Government Polarization Essay

lable at ScienceDirect

Electoral Studies 41 (2016) 12e22

Contents lists avai

Electoral Studies

journal homepage: www.elsevier.com/locate/electstud

The rise of negative partisanship and the nationalization of U.S. elections in the 21st century

Alan I. Abramowitz*, Steven Webster Emory University, United States

a r t i c l e i n f o

Article history: Received 13 July 2015 Received in revised form 22 September 2015 Accepted 11 November 2015 Available online 23 November 2015

Keywords: U.S. elections Partisanship

* Corresponding author. E-mail address: [email protected] (A.I. Abramow

http://dx.doi.org/10.1016/j.electstud.2015.11.001 0261-3794/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

One of the most important developments affecting electoral competition in the United States has been the increasingly partisan behavior of the American electorate. Yet more voters than ever claim to be independents. We argue that the explanation for these seemingly contradictory trends is the rise of negative partisanship. Using data from the American National Election Studies, we show that as partisan identities have become more closely aligned with social, cultural and ideological divisions in American society, party supporters including leaning independents have developed increasingly negative feelings about the opposing party and its candidates. This has led to dramatic increases in party loyalty and straight-ticket voting, a steep decline in the advantage of incumbency and growing consistency between the results of presidential elections and the results of House, Senate and even state legislative elections. The rise of negative partisanship has had profound consequences for electoral competition, democratic representation and governance.

© 2015 Elsevier Ltd. All rights reserved.

In the 21st century, the United States has entered a new age of partisanship. Sharp party divisions now characterize all of the na- tion's major political institutions. In Congress, the ideological divide between Democrats and Republicans in both the House and Senate is now larger than at any time in the past century (Ansolabehere et al., 2001; Theriault, 2008; Bafumi and Herron, 2010; Mann and Ornstein, 2013; Kraushaar, 2014). Party unity on roll call votes has increased dramatically in both chambers in recent years (Izadi, 2014; McCarty et al., 2008; Theriault, 2008; Sinclair, 2006). Moreover, the party divide in Washington is not limited to the elected branches of government. On the Supreme Court, the justices now divide along party lines on major cases with greater frequency than at any time in recent history (Clark, 2009; Bartels, 2015; Stone, 2014). And deep party divisions are not confined to the federal government. In many of the states, Democrats and Re- publicans are even more divided along ideological lines than Democrats and Republicans in Congress (Shor and McCarty, 2011).

The resurgence of partisanship in American politics has not been limited to political elites. Indeed, the growing intensity of partisan conflict among political elites in recent years cannot be understood without taking into account the increasingly partisan behavior of the American electorate. In this article we argue that one of the most

itz).

important trends in American politics over the past several decades has been the rise of negative partisanship in the electorate. The rise of negative partisanship, a development not captured by the tradi- tional party identification scale, has led to a sharp increase in party loyalty in voting for elected offices at all levels, a concurrent increase in straight-ticket voting and a growing connection between the results of presidential elections and the results of House, Senate and even state legislative elections. To a greater extent than at any time in the post-World War II era, the outcomes of elections below the presidential level reflect the outcomes of presidential elections.

1. The growth of partisan behavior: party loyalty and straight-ticket voting

Recent elections in the United States have been characterized by the highest levels of party loyalty and straight-ticket voting since the American National Election Studies first began measuring party identification in 1952. In 2012, according to the ANES survey, 91 percent of party identifiers and leaners voted for their party's presidential candidate.1 That tied the record first set in 2004 and

1 All analyses of the 2012 ANES survey are based on the personal interviews only in order to facilitate comparisons with surveys done in earlier years. Results including the Internet-based component of the 2012 survey show slightly higher levels of party loyalty in voting for president, House and Senate.

Fig. 1. Trends in consistent party loyalty among democratic and Republican voters, 1972e2012. Source: ANES Cumulative File.

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e22 13

matched in 2008. The 90 percent rate of party loyalty in the 2012 House elections tied the record set in 1956 and the 89 percent rate of party loyalty in the 2012 Senate elections broke the previous record of 88 percent set in 1958. As one would expect, these high rates of party loyalty were accompanied by very high rates of straight-ticket voting. The 89 percent rate of straight-ticket voting in the presidential and House elections in 2012 broke the record of 87 percent set in 1952 and the 90 percent rate of straight-ticket voting in the presidential and Senate elections in 2012 broke the record of 89 percent set in 1960.

