of “marginal” leads to an inconsistent relationship with legislative behavior. The relationship between being cross-pressured and legislative behavior is stronger and more consistent. As a result, it seems prudent that research should focus on cross-pressured members rather than marginal districts.

Table 2

Correlation between Marginal and Cross-pressured.

Variable	Marginal 40	Marginal 45	Marginal 40 last election	Marginal 45 last election
Marginal 40
Marginal 45	.615
Marginal 40 last election	.400	.290
Marginal 45 last election	.353	.284	.607
Cross-pressured	.280	.253	.215	.202

I test MC behavior through the use of DW-NOMINATE scores. I conduct initial tests with the ideal point estimates. This allows us to see if there is an impact for being a cross-pressured member. The main goal is to see if cross-pressured MCs deviate from their party in a discernible fashion. Thus, I used the ideal point estimates to derive the median and mean point estimates for each party and each session of Congress (see Table 3). The DW-NOMINATE system uses a positive score to show one is a conservative, and a negative score indicates one is liberal. As Table 3 shows, there are large and increasing differences between the parties’ ideal points.

Table 3

Median and Mean Ideal Point Estimates by Party.

Variable	109th house	110th house	111th house	112th house	113th house
Median GOP point	0.595	0.632	0.686	0.723	0.732
Mean GOP point	0.610	0.641	0.677	0.707	0.723
Median DEM point	−0.368	−0.352	−0.341	−0.393	−0.388
Mean DEM point	−0.365	−0.343	−0.333	−0.384	−0.378
Median difference	0.963	0.984	1.027	1.116	1.112
Mean difference	0.975	0.984	1.010	1.091	1.101

I derive the difference between each representative and the median and mean for their party. The variable is coded so that a negative score means one is more liberal than the median of their party. A positive score indicates that one is more conservative than the median or mean of the party. This is the same for both Democrats and Republicans.

This measurement choice may seem counter-intuitive at first. Yet I do this for a few reasons. First, this is the same classification used by the DW-NOMINATE measure where a negative score is liberal and a positive score is conservative. Second, I test the hypotheses separately for Democrats and Republicans. This means, we expect cross-pressured Democrats to have a positive number as it relates to the Democratic median, but we expect cross-pressured Republicans to have scores below the Republican median. As a result, this choice does not change the outcome of analysis; it only means the expectations for each party is the reverse of each other.

There are several other variables I include in the analysis. First, I create a variable for the number of terms served, with partial terms counted as a whole term. I include this variable because long-serving MCs have stronger reelection chances. This allows me to determine if those who are cross-pressured and long serving behave differently than those who have not been in Congress long. It also allows for testing the relationship between terms served and moderation as well as reelection rates.

Second, I include a dummy variable for political party because of the central role that partisanship has in Congressional voting behavior (McCarty, Poole, and Rosenthal 2008). Third, I create a dummy variable looking at whether a member seeks reelection. One cannot win reelection if they do not seek it. This allows for testing the relationship between seeking reelection and being cross-pressured. Next, I create a dummy variable looking at whether the representative wins reelection. This allows me to compare rates of reelection.

Finally, I create variables that look at the impact of spending on campaigns and challenger quality. We know that campaign spending has increased drastically in recent elections (Jacobson 2008). I look at reported incumbent spending, and the ratio between incumbents and challengers. I also include a dummy variable for having a quality challenger. Quality challengers are more rare than ever before (Carson, Engstrom, and Roberts 2007), but we know they can impact incumbent strategy (Druckman, Kifer, and Parkin 2009). Table 4 shows the descriptive statistics of these variables. For example, 235 out of the 438 who served in the 109th U.S. House were Republicans. The average House member had served 5.8 terms heading into the 2006 election. A total of 404 MCs sought reelection, and 375 were successful, for a retention rate of 93 percent. Only two of the 20 who lost did so in a primary.

Table 4

Descriptive Statistics.

Variable	109th house	110th house	111th house	112th house	113th house	Total109th–113th
GOP served	235	205	183	243	239	1,105
Total in analysis	438	446	444	442	443	2,213
Average House terms served	5.80	5.72	5.78	5.60	5.37	5.73
Number seeking reelection	404	402	397	395	395	1,993
Number winning reelection	375	383	341	356	377	1,832
Number losing in primary	2	4	4	18	5	33
Median incumbent spending	$973,085	$1,081,689	$1,258,857	$1,233,090	$1,114,806	$1,130,621
Median incumbent spending ratio	8.16	8.64	6.06	8.90	10.18	7.83
Number of quality challengers	55	58	81	85	64	343

Findings

Table 5 looks at whether being a cross-pressured MC impacts one’s legislative behavior. I test for the impact of several variables on ideal point estimates. I use a dummy variable for being a Republican and if one is running for reelection and a variable for how many terms one has completed. I use the PVI as a continuous variable because the range is 75 points from DEM+43 to GOP +32. I also use a variable for the money ratio between incumbent and challenger. Due to the existence of heteroskedasticity, I use robust standard errors in each of my models. I tested the models for multicollinearity, and in none of the models is the variance inflation factor (VIF) larger than 2.

Table 5

OLS with Ideal Point as Dependent Variable.