The extraordinarily high rates of party loyalty in the 2012 presidential, House and Senate elections represent a continuation of a long-term trenddone that has been evident since partisanship reached a low-point in the elections of the 1970s and 1980s (Bartels, 2000; Hetherington, 2001; Green et al., 2002). Moreover, the rise of partisan behavior has involved supporters of both major parties. This can be seen in Fig. 1 which displays the trends in consistent party loyalty among Democratic and Republican iden- tifiers, including leaning independents, between 1972 and 2012. Consistent loyalty here means voting for the candidates of your own party for President, House of Representatives and U.S. Senate in the same election.

The results in Fig. 1 show that party loyalty among Democrats and Republicans has increased dramatically since the 1980s. Among all party supporters including leaning independents the 81 percent rate of consistent loyalty in 2012 was an all-time record, breaking the previous record of 79 percent set in 1960. This loyalty rate represented a very sharp increase from the 55 to 63 percent loyalty rates among all party supporters between 1972 and 1988. For Republicans, the 79 percent rate of consistent loyalty in 2012 was somewhat lower than the loyalty rates of the 1952, 1956 and 1960 elections but substantially higher than the loyalty rates of the

1970s and 1980s. For Democrats, the 84 percent rate of consistent loyalty in 2012 was the highest ever recorded in an ANES survey, easily surpassing the 80 percent loyalty rate recorded in 2004.

The sharp increase in party loyalty in voting documented in Fig. 1 was not simply a result of party realignment in the South. Party loyalty has increased substantially in every region of the United States since the 1970s. In fact, the largest increase in party loyalty among voters was not in the South but in the Northeast. Between the 1972e80 elections and the 2004e2012 elections, the average rate of consistent loyalty increased from 55 percent to 78 percent in the South, from 54 percent to 79 percent in the North- east, from 60 percent to 75 percent in the Midwest and from 63 percent to 83 percent in the West. Resurgent partisanship in voting behavior is clearly a national phenomenon.

It is puzzling that record levels of party loyalty and straight- ticket voting in elections have occurred at the same time that the proportion of Americans identifying with either major party has reached its lowest level in recent history. In the 2012 ANES survey, only 63 percent of voters identified with either the Democratic or Republican Party in response to the initial party identification question. That was the lowest percentage of party identifiers in the history of the American National Election Studies. In contrast, be- tween 1952 and 1964 about 80 percent of voters readily identified with one of the two major parties. Even during the 1970s and 1980s when party loyalty in voting was at its nadir, the percentage of party identifiers never fell below 66 percent. And the ANES surveys are not the only ones that have picked up this trend. The Gallup Poll, using a slightly different question, has also reported a substantial increase in the proportion of Americans identifying themselves as independents in recent years (Jones, 2015).

It appears that a large proportion of American voters today are reluctant to openly acknowledge any affiliation with a political

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e2214

party. This may reflect a kind of social desirability effect: because partisanship has a negative connotation, the independent label is attractive to many Americans, especially the college educated (Petrocik 1974; Keith et al., 1992). These voters like to believe that they base their choices in elections on candidates and issues, not party labels. Being an independent can itself be an important social identity for some voters (Greene, 1999; Klar, 2014).

Today independents outnumber both Democrats and Re- publicans in the American electorate. When pressed about their party preference, however, between two-thirds and three-fourths of independent identifiers acknowledge that they lean toward one of the two major parties. In recent elections, fewer than ten percent of voters have fallen into the “pure independent” category. Moreover, when we shift our focus from partisan identification to partisan behavior we find that leaning independents as well as strong and weak party identifiers are voting along party lines to a greater extent now than at any time in the past forty years. This can be seen in Fig. 2 which shows the trends in consistent party loyalty among all three types of party supportersdstrong identifiers, weak identifiers and leaning independentsdbetween 1972 and 2012. Among each group of party supporters there was a marked increase in party loyalty between the elections of the 1970s and 1980s and the present.

Not surprisingly, the data displayed in Fig. 2 show that strong party identifiers had the highest rate of consistent party loyalty in every election. However, even among strong party identifiers, the rate of consistent loyalty rose from an average of around 75 percent during the 1970s and 1980s to close to 90 percent in the 2004e2012 elections. What is perhaps more impressive in these data, however, is the surge in party loyalty among weak party identifiers and leaning independents since the 1980s. During the 1970s and 1980s the rate of consistent party loyalty among these two groups was slightly below 50 percent. By 2012, however, the loyalty rate of both groups was close to 75 percent. Leaning in- dependents are now as loyal as strong party identifiers were during the 1970s and 1980s.

Fig. 2. Trends in consistent party loyalty among strong identifiers, weak identifiers and lea Source: ANES Cumulative File.