Variable	109th House × Robust Standard Errors	110th House × Robust Standard Errors	111th House × Robust Standard Errors	112th House × Robust Standard Errors	113th House × Robust Standard Errors	109th–113th House × Robust Standard Errors
Cross-pressured	0.001(0.017)	0.010(0.023)	0.154*** (0.023)	0.050** (0.024)	0.064** (0.025)	0.058*** (0.012)
GOP	0.833*** (0.015)	0.850*** (0.020)	1.029*** (0.017)	0.976*** (0.019)	0.968*** (0.020)	0.948*** (0.008)
Terms completed	−0.007*** (0.001)	−0.006*** (0.002)	−0.007*** (0.002)	−0.005*** (0.002)	−0.005*** (0.002)	−0.006*** (0.001)
Running for reelection	0.001(0.023)	0.067** (0.027)	0.017(0.027)	−0.008(0.028)	−0.007(0.023)	0.014(0.012)
GOP PVI	0.007** (0.0005)	0.007*** (0.0006)	0.002*** (0.0005)	0.005*** (0.0006)	0.005*** (0.0007)	0.004*** (0.0002)
Money ratio	−0.00003(0.00004)	1.68e⁻⁶ (3.70e⁻⁶)	−3.12e⁻⁷*** (1.69e⁻⁸)	−9.81e⁻⁶*** (0.00002)	−1.54e⁻⁶ (3.43e⁻⁶)	−4.13e⁻⁷*** (1.17e⁻⁸)
Constant	−0.246***	−0.304***	−0.342***	−0.286***	−0.275***	−0.296***
Model R²	.944	.934	.923	.937	.942	0.931
n	438	446	444	442	443	2,213

Note. OLS = ordinary least square; PVI = Partisan Voting Index.

Coefficient significance = .1. **Coefficient significance = .05. ***Coefficient significance < .01.

These models only test for a relationship with ideal points and not if one is more moderate than their party. Yet the results show there is a relationship between being cross-pressured and conservative behavior. This is true for three sessions and the aggregate series. The aggregate effect is 0.058, which is equal to 11 percent of a standard deviation in DW-NOMINATE scores throughout the series. This finding makes sense, given that 173 of 246 cross-pressured members are Democrats.

There are other important findings to address related to Table 5. Being a Republican has the strongest relationship with ideal point conservatism. For every term that an incumbent completes, they are more likely to have liberal ideal points. Interestingly, running for reelection has no consistent relationship with legislative behavior. This is most likely because of the strong incumbency advantage. Most MCs seek to make a career out of serving in Congress (Polsby 1968). So, many members act as if they are always running for reelection (King 1997). Finally, the PVI plays a significant role in ideal point estimation. Increasing the Republican leaning of a district by 1 point leads an MC to be 0.004 more conservative. The difference between the most Democratic district and most Republican district is 75 units of PVI. The maximal effect of PVI is 0.300 as it relates to DW-NOMINATE scores.

Table 6 presents the results of a model where I look at the distance to the party median for Democratic MCs. This specifically tests the legislative behavior of Democrats compared with each other. I do not include party in these models because I have already controlled for this in the dependent variable. Once again, I use robust standard errors due to heteroskedasticity in some models of Table 6. I also tested to make sure that multicollinearity was not a concern. We expect to find a positive relationship with being cross-pressured and difference to the median. This would mean that cross-pressured MCs are more conservative than the median Democrat.

Table 6

OLS with Dependent Variable Being Distance to Democratic Party Median.

Variable	109th house	110th House × Robust Standard Errors	111th house	112th house	113th house	109th–113th house
Cross-pressured	0.100*** (0.033)	0.076*** (0.029)	0.201*** (0.015)	0.048* (0.027)	0.106*** (0.036)	0.144*** (0.011)
Terms completed	−0.005*** (0.002)	−0.005*** (0.002)	−0.007*** (0.002)	−0.004*** (0.001)	−0.004*** (0.001)	−0.006*** (0.0007)
Running	−0.020(0.005)	0.013(0.025)	−0.008(0.024)	0.029(0.020)	−0.025(0.022)	−0.012(0.012)
GOP PVI	0.005*** (0.0006)	0.006*** (0.0009)	0.001*** (0.0005)	0.045*** (0.0007)	0.004*** (0.0006)	0.003*** (0.0004)
Money ratio	1.18e⁻⁶ (0.00005)	4.79e⁻⁶ (3.79e⁻⁶)	−2.84e⁻⁷** (1.38e⁻⁷)	0.00006(0.00007)	−0.00002(0.00008)	−3.82e⁻⁷*** (1.27e⁻⁸)
Cross × GOP PVI	−0.001(0.003)	0.0004(0.003)	−0.0004(0.001)	0.004** (0.002)	0.002(0.005)	−0.0004(0.0008)
Constant	0.101	0.727	0.013	0.063	0.108	0.062
Model R²	.495	.516	.517	.420	.456	.398
n	202	241	261	199	204	1,108
Standard deviation of distance to DEM median	0.125	0.150	0.150	0.124	0.113	0.135
Effect of cross-pressured	0.80 SD	0.51 SD	1.34 SD	0.39 SD	0.94 SD	1.07 SD
Maximal effect of GOP PVI for cross-pressured	0.72 SD	0.73 SD	0.20 SD	0.72 SD	0.63 SD	0.55 SD

Note. OLS = ordinary least square; PVI = Partisan Voting Index.

Coefficient significance = .1. **Coefficient significance = .05. ***Coefficient significance < .01.

The data presented in Table 6 provide strong support for Hypothesis 1. In every single session being a cross-pressured Democrat makes one more likely to moderate compared with other Democrats. The substantive effect of being a cross-pressured Democrat is striking. In the aggregate series, we expect a cross-pressured Democrat to be 1.07 standard deviations more conservative than other Democrats. In three of the five sessions, the effect is at least 0.80 standard deviations. I use a t test to compare the average difference with the Democratic median for cross-pressured members compared with copartisans. Although I do not show this analysis, there is a significant difference in each session of the House.

One may argue that being cross-pressured does not impact voting behavior. Instead, it may be district-level ideology. As a proxy for this, I use the PVI. This allows me to test for the impact of each variable on voting behavior, controlling for the other. Similar to being cross-pressured, PVI has a significant effect in all five sessions. Table 6 highlights the maximal effect of PVI on a cross-pressured Democrat. I multiply the coefficient by the largest Republican advantage for a Democrat. In the 109th session this is GOP+18, and in the 111th House this is GOP+25. This creates the maximal effect for cross-pressured Democrats in the House. This effect is in addition to being cross-pressured. Hence, the representative who was in a GOP+18 district in the 109th House was likely to be 1.52 standard deviations more conservative than other Democrats. Thus, district-level ideology is a separate factor from being cross-pressured and has its own substantive effect on legislative behavior.

The interaction between being cross-pressured and PVI is significant during the 112th House. This means that there is a stronger impact to being cross-pressured added to district ideology. For a cross-pressured Democrat in a district that is GOP+10, we expect that individual to be 0.137 points more conservative than the median. This is 1.1 standard deviations away from the median.

The appendix shows the results of looking at differences to the Democratic mean with the same results. The only variable that leads Democrats toward liberal voting behavior is the length of service. As a Democratic representative serves in Congress more, he or she will move to the left. But Table 6 clearly shows that, controlling for several factors, being cross-pressured leads a Democrat to vote more conservatively than his or her party. At the same time, as a district becomes more Republican, we see the legislative behavior moving in the same direction. As a result, I confirm Hypothesis 1 as it relates to Democrats.

Table 7 looks at the results of the same tests for Republicans. As a reminder, we expect being a cross-pressured Republican to have a negative relationship with difference to the median. This is because liberal scores are below the median for the party. The findings for Republicans (Table 7) are somewhat different than for Democrats (Table 6). Only in two sessions, the 110th and 111th, is there a significant moderating effect for being a cross-pressured Republican. These were sessions leading into the 2008 and 2010 election. When being cross-pressured does lead to moderation the effects are strong. In the 110th House, being cross-pressured leads a Republican to be 0.9 standard deviations more liberal than the median Republican MC. The pooled analysis predicts that cross-pressured Republicans are more liberal than those who are not cross-pressured. The effect for the entire series is equal to 0.53 standard deviations from the median. Yet this is smaller than the effect for being a cross-pressured Democrat, which was 1.07 standard deviations.

Table 7

OLS with Dependent Variable Being Distance to Republican Party Median.

Variable	109th house	110th house	111th House × Robust Standard Errors	112th house	113th house	109th–113th house
Cross-pressured	−0.032(0.041)	−0.151** (0.075)	−0.246*** (0.080)	0.027(0.053)	−0.017(0.083)	−0.090*** (0.021)
Terms completed	−0.008*** (0.002)	−0.006* (0.003)	−0.005(0.003)	−0.007** (0.003)	−0.007*** (0.002)	−0.006*** (0.001)
Running for reelection	0.010(0.031)	0.066** (0.027)	0.035(0.039)	−0.041(0.041)	0.003(0.035)	0.024(0.016)
GOP PVI	0.054(0.002)	0.005*** (0.002)	0.001(0.001)	0.004*** (0.002)	0.005*** (0.002)	0.003*** (0.0006)
Money ratio	−0.00001(0.0001)	−0.00004(0.0002)	−0.0003(0.001)	8.21e⁻⁶ (0.00002)	−2.10e⁻⁶ (0.00002)	−2.75e⁻⁶ (0.00002)
Cross × GOP PVI	0.014(0.012)	−0.029(0.027)	−0.007* (0.004)	−0.002(0.006)	0.008(0.035)	−0.004(0.003)
Constant	0.011	−0.013	−0.004	0.010	−0.034	−0.004
Model R²	.181	.201	.080	.053	.076	.077
n	235	205	183	243	239	1,105
Standard deviation of distance to GOP median	0.154	0.169	0.181	0.174	0.176	0.171
Effect of being cross-pressured	NA	0.89 SD	1.36 SD	NA	NA	0.53 SD
Maximal effect of cross-pressured PVI	NA	0.16 SD	0.39 SD	0.30 SD	0.14 SD	0.22 SD

Note. OLS = ordinary least square; PVI = Partisan Voting Index.

Coefficient significance = .1. **Coefficient significance = .05. ***Coefficient significance < .01.

One finds a less consistent relationship between Republican legislative behavior and PVI as well. The models in Table 7 show that district ideology only predicts differences from Republican Party median in three of the five sessions. Thus, it does nothing to test partisan identification of the MC. Finally, the number of terms completed plays a significant role in legislative behavior in most sessions. The results show that a Republican becomes slightly more liberal as they complete each term. Thus, no relationships are as strong for Republicans (Table 7) as those found for Democrats (Table 6). But looking at the pooled analysis shows that cross-pressured Republicans are more moderate than their copartisans. As a result, this study confirms Hypothesis 1 and agrees that being cross-pressured predicts moderation.

This effect occurs even when controlling for district-level ideology separately through PVI. We also find a consistent relationship when looking at PVI. Both Democrats and Republicans are more conservative for every increase in Republican presidential support. So, being cross-pressured increases moderation. However, there is a significant and independent effect for each unit change in district ideology. This variable includes those who are cross-pressured and those who are not cross-pressured. The effect of PVI does mute the overall impact of being cross-pressured. But these variables are complementary in nature. In essence, a better way of thinking about this is through considering them as two effects leading to the same thing for cross-pressured members.

It is important to extrapolate on why there are clearer results for Democrats than for Republicans. Remember, this analysis shows that cross-pressured members, regardless of party, are more moderate than their copartisans. Cross-pressured Democrats are more conservative and cross-pressured Republicans are more liberal than their peers. It also shows that district-level ideology predicts differences with the party. So, district composition—in terms of ideology and being cross-pressured—impacts Democrats more than it does Republicans.