2. A theory of negative partisanship

The trends shown in Fig. 2 demonstrate that partisan behavior has surged in the American electorate in recent elections and that this surge has cut across the traditional division of party supporters into strong identifiers, weak identifiers and leaning independents. Among all types of party supporters, consistent loyalty reached its highest levels in half a century in 2012. We believe that this surge in partisan behavior reflects a fundamental change in the nature of partisan affect in the American electoratedthe rise of negative partisanship.

Affect towardpoliticalpartiesas groups has longbeen recognized as a crucial component of partisanship (Campbell et al.,1960; Green et al., 2002). Changes in affective evaluations of parties therefore have the potential to significantly alter partisan behavior and over the past two decades there has been an important change in affec- tive evaluations of the two major political parties in the U.S. elec- torate. While the feelings of Democrats and Republicans about their ownparty havechanged verylittle, their feelings about the opposing party have become much more negative (Bafumi and Shapiro, 2009; Greenberg, 2004; Jacobson, 2007; Iyengar et al., 2012; Iyengar et al., 2012; Mason, 2013, 2015; Abramowitz, 2015; Huddy et al., 2015).

Increasingly negative feelings toward the opposing party are partially a reflection of changes in the composition of the Demo- cratic and Republican electoral coalitions. Over the past several decades, partisan identities in the United States have become increasingly aligned with other salient social and political divisions in American society, most notably race and religion (Abramowitz, 2013; Frey, 2015). As a result, supporters of each party have come to perceive supporters of the opposing party as very different from themselves in terms of their social characteristics, political beliefs and values and to view opposing partisans with growing suspicion and hostility (Mason, 2015). Moreover, these negative feelings to- ward the opposing party are increasingly reinforced by exposure to partisan media, which have proliferated in recent years (Mutz, 2006, 2007; Prior, 2007; Iyengar and Hahn, 2009; Levendusky, 2013).

ning independents, 1972e2012.

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e22 15

The rise of negative partisanship is readily apparent in Fig. 3, which displays the trends in the mean rating by party supporters, including leaning independents, of their own party and the opposing party on the ANES feeling thermometer scale. This is a scale that ranges from zero degrees, the most negative rating, to 100�, the most positive rating. A rating of 50� on the scale is labeled neutral. The data in Fig. 3 show that party supporters' ratings of their own party have changed very little over this time period. The average rating by voters of their own party went from 72� in 1980 to 70� in 2012. In contrast, ratings by voters of the opposing party have fallen sharply over this time period. The average rating of the opposing party fell from 45� in 1980 to 30� in 2012. Thus, the substantial increase in polarization of party feeling thermometer scores over the past three decades was entirely due to increasingly negative ratings of the opposing party. Moreover, this increasing negativity toward the opposing party has affected all types of party supporters. Between 1980 and 2012 the mean rating of the opposing party on the feeling thermometer scale fell from 41� to 24� among strong party identifiers, from 47� to 36� among weak party identifiers and from 48� to 35� among leaning independents.

Today, far larger proportions of Democratic and Republican voters hold strongly negative views of the opposing party than in the past. In 1980, 55 percent of voters gave the opposing party a neutral or positive rating while only 27 percent gave the opposing party a rating of 30� or lower. In contrast, in 2012, only 26 percent of voters gave the opposing party a neutral or positive rating while 56 percent gave the opposing party a rating of 30� or lower.

The trends seen in Fig. 3 suggest that partisans' feelings about their own party and the opposing party are largely independent of one another. As partisans' ratings of the opposing party declined sharply in recent years, there was no corresponding increase in their ratings of their own party. Moreover, an examination of this relationship at the individual level supports the conclusion that partisans' feelings about their own party and the opposing party are largely independent. For all elections between 1978 and 2012, the average correlation between partisans' ratings of their own party

Fig. 3. Feeling thermometer ratings of own party and opposing party, 1980e2012. Source: ANES Cumulative File.

and the opposing party is a modest �.09. For 2012 alone, the cor- relation is a slightly stronger �.15. After controlling for strength of party identification, however, since strong partisans would be ex- pected to rate their own party more positively and the opposing party more negatively than weak or independent partisans, the correlation drops to a miniscule �.01 for all elections between 1978 and 2012 and to a very modest �.05 for 2012 alone.

3. Negative partisanship and voting behavior

The political significance of negative partisanship is that it has the potential to strongly influence voter decision-making. Along these lines, a number of studies have found evidence for the importance of negative partisanship in countries with multi-party systems. In a study of Canadian elections, Caruana et al. (2014) found that individuals who held negative views toward one of the three major Canadian parties were less likely to cast a vote for that party. They also find that individuals who disliked a particular party were more likely to vote and to engage in activities beyond voting such as attending a protest rally. In analyzing elections across four Western democracies, Medeiros and No€el (2014) also found evidence of negative partisanship. They concluded that the “commitment never to vote for a disliked party” plays an important and largely overlooked role in voter behavior. Along the same lines, Rose and Mishler (1998) found strong evidence that negative partisanship influenced voter decision-making and election out- comes in various post-communist Eastern European democracies.