One may argue this is because Republicans have grown much more conservative in recent years. The mean Republican ideal point has increased from 0.6 in the 109th session to 0.723 in the 113th session. To provide context, Paul Ryan had an ideal point estimate of 0.75 in the 109th House. This made him the 23rd most conservative member in the House. Yet, in the 113th session, Ryan’s ideal point estimate had grown more conservative. His new estimate was 0.858, but it was only the 53rd most conservative.

Democrats, however, have stayed more consistent throughout the five sessions. In the 109th session, the Democratic mean was −0.365, and in the 113th session, it had become slightly more liberal at −0.378. Maxine Waters, a Representative from California’s 43rd district, had an ideal point estimate of −0.666 in the 109th session making her the fourth most liberal member in the House. In the 113th session, her ideal point estimate had moved to −0.554, making her the 13th most liberal member of the House.

Yet this does not explain why Democrats are more influenced by being cross-pressured than are Republicans. A more moderate Congressional Caucus would actually suggest smaller room for divergence. As seen in Tables 6 and 7, the standard deviation around the median ideal point is smaller for Democrats than Republicans. But, it is the members of the more moderate Democratic Party that are the most likely to moderate. They respond to the effects of being cross-pressured more clearly. Democrats also respond more consistently to district ideology as it relates to PVI. This result occurs even though it is Republicans who have become more extreme in recent years.

Another potential reason for why cross-pressured Republicans are less likely to moderate is the larger standard deviation in the party. Table 7 shows that as Republicans have become more conservative, there is also more disagreement within the party. The Republican caucus housed Walter Jones (NC-3) with an ideal point estimate of 0.042. But, it also had Jim Sensenbrenner (WI-5) with an estimate of 1.234.² In short, there is more variance among Republicans even as they become more conservative. Democrats have a smaller difference. Their caucus included Ron Barber (AZ-2), who lost reelection at −0.088, and Jim McDermott (WA-7) at −0.678.

Yet there are reasons to suggest that Democrats would be less likely to moderate than Republicans. The Blue Dog Democrat Coalition consists of close to 20 members that label themselves conservative. The Tuesday group consists of close to 50 members of the Republican party who seek to counterbalance the power of the Freedom Caucus.

There is also wider variance in the PVI represented by Democrats than by Republicans. For instance, Jim Matheson was a Democratic MC representing a constituency with a PVI of GOP+18. Charlie Rangel represented a constituency in the same caucus with a PVI of DEM+42. For Republicans, the largest PVI deficit was Joseph Cao at DEM+24. It should be noted that Cao won his seat (LA-2) largely because of the criminal trial facing the incumbent William Jefferson in 2008. Cao lost his reelection bid in 2010. The largest Republican advantage is GOP +32. Thus, the PVI range for Democrat MCs is larger at 60 than it is for Republicans at 56.

This difference in moderation may also be the result of more Democratic districts that are safe. Only four districts have a PVI greater than GOP+30, whereas there are 102 Democratic districts with the same advantage. It is beyond the scope of this article to test the effects of this asymmetry. However, this packing of Democrats into safe districts could lead those Democrats in less safe districts to moderate.

The simplest alternative for this is that fewer Republicans are cross-pressured (see Table 1). Only 73 of the 246 cross-pressured MCs in the five sessions were Republicans. Put another way, only 6.7 percent of Republicans in these sessions were cross-pressured, and 15.6 percent of Democrats were cross-pressured. As a result, fewer Republicans feel the need to moderate behavior compared with their copartisans. They are less likely to be cross-pressured and represent districts that heavily favor the Democratic presidential candidate.

In the appendix, I test for the independent impact that being in a marginal district may have. The effects of being in a marginal district sometimes have a significant relationship with moderating legislative behavior. Nevertheless, the effect is much smaller than it is for being a cross-pressured member. The results from Tables 6 and 7 highlight that the district’s composition impacts MC voting behavior. It is not through being in a “marginal” district. Instead, cross-pressured MCs are likely to moderate voting behavior compared with copartisans (Hypothesis 1).

The second set of appendix tables includes models with a dummy variable for being in the South (confederate states). The results from Tables 6 and 7 remain consistent. Regardless of party, being from the South makes one more conservative. But, being cross-pressured remains a significant predictor of legislative behavior. The final models I run create an interactive variable for being cross-pressured and the number of terms completed. In none of the terms is the interactive term significant. This suggests that long-serving cross-pressured members are no different in voting behavior than are newer cross-pressured members. None of these different models affect the results in Tables 6 and 7.

Next, the analysis turns to Hypothesis 2 which argues that even though cross-pressured members moderate, they are less likely to win reelection than other members of their party. I measure this through creating a variable for moderation by including any member who is more than one standard deviation toward the center from his or her party. Thus, if a Democrat is more than one standard deviation more conservative than the median Democrat, I label them as a moderate. If a Republican is one standard deviation more liberal than the median Republican, I measure them as a moderate. For instance, this could apply to a Republican with an ideal point lower than 0.44 in the 109th House and a Democrat with an ideal point greater than −0.243 in the 109th House. There are 185 Democrat moderates, with 114 of those being cross-pressured. There are 163 Republican moderates, and 33 of these individuals are cross-pressured. The first way I test Hypothesis 2 is to look at reelection rates for those who seek reelection (Table 8).

Table 8

Logit Model of Winning.