While the concept of negative partisanship has thus far been used mainly to analyze voter decision-making in countries with multi-party systems, we believe that it can also help to explain recent electoral phenomena in the United States. In contrast to the situation in a multi-party system, however, in a two-party system, strongly negative feelings toward one party leave voters with only one choicedsupporting the other party. Thus, according to the theory of negative partisanship, increasingly negative feelings to- ward the opposing party should result in higher levels of loyalty in

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e2216

all types of elections among all types of partisans including leaning independents.

An examination of the bivariate relationships in the ANES data provides preliminary support for the negative partisanship hy- pothesis. The results indicate that for all elections between 1980 and 2012, negative ratings of the opposing party were more strongly related to loyalty than either positive ratings of one's own party or strength of party identification: the average correlation between feeling thermometer ratings of the opposing party and our measure of consistent loyalty is �.32 while the average correlation between feeling thermometer ratings of one's own party and consistent loyalty is only.22; the average correlation between strength of party identification and consistent loyalty is only .20. These differences are highly statistically significant (p < .001). For the three elections since 2004, the results are very similar: the average correlation between feeling thermometer ratings of the opposing party and consistent loyalty is �.35, the average correlation between feeling thermometer ratings of one's own party and consistent loyalty is only .21 and the average correlation between strength of party identification and consistent loyalty is only .20. Once again, these differences are highly statistically significant (p < .001).

Fig. 4 displays the trend in the relationship between feeling thermometer ratings of the opposing party and consistent loyalty between 1980 and 2012. Here we have divided voters into three groupsdthose with negative feelings toward the opposing party (ratings below 50�), those with neutral feelings toward the opposing party (ratings of exactly 50�) and those with positive feelings toward the opposing party (ratings above 50�). It is important to note that over the nine elections included in this figure, the relative size of these three groups changed dramatically. Feelings toward the opposing party were far more negative in 2012 than in 1980. In 1980, 34 percent of voters rated the opposing party above 50� while 45 percent of voters rated the opposing party below 50�. In contrast, in 2012 only 11 percent of voters rated the opposing party above 50� while 74 percent rated the opposing party below 50�.

Fig. 4. Trends in consistent loyalty between 1980 and 2012 by feeling thermometer rating Source: ANES Cumulative File.

The data displayed in Fig. 4 provide additional support for the negative partisanship hypothesisdin all nine elections, voters rat- ing the opposing party below 50� on the feeling thermometer were much more loyal than voters rating the opposing party at exactly 50� or above 50�. Moreover, the relationship between negative partisanship and loyalty appears to have gotten stronger over time. In the first two elections in this series,1980 and 1984, the difference in loyalty between voters rating the opposing party below 50� and voters rating the opposing party above 50� averaged just over 20 percentage points. In contrast, in the last two elections in this se- ries, 2008 and 2012, the difference in loyalty between these two groups averaged almost 40 percentage points. The proportion of voters with negative feelings toward the opposing party was much larger in 2008e2012 than in 1980e1984 and the impact of negative feelings on party loyalty appears to have been much stronger in 2008e2012 than in 1980e1984.

As a further test of the negative partisanship hypothesis, we conducted a logistic regression analysis of party loyalty among all voters supporting a party, including leaning independents, in the 2008e2012 elections. These are the two elections with the highest levels of party loyalty in recent history. For this analysis, the dependent variable is our dichotomous measure of consistent party loyalty. Our independent variables are strength of party identifi- cation, measured by dummy variables for strong party identifiers and weak party identifiers with leaning independents as the contrast category; feeling thermometer ratings by voters of their own party and the opposing party; and two variables measuring the incumbency status of the House and Senate contests with þ1 representing an incumbent from the same party as the voter, 0 representing an open seat contest and �1 representing an incumbent from the opposing party. The results of this logistic regression analysis are displayed in Table 1.

The results in Table 1 provide strong support for the negative partisanship hypothesis. Not surprisingly, voters were more loyal to their party when there was a House incumbent or a Senate incumbent from their own party and less loyal when there was a

of opposing party.

Table 1 Results of logistic regression analysis of party loyalty in 2008e2012 elections.