Variable	109th house	110th house	111th house	112th house	113th house	109th–113th house
Cross-pressured	−1.540*** (0.463)	−1.291* (0.738)	−3.321*** (0.513)	−2.94*** (0.593)	−1.965** (0.797)	−2.449*** (0.218)
Terms completed	−0.017(0.053)	0.078(0.079)	0.129** (0.056)	0.120(0.077)	0.032(0.082)	0.070*** (0.026)
Ideal point	1.942(1.492)	1.484(1.914)	1.637(1.655)	1.483(1.801)	−6.410* (3.360)	0.504(0.678)
GOP	−3.078** (1.401)	−3.454* (1.923)	1.115(1.457)	−2.115(1.910)	8.402** (3.873)	−0.570(0.660)
Money ratio	0.002(0.005)	0.432** (0.216)	0.011(0.010)	−0.0006* (0.0004)	0.176(0.157)	1.94e⁻⁶ (0.00002)
Quality challenger	−1.125** (0.462)	−1.405** (0.627)	−2.12*** (0.464)	−1.334** (0.532)	−1.325* (0.721)	−1.450*** (0.207)
Constant	4.748***	4.061***	3.059***	4.298***	0.965	3.581***
Model pseudo-R²	.154	.244	.537	.243	.425	.237
n	402	398	391	370	387	1,948
Odds ratio of cross-pressured	0.232	0.275	0.036	0.053	0.140	0.086

Coefficient significance = .1. **Coefficient significance = .05. ***Coefficient significance < .01.

The most important finding is also the clearest in Table 8. There is a negative relationship between being cross-pressured and winning reelection. In all five sessions, the effect is strong. Because the coefficients of logit models are difficult to interpret, I include the odds ratio at the bottom of Table 8. It suggests that a cross-pressured member in the 109th House, holding all else equal, was 23.2 percent as likely to return to Congress than one who is not cross-pressured. The highest odds ratio for a cross-pressured member is 27.5 percent. Throughout the entire series, cross-pressured members are 0.086 times as likely to be reelected compared with MCs who are not cross-pressured. One also notices the consistent negative relationship when facing a quality challenger and winning reelection. Furthermore, the number of terms completed and ideal point estimates are only significant in one session.

The impact of party is important to disentangle. Being a Republican harms reelection rates in 2006 (following 109th House) and 2008. However, it helps in 2014 (following 113th House). This makes sense, given the political climate in those elections. In sum, the above confirms the first part of Hypothesis 2: being cross-pressured makes it less likely that one will win reelection.

These models will help to determine if it is being a moderate or being cross-pressured that makes one less likely to be reelected. There is a relationship between being cross-pressured and being a moderate, as discussed above. I try to solve this problem by testing for endogeneity. But they also test somewhat different groups. In total, 114 of the 173 cross-pressured Democrats are also moderates. And 33 of the 73 cross-pressured Republicans are moderate. This represents overlap, but it does not represent the exact same set of individuals.

The first model of Table 9 looks at Democrat chances of winning reelection. The results show that being cross-pressured makes one less likely to win reelection. Holding all else constant, the probability of a cross-pressured Democrat winning reelection is 0.692 (0.553, 0.831). This compares with a probability of 0.969 for Democrats who are not cross-pressured. Being a moderate Democrat has no impact on one’s chances of reelection. But being a cross-pressured and moderate Democrat does have an impact. What this means is that one who falls into the category of being cross-pressured and one standard deviation more conservative than the average Democrat is more likely to win reelection than one who is not moderate. In other words, being a moderate Democrat does not help reelection. Yet being a moderate cross-pressured Democrat helps one mitigate these factors slightly. This reelection rate is still about 20 percent below one who is not cross-pressured (0.791 compared with 0.969).

Table 9

Logit Model of Winning, with Moderation Included for 109th to 113th U.S. House.

Variable	Democrats	Republicans	Both parties
Cross-pressured	−2.868*** (0.467)	−1.942*** (0.487)	−2.732*** (0.294)
Terms completed	0.051(0.034)	0.092** (0.045)	0.069*** (0.027)
Ideal point	−2.082(1.465)	2.554** (1.286)	0.281(0.683)
GOP			−0.340(0.656)
Money ratio	0.001(0.003)	−0.0004* (0.0002)	1.77e⁻⁶ (0.00002)
Quality challenger	−1.514*** (0.291)	−1.145*** (0.318)	−1.408*** (0.208)
DEM moderate	−0.587(0.595)
Democrat cross-pressured moderate	1.175* (0.660)
Republican moderate		0.127(0.562)
Republican cross-pressured moderate		0.518(0.761)
Moderate			−0.611* (0.347)
Cross-pressured moderate			0.944** (0.475)
GOP PVI	0.024* (0.013)
DEM PVI		0.954*** (0.016)
Constant	3.153***	1.096	3.586***
Model pseudo-R²	.308	.190	.241
n	996	952	1,948

Note. PVI = Partisan Voting Index.

Coefficient significance = .1. **Coefficient significance = .05. ***Coefficient significance < .01.

The second model of Table 9 looks at Republicans. It continues to find evidence indicating the problem of being cross-pressured. The probability of a Republican who is not cross-pressured being reelected is 0.957, and that probably decreases to 0.791 for one who is cross-pressured. Being cross-pressured does not help Republicans or Democrats win reelection. Furthermore, moderating behavior as a Republican does not help win reelection. The interaction between being a moderate and being cross-pressured has no statistical significance. This implies that there is no electoral incentive for Republicans to moderate, even those who are cross-pressured. These two models indicate that it is being a cross-pressured representative, and not a moderate, from both parties that makes one less likely to win reelection. Having a quality challenger continues to make it more difficult for representatives from both parties to win reelection.

The final model of Table 9 combines Democrats and Republicans throughout the five sessions. Cross-pressured MCs continue to have the electoral sanction. Being a moderate (1 standard deviation away from the party median) makes one less likely to win reelection. Yet the interaction variable for those who are cross-pressured and moderate MCs has a positive and corrective effect on winning; t tests show (see Table 10) that being a cross-pressured MC has the lowest probability of being reelected compared with others. Those who are moderate, regardless of whether they are cross-pressured are reelected at a rate of 81.8 percent, while others are reelected at a rate of 93.8 percent. But, those who are cross-pressured only win reelection at a rate of 65.6 percent. The final t test looks at whether there is a difference between representatives that are cross-pressured and moderate and those who are only cross-pressured and produces the only nonsignificant difference.

Table 10

t Tests for Reelection Rates.

Variable	Yes	No
Cross-pressured (p < .001)	0.656 (0.591, 0.720)	0.951 (0.941, 0.961)
Moderate (p < .001)	0.818 (0.775, 0.862)	0.938 (0.927, 0.950)
Cross-pressured moderate (p < .001)	0.682 (0.601, 0.762)	0.937 (0.925, 0.948)
Cross-pressured moderate versus cross-pressured (p = .153)	0.682 (0.601, 0.762)	0.613 (0.503, 0.722)

Thus, Table 9 highlights the fact that cross-pressured members are less likely to win reelection. This is true even when controlling for moderation in voting behavior. Furthermore, there is no difference in reelection rates between those who are cross-pressured and those who are both cross-pressured and moderate (Table 10). Being moderate harms one’s chances of winning reelection, but not as much as being cross-pressured. As a result, Hypothesis 2 is confirmed. There is no incentive for cross-pressured members to moderate as they win reelection at statistically similar rates than those who are cross-pressured and not moderates.

Conclusion

The findings of this study highlight the fact that cross-pressured members of the House moderate themselves vis-à-vis their fellow partisans. They are more likely to break with the party in voting behavior. When it comes to reelection, however, there is a significant and negative probability of winning when one is a cross-pressured member. What this means is that even though cross-pressured members moderate their voting behavior, they still have a lower return rate to Congress. This leads one to conclude the reason for this is largely the result of one’s partisan affiliation rather than voting behavior in Congress.

These findings point to the paradox of Congressional representation. Representatives behave as if members of the public pay attention to their voting behavior, but the public is willing to vote out of office those members who are cross-pressured, even if they moderate their voting behavior. We know that measuring these districts by marginality plays a significant, but smaller role in legislative voting behavior. We also know that being from the confederate states, as well as length of service both do not have an interactive effect with being cross-pressured in the likelihood of returning to office.

What is not clear is what these findings mean for representation. One could argue that members from cross-pressured districts should have a harder time getting reelected because they do not conform to their constituents’ preferences. A counter argument is that these members do moderate their behavior and still have difficulty getting reelected. Furthermore, if these members were perceived to be strong partisans they would most likely have difficulty being elected to Congress in the beginning. This paradox remains.

Ultimately, what this means is that districts that are willing to vote in a member of the other party often are in moderate districts. They prefer moderate representation (all else being equal). These districts, like many others, are susceptible to national waves in which members are elected that are not ideologically consistent with constituent ideology. Districts that are willing to vote for presidents of one party, and representatives of the other, would lead one to presume that either they are moderate districts, or districts with highly balanced partisan composition. Only in districts with moderate ideology would cross-pressured members be likely to moderate their behavior, as I find to be the case.

However, what all of this suggests is that it is important to understand what types of districts elect cross-pressured members. This article presents a first test of this when it incorporates district ideology. I use the PVI since this allows me to determine how much a party supports the other party’s presidential candidate. It also allows me to make more sense of the continuous nature of PVI in the analysis. I find that district-level ideology, measured in this way, does help us understand legislative behavior, and reelection rates. Yet the effects should not be seen as entirely independent of being cross-pressured. Members not only are impacted by representing cross-pressured districts as a binary variable but also have the continuous measure of district PVI impacting behavior and reelection rates.

In sum, cross-pressured districts present a way to better understand the increase of polarization in American politics. Citizens who are willing to vote for both parties, but do not reward members for moderating their behavior, are an important segment of the population to study. Although this study is indeed a preliminary one, it does show that cross-pressured members are a different phenomenon than marginal districts. Furthermore, we know that the paradox between representatives who seek to listen to constituent ideology, but constituents appear not to notice. Representation requires knowing about the behavior of the representative as well as the represented. The next step in forthcoming research necessarily should be to look at the behavior and opinions of citizens in cross-pressured district as well as determining what kind of candidates are successful in these districts. Future research on the behavior of legislators and constituents in cross-pressured districts could help us understand more about the depth and nature of polarization in the current era.

Footnotes

Appendix

Table A6

Difference to Republican Median with South Included.

Variable	109th house	110th House × Robust Standard Errors	111th house	112th house	109th–112th House × Robust Standard Errors
Cross-pressured	−0.149*** (0.047)	−0.129*** (0.048)	−0.213 (0.162)	−0.042 (0.072)	−0.125*** (0.032)
Terms completed	−0.006** (0.003)	−0.003 (0.003)	−0.004 (0.003)	−0.003 (0.003)	−0.004*** (0.001)
Running for reelection	0.017 (0.030)	0.069*** (0.022)	0.017 (0.034)	0.002 (0.044)	0.029* (0.017)
Super cross-pressured	−0.102* (0.060)	−0.002 (0.057)	−0.218 (0.173)	−0.122 (0.183)	−0.111** (0.044)
Cross-pressured × Terms	0.002 (0.066)	−0.004 (0.006)	0.007 (0.027)	0.005 (0.011)	0.001 (0.004)
Constant	0.055	−0.019	0.033	0.015	0.016
Model R²	.168	.157	.093	.009	.085
n	235	205	183	240	863
Standard deviation of distance to GOP median	0.149	0.147	0.159	0.171	0.157

Coefficient significance = .1. **Coefficient significance = .05. ***Coefficient significance < .01.