Independent variable Dependent variable: consistent loyalty

Strong partisan .297 (.201)

Weak partisan �.120 (.183)

Own party FT .026*** (.005)

Opposing party FT �.044*** (.004)

Party of house incumbent .487*** (.084)

Party of senate incumbent .248** (.082)

2012 Election .064 (.151)

Constant .985** (.356)

Observations 1406 Log likelihood �568.7 Nagelkerke R-Square .275

**p < 0.01***p < 0.001. Source: ANES Cumulative File.

Table 2 Results of logistic regression analysis of party loyalty in 2012 presidential election.

Independent Variable Dependent variable: party loyalty

Strong partisan 1.006** (.381)

Own party FT �.005 (.009)

Opposing party FT �.029*** (.008)

Pre-election candidate 6.390*** Preference (.376) Constant �.902

(.609)

Observations 3287 Log Likelihood 215.9 Nagelkerke R-Square .763

**p < 0.01***p < 0.001. Source: ANES Cumulative File

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e22 17

House incumbent or a Senate incumbent from the opposing party. Somewhat surprisingly, strength of party identification had little or no influence on loyalty after controlling for the other predictors in the model. Most importantly for our theory, feeling thermometer ratings of the opposing party were by far the strongest predictor of party loyalty: the more voters disliked the opposing party, the more loyal they were to their own party's candidates. Voters' feelings toward their own party also had a significant effect on loyalty but this effect was considerably weaker than that of feelings toward the opposing party.

4. An empirical test of the causal ordering hypothesis

The negative partisanship hypothesis is based on the assump- tion that negative affect toward the opposing party causes partisans to view that party as an unacceptable alternative and, therefore, to vote for their own party's candidates. The results in Table 1, which include controls for other likely influences on party loyalty, are consistent with this theory. However, we cannot rule out the pos- sibility that voters decide whether to vote for their party's candi- dates for some other reason and that this decision in turn influences their opinion of the opposing party.

Fortunately, we were able to take advantage of the panel format of the 2012 ANES survey to conduct an empirical test of our causal ordering hypothesis. In addition to the feeling thermometer ques- tions, the pre-election wave of the 2012 survey included a question about voting intentions in the presidential election.2 Presidential vote choice was, of course, measured in the post-election wave. In order to test the causal ordering hypothesis, we conducted a logistic regres- sion analysis of party loyalty in the presidential election while con- trolling for pre-election presidential voting intention. Out dependent variable was vote choice (loyal or disloyal). Our independent vari- ables were pre-election voting intention (loyal or disloyal), a dummy variable for strong partisans, and pre-election feeling thermometer ratings of one's own party and the opposing party.3

2 There were no questions in the pre-election wave about voting intentions in House or Senate elections.

3 For this analysis to obtain a larger sample size we used the full ANES sample including Internet and face-to-face interviews. However, results for the face-to-face sample only were very similar. Two percent of eventual voters who were undecided in the pre-election wave were not included in the analysis.

Not surprisingly, since the pre-election wave of the ANES survey was conducted only a few weeks before Election Day, there was a very strong relationship between pre-election voting intentions and vote choice. Over 98 percent of voters chose the same candi- date in both waves. Nevertheless, the results in Table 2 show that pre-election feeling thermometer ratings of the opposing party had a significant effect on party loyalty in the expected direction. The more negative their rating of the opposing party in the pre-election wave, the more likely voters were to support their own party's presidential candidate even after controlling for their pre-election candidate preference. These results are consistent with our causal ordering hypothesis. The results in Table 2 also show that strong partisans were significantly more likely to vote for their party's candidate than weak partisans or leaning independents even after controlling for pre-election voting intention. However, pre-election feeling thermometer rating of one's own party had no discernible effect on vote choice in this analysis.

5. Estimating the effects of negative partisanship on party loyalty

Based on the results in Table 1 we can estimate the effects of changes in feeling thermometer ratings of the opposing party on the probability of consistent loyalty under different incumbency conditions.4 For voters with running House and Senate incumbents from the opposing party, as the feeling thermometer rating of the opposing party increases from zero degrees to 50�, the expected probability of consistent loyalty falls from 89 percent to 48 percent; for voters with no running House or Senate incumbents, the ex- pected probability falls from 94 percent to 65 percent; finally, for voters with running House and Senate incumbents from their own party, the expected probability falls from 97 percent to 80 percent.

These results indicate that the impact of negative partisanship on loyalty is greatest for voters with opposing party incumbents and smallest for voters with incumbents from their own party. This finding is consistent with evidence on trends in cross-party voting in House and Senate elections. According to data from the ANES cumulative file, the loyalty rate of House voters with an incumbent from their own party increased only slightly between the 1980s and 2008e2012dfrom an average of 93 percent to an average of 95 percent. In contrast, the loyalty rate of House voters with an incumbent from the opposing party increased dramatically

4 In all calculations, feeling thermometer ratings of an individual's own party is set at the mean value and all other predictors are set at their medians.