1

Mathematically, I determine the national PVI by adding the entire votes cast for the previous two elections for the Democratic candidate. Then I divide this number by the total votes cast. The Cook PVI method looks at only two-party votes cast, and I follow suit. I follow the same method for creating the PVI for each district.

2

Jones’s ideal point estimate is interesting. In the 112th House, it was 0.082, and in the 113th, it was 0.042. The second dimension estimate for Jones in the 113th session was 0.831. Thus, he was more moderate on economic issues than on social issues. Yet Jones represents a district that is GOP+10 in the 113th House. Similarly, the estimate for Sensenbrenner is outside of the traditional range of −1 to 1. I used the updated version from September 2015 found on .

About the Author

Benjamin T. Toll is an assistant professor of political science at Lake Superior State University in Michigan. He received his PhD in political science from Indiana University. His research interests lie at the intersection of public opinion and public policy in the United States. His current project looks at the politics of public higher education funding in the United States as it relates to public opinion on the topic. He can be reached at btoll@lssu.edu.

References

Abramowitz

Alan I.

2010. The Disappearing Center: Engaged Citizens, Polarization and American Democracy. New Haven: Yale University Press.

Adams

Greg D.

1997. “Abortion: Evidence of an Issue Evolution.” American Journal of Political Science 41 (3): 718–37.

Ansolabehere

Stephen

Jones

Philip Edward

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

Ansolabehere

Stephen

Rodden

Jonathan

Snyder

James M.

Jr.

2006. “Purple America.” Journal of Economic Perspectives 20 (2): 97–118. https://www.aeaweb.org/articles?id=10.1257/jep.20.2.97.

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.

Ashworth

Scott

de Mesquita

Ethan Bueno

. 2006. “Delivering the Goods: Legislative Particularism in Different Electoral and Institutional Settings.” Journal of Politics 68 (1): 168–79.

Bafumi

Joseph

Herron

Michael C.

2010. “Leapfrog Representation and Extremism.” American Political Science Review 104 (3): 519–42.

Bartels

Larry M.

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

Bartels

Larry M.

2008. Unequal Democracy: The Political Economy of the New Gilded Age. Princeton: Princeton University Press.

10.

Bauer

Nicole M.

Yong

Laurel Harbridge

Krupnikov

Yanna

. 2017. “Who Is Punished? Conditions Affecting Voter Evaluations of Legislators Who Do Not Compromise.” Political Behavior 39 (2): 279–300.

11.

Brady

David W.

Han

Hahrie

Pope

Jeremy C.

2007. “Primary Elections and Candidate Ideology: Out of Step with the Primary Electorate?” Legislative Studies Quarterly 32 (1): 79–105.

12.

Brady

David W.

Schwartz

Edward P.

1995. “Ideology and Interests in Congressional Voting: The Politics of Abortion in the U.S. Senate.” Public Choice 84 (1/2): 25–48.

13.

Bullock

John G.

2011. “Elite Influence on Public Opinion in an Informed Electorate.” American Political Science Review 105 (3): 496–515.

14.

Carson

Jamie L.

Engstrom

Erik J.

Roberts

Jason M.

2007. “Candidate Quality, the Personal Vote, and the Incumbency Advantage in Congress.” American Political Science Review 101 (2): 289–301.

15.

Carson

Jamie L.

Koger

Gregory

Lebo

Matthew J.

Young

Everett

. 2010. “The Electoral Costs of Party Loyalty in Congress.” American Journal of Political Science 54 (3): 598–616.

16.

Converse

Philip E.

1964. “The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, edited by Apter

David E.

, 206–61. New York: The Free Press.

17.

Converse

Philip E.

1990. “Popular Representation and the Distribution of Information.” In Information and Democratic Processes, edited by Ferejohn

John A.

Kuklinski

James H.

, 369–90. Urbana-Champaign: University of Illinois Press.

18.

Dalton

Russell

. 2013. The Apartisan American: Dealignment and Changing Electoral Politics. Thousand Oaks: CQ Press.

19.

Druckman

James N.

Kifer

Martin J.

Parkin

Michael

. 2009. “Campaign Communication in U.S. Congressional Elections.” American Political Science Review 103 (3): 343–66.

20.

Erikson

Robert S.

Wright

Gerald C.

2005. “Voters, Candidates, and Issues in Congressional Elections.” In Congress Reconsidered, edited by Dodd

Lawrence C.

Oppenheimer

Bruce I.

, 132–61. Washington, DC: CQ Press.

21.

Fenno

Richard

. 1978. Homestyle: U.S. House Members in Their Districts. New York: Little Brown.

22.

Gelman

Andrew

Park

David

Shor

Boris

Cortina

Jeronimo

. 2010. Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do. Princeton: Princeton University Press.

23.

Gerber

Alan S.

Huber

Gregory A.

Doherty

David

Dowling

Conor M.

2011. “Citizens’ Policy Confidence and Electoral Punishment: A Neglected Dimension of Electoral Accountability.” Journal of Politics 73 (4): 1206–24.

24.

Gilens

Martin

. 2001. “Political Ignorance and Collective Policy Preferences.” American Political Science Review 95 (2): 379–96.

25.

Gilens

Martin

. 2005. “Inequality and Democratic Responsiveness.” Public Opinion Quarterly 69 (5): 778–96. https://www-jstor-org-443.web.bisu.edu.cn/stable/3521574.

26.

Gilens

Martin

. 2012. Affluence and Influence: Economic Inequality and Political Power in America. Princeton: Princeton University Press.

27.

Gohmann

Stephan F.

Ohsfeldt

Robert L.

1994. “Voting in the U.S. House on Abortion Funding Issues: The Role of Constituents’ and Legislators’ Ideology, before and after the Webster Decision.” American Journal of Economic and Sociology 53 (4): 455–74.

28.

Grofman

Bernard

Koetzle

William

McDonald

Michael P.

Brunell

Thomas L.

2000. “A New Look at Split-ticket Outcomes for House and President: The Comparative Midpoints Model.” Journal of Politics 62 (1): 34–50.

29.

Harbridge

Laurel

Malhotra

Neil

. 2011. “Electoral Incentives and Partisan Conflict in Congress: Evidence from Survey Experiments.” American Journal of Political Science 55 (3): 494–510.

30.

Harbridge

Laurel

Malhotra

Neil

Harrison

Brian F.

2014. “Public Preferences for Bipartisanship in the Policymaking Process.” Legislative Studies Quarterly 39 (3): 327–55.

31.

Hayes

Matthew

Hibbing

Matthew V.

Sulkin

Tracy

. 2010. “Redistricting, Responsiveness, and Issue Attention.” Legislative Studies Quarterly 35 (1): 91–115.

32.

Hibbing

John R.

Theiss-Morse

Elizabeth

. 1995. Congress as Public Enemy: Public Attitudes toward American Political Institutions. Cambridge: Cambridge University Press.

33.

Hill

Kim Quayle

Hurley

Patricia

. 1999. “Dyadic Representation Reappraised.” American Journal of Political Science 43 (1): 109–37.

34.

Hussey

Wesley

Zaller

John

. 2011. “Who Do Parties Represent?” In Who Get’s Represented, edited by Enns

Peter K.

Wlezien

Christopher

, 311–44. New York: Sage.

35.

Jacobson

Gary C.

2008. The Politics of Congressional Elections. 7th ed. New York: Longman Classics in Political Science.

36.

Karol

David

. 2009. Party Position Change in American Politics: Coalition Management. Cambridge: Cambridge University Press.

37.

Kinder

Donald R.

Kalmoe

Nathan P.

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

38.

King

Anthony

. 1997. “The Vulnerable American Politician.” British Journal of Political Science 27 (1): 1–22.

39.

Klar

Samara

Krupnikov

Yanna

. 2016. Independent Politics: How American Disdain for Parties Leads to Political Inaction. New York: Cambridge University Press.

40.

Layman

Geoffrey C.

Carsey

Thomas M.

Green

John C.

Herrera

Richard

Cooperman

Rosalyn

. 2010. “Activists and Conflict Extension in American Party Politics.” American Political Science Review 104 (2): 324–46.

41.

Levendusky

Matthew C.

2009. The Partisan Sort: How Liberals Became Democrat and Conservatives Became Republicans. Chicago: The University of Chicago Press.

42.

Levendusky

Matthew C.

2013. How Partisan Media Polarize America. Chicago: The University of Chicago Press.

43.

Mayhew

David

. 1974. “Congressional Elections: The Case of the Vanishing Marginals.” Polity 6 (3): 295–317.

44.

Mayhew

David

. 2004. Congress: The Electoral Connection. 2nd ed. New Haven: Yale University Press.

45.

McCarty

Nolan

Poole

Keith T.

Rosenthal

Howard

. 2008. Polarized America: The Dance of Ideology and Unequal Riches. Boston: MIT Press.

46.

Miller

Gary

Schofield

Norman

. 2003. “Activists and Partisan Realignment in the United States.” American Political Science Review 97 (2): 245–60.

47.

Miller

Gary

Schofield

Norman

. 2008. “The Transformation of the Republican and Democratic Party Coalitions in the U.S.” Perspectives on Politics 6 (3): 433–50.

48.

Miller

Warren E.

Stokes

Donald E.

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

49.

Monroe

Alan D.

1998. “Public Opinion and Public Policy, 1980-1993.” Public Opinion Quarterly 62 (1): 6–28.

50.

Nichols

John

. 2013. “America Gets Primaried: The Government Shutdown Is All about Electoral Politics—Particularly the Gerrymandering of Republican Districts and Primaries that Make Compromise Impossible.” The Nation, October 1. https://www.thenation.com/article/america-gets-primaried/.

51.

Page

Benjamin I.

Shapiro

Robert Y.

1983. “Effects of Public Opinion on Policy.” American Political Science Review 77 (1): 175–90.

52.

Peterson

David A. M.

Grossback

Lawrence J.

Stimson

James A.

Gangl

Amy

. 2003. “Congressional Response to Mandate Elections.” American Journal of Political Science 47 (2): 411–26.

53.

Pew Research Center. 2014. “Political Polarization in the American Public: Section 4-Political Compromise and Divisive Policy Debates.” Pew Research Center, June 10. http://www.people-press.org/2014/06/12/section-4-political-compromise-and-divisive-policy-debates/

54.

Polsby

Nelson

. 1968. “The Institutionalization of the U.S. House of Representatives.” American Political Science Review 62 (1): 144–68.

55.

Powell

Lynda W.

1982. “Issue Representation in Congress.” Journal of Politics 44 (3): 658–78.

56.

Wright

Gerald C.

1989. “Policy Voting in the U.S. Senate: Who Is Represented.” Legislative Studies Quarterly 14 (4): 465–86.

57.

Wright

Gerald C.

Berkman

Michael B.

1986. “Candidates and Policy in United States Senate Elections.” American Political Science Review 80 (2): 567–88.