Table 3 Results of regression analyses of house election outcomes, 1972e2014.

Year Unstandardized coefficients R Square

Presidential Partisanship Party of incumbent

1972 .434 13.7 .77 1974 .316 12.6 .70 1976 .712 11.9 .76 1978 .638 12.9 .74 1980 .656 13.1 .79 1982 .649 11.5 .78 1984 .614 13.1 .86 1986 .479 15.4 .83 1988 .519 15.4 .86 1990 .424 11.6 .75 1992 .636 9.9 .75 1994 .674 10.6 .83 1996 .726 9.4 .86 1998 .702 10.9 .87 2000 .631 12.1 .89 2002 .572 12.6 .89 2004 .626 11.3 .92 2006 .626 10.2 .85 2008 .649 9.6 .89 2010 .848 6.6 .88 2012 .854 4.9 .94 2014 .828 4.9 .92

Avg. 72e80 .551 12.8 .75 Avg. 82e90 .537 13.4 .82 Avg. 92e00 .674 10.6 .84 Avg. 02e10 .664 10.1 .89 Avg. 12-14 .841 4.9 .93

Note: Dependent variable is percentage of major party vote for Democratic House candidate in contested races. Estimates for intercepts not shown. All estimated coefficients are statistically significant at .001 level based on one-tailed t-tests. Sources: Gary Jacobson and data compiled by authors.

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e2218

between the 1980s and 2008e2012dfrom an average of only 53 percent to an average of 78 percent.

The results for Senate elections were similar to those for House elections: the loyalty rate of Senate voters in elections with an incumbent from their own party increased only slightly between the 1980s and 2008e2012dfrom an average of 89 percent to an average of 95 percent. In contrast, the loyalty rate of Senate voters in elections with an incumbent from the opposing party increased sharply during this time perioddfrom an average of only 66 percent during the 1980s to an average of 81 percent in 2008e2012.

The main impact of negative partisanship on recent House and Senate elections appears to be a substantial decline in the will- ingness of voters to support incumbents from the opposing party. This decline in the advantage of incumbency is a direct result of the increasing influence of partisanship in House and Senate elections.5

Voters now appear to be less concerned about which individual will represent their district or state and more concerned about which party will control the House or Senate. As we will see, this is making it increasingly difficult for candidates from the minority party to win seats in the House and Senate and, when they do manage to win against the party grain, to hold onto these seats.

6. Nationalization of U.S. House elections

Thus far we have demonstrated that the rise of negative parti- sanship has contributed to a substantial increase in party loyalty and straight-ticket voting. These results imply that voters no longer view House, Senate and local elections as separate arenas of competition from presidential elections. On the contrary, voters now view their choices in elections at all levels through the lens of negative partisanship: at all levels of government, the greatest concern of party supporters is preventing the opposing party from gaining power. For this reason, negative partisanship has nation- alized American elections.

An analysis of aggregate-level data on the results of U.S. House elections since the 1970s supports this conclusion. These data show very clearly that growing party loyalty and straight-ticket voting have produced a dramatic increase in the connection between presidential and House elections: the correlation between the Democratic share of the House vote and the Democratic share of the presidential vote in districts with contested races averaged .54 between 1972 and 1980, .65 between 1982 and 1990, .78 between 1992 and 2000, .83 between 2002 and 2010 and .94 in 2012e2014. In terms of shared variance, the relationship between presidential and House election outcomes is now three times stronger than it was in the 1970s.

These results indicate that there have been important changes in the determinants of House election outcomes over the past several decades. In particular, these results suggest the relative importance of district presidential partisanship may be increasing at the expense of more local factors, especially the personal advantage of incumbency. In order to test this hypothesis and to measure changes in the relative influence of presidential parti- sanship and incumbency on the outcomes of individual House races over time, we conducted regression analyses of these outcomes for every election between 1972 and 2014. Our dependent variable in these analyses is the Democratic percentage of the major party vote. The independent variables in our regression equations are the Democratic share of the major party presidential vote in each dis- trict in relation to the nation and the party of the House incumbent, coded as þ1 for contests with Democratic incumbents, 0 for open seat contests and �1 for contests with Republican incumbents.

5 See Jacobson (2015) for additional evidence of this trend in U.S. House elections.

The results in Table 3 show that there were drastic changes in the effects of our two independent variables during the five de- cades included in this analysis. The overall explanatory power of our model increased substantially over time. However, this increase was due entirely to the increased influence of district partisanship. An examination of the unstandardized regression coefficients shows that the effect of district presidential partisanship increased considerably over time with the largest increase occurring during the most recent time period. By contrast, the effect of incumbency decreased slightly during the 1990s and 2000s and then dropped dramatically in the most recent time period. Over these 22 elec- tions, the three largest estimated coefficients for the effect of presidential partisanship as well as the three smallest estimated coefficients for the effect of incumbency occurred in the three most recent elections.6 Compared with the 1970s and 1980s, the electoral fortunes of House incumbents now depend much less on how effectively they cultivate their constituencies and much more on the partisan makeup of those constituencies.

These results demonstrate that House elections have experi- enced a strong trend toward nationalization over the past several decades. The outcomes of House elections are now consistent with the results of presidential elections at the district level determined to a much greater degree than in the recent past by the relative strength of the Democratic and Republican presidential candidates in House districts. This shift is readily apparent in Fig. 5 which displays the trends in the percentage of House contests won by Democrats and Republicans in House districts that favor each party

6 The differences between the three most recent elections and earlier elections for both the district partisanship and incumbency variables are highly statistically significant (p < .001).

Fig. 5. Percentage of house winners consistent with presidential party advantage by decade. Sources: Gary Jacobson and data compiled by authors.

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e22 19

based on the results of presidential elections. The favored presi- dential party is determined here by the Democratic share of the major party presidential vote in the district compared with the Democratic share of the major party presidential vote in the nation. A district in which the Democratic share of the major party presi- dential vote is greater than the Democratic share of the major party presidential vote in the nation is considered a Democratic-leaning district while a district in which the Democratic share of the ma- jor party presidential vote is less than the Democratic share of the major party presidential vote in the nation is considered a Republican-leaning district.7

The results displayed in Fig. 5 show that since 1990 there has been a very sharp increase in the proportion of House contests won by the candidate of the favored presidential party. However, this increase has been much greater in Republican-leaning districts than in Democratic-leaning districts. Between the 1960s and the 1980s, Republicans won less than 60 percent of House races in districts that were more Republican than the nation in presidential elections. In contrast, during this time period, Democrats won about 80 percent of House races that were more Democratic than the nation in presidential elections. Since 1994, however, and especially since 2010, GOP candidates have enjoyed much more success in Republican-leaning districts. This improvement in Republican fortunes is directly related to the increasing connection between presidential and House voting.

In the most recent time period, 2012e2014, there has been an extremely strong relationship between presidential and House voting. As a result, in these elections Republicans have won a remarkable 95 percent of contests in Republican-leaning districts

7 By controlling for short-term factors favoring one party or the other in a specific presidential election, this measure provides a more accurate gauge of the under- lying partisan orientation of House districts than simply using the winner of the presidential election.

while Democrats have won 93 percent of contests in Democratic- leaning districts. This pattern of strong partisan consistency fa- vors Republicans because there are more Republican-leaning dis- tricts than Democratic-leaning districts. However, this Republican advantage is not new.

Republicans have long enjoyed an advantage in House elections as a result of the fact that, from the standpoint of winning indi- vidual seats, Democratic voters are inefficiently distributed across House districts (Erikson 1972, 2002; Chen and Rodden, 2009; Rodden, 2010; Sides and McGhee, 2013; Cohn, 2014). What has changed in recent years is not the proportion of Republican-leaning districts. It is the ability of Republicans to convert their advantage in district presidential partisanship into actual majorities of House seats and this change is directly attributable to the nationalization of House elections.

7. Nationalization of U.S. Senate elections

Senate elections have also become increasingly nationalized in recent years as a result of growing party loyalty and straight-ticket voting. The trend there has been very similar to the trend in House elections but even more dramatic because the connection between presidential and Senate elections was considerably weaker than the connection between presidential and House elections during the 1970s and 1980s. The average correlation between the Democratic share of the presidential vote and the Democratic share of the Senate vote in states with contested races has risen from .16 be- tween 1972 and 1980 to .25 between 1982 and 1990, to .42 between 1992 and 2000, to .66 between 2002 and 2010 and to .84 in 2012e2014. This means that, in terms of shared variance, the relationship is now more than four times stronger than it was during the 1990s and more than 25 times stronger than it was during the 1970s.

As was true for House elections, these results indicate that there have been important changes in the determinants of Senate

Table 4 Results of regression analyses of senate election outcomes by decade.

Decade Unstandardized coefficients R Square

Presidential Partisanship Party of incumbent

1972e80 .395 6.5 .35 1982e90 .362 10.0 .56 1992e00 .461 9.5 .59 2002e10 .602 9.2 .71 2012e14 .683 6.1 .86

Note: Dependent variable is percentage of major party vote for Democratic Senate candidate in contested races. Estimates for intercepts and election-year fixed effects not shown. All estimated coefficients are statistically significant at .001 level based on one-tailed t-tests. Sources: Data compiled by authors.

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e2220

election outcomes over the past three decades. And just as with House elections, these results suggest the relative importance of presidential partisanship has been increasing at the expense of more local factors, especially the personal advantage of in- cumbency. In order to test this hypothesis and to measure changes in the relative influence of presidential partisanship and in- cumbency on the outcomes of individual Senate races over time, we conducted regression analyses of these outcomes for each decade since the 1970s. We combined Senate elections by decade because of the relatively small number of Senate contests in each election year. As with House elections, our dependent variable in these analyses is the Democratic percentage of the major party vote in contested races.

The independent variables in our Senate regression equations are the Democratic share of the major party presidential vote in each state in relation to the nation, and the party of the incumbent, coded as þ1 for contests with Democratic incumbents, 0 for open seat contests and �1 for contests with Republican incumbents. In addition, because our data is measured at the state-level and the errors in explaining election outcomes in state i at time t are likely to be correlated with outcomes in that same state at time t � 6, we included election-year fixed effects in each model. For the sake of clarity in presentation, we did not include the estimates for the election-year fixed effects in Table 3.

The results in Table 4 show that the trends in Senate elections were very similar to those that we found in House elections. There were drastic changes in the effects of our two key independent variables during the five decades included in this analysis. The overall explanatory power of our model increased substantially over time and, as with House elections, this increase was due entirely to the increased influence of presidential partisanship. An examination of the unstandardized regression coefficients shows that the effect of state presidential partisanship increased consid- erably over time with the largest increases occurring during the past two decades. As was true for House elections, the effect of incumbency declined dramatically in the most recent time period. As a result, the electoral fortunes of Senate incumbents, like the electoral fortunes of House incumbents, now depend less on their personal popularity and more on the partisan makeup of their states than in the past.8

Just as was the case for House elections, the nationalization of Senate elections has resulted in a sharp increase in consistency between the outcomes of these two types of elections. We classified states as Democratic-leaning or Republican-leaning in the same way that we classified House districts. States in which the

8 The differences between the coefficients for the state partisanship and in- cumbency variables in the most recent period and the coefficients for the first three time periods are both highly statistically significant (p < .001).

Democratic presidential candidate's share of the major party vote was greater than his share of the national major party vote were classified as Democratic-leaning while those in which the Demo- cratic presidential candidate's share of the major party vote was less than his share of the national major party vote were classified as Republican-leaning. Fig. 6 displays the trends in the proportions of Senate contests won by the Democratic candidate in Democratic- leaning states and by the Republican candidate in Republican- leaning states over the past five decades.

The data displayed in this figure show that since the 1980s there has been a dramatic increase in the proportion of Senate races won by the advantaged party. This trend has occurred in both Democratic-leaning states and Republican-leaning states. As a result, Democratic-leaning states are now much more likely to elect Democratic senators than they were during the 1980s and Republican-leaning states are now much more likely to elect Republican senators than they were during the 1980s. Thus, in 2014, 33 of 36 Senate contests were won by the candidate of the advantaged partydDemocrats won 12 of 15 contests in states that were more Democratic than the nation in the 2012 presidential election and Republicans won all 21 contests in states that were more Republican than the nation in the 2012 presidential election.

The results displayed in Fig. 6 show that between the 1972e1980 elections and the 2012e2014 elections, both parties substantially improved their performance in states in which they were advantaged based on presidential election results. In each decade, however, Democratic candidates won a somewhat larger share of seats in Democratic-leaning states than Republican can- didates won in Republican-leaning states. In the 2012e2014 elec- tions, for example, Democratic candidates won 88 percent (30 of 34) of contests in Democratic-leaning states while Republican candidates won 80 percent (28 of 35) of contests in Republican- leaning states.

Until 2014, Democrats had fared better in recent Senate elec- tions than in recent House elections because of their success in winning Senate contests in states that lean Republican in presi- dential elections. The question now is whether, given the increasing nationalization of Senate elections, Democrats will be able to continue winning such contests in the future. If they cannot, we may see a Republican advantage in Senate elections similar the party's recent advantage in House elections.

8. A note on nationalization of state legislative elections

The rise of negative partisanship is affecting elections for state and local offices as well as federal offices. As a result of growing party loyalty and straight-ticket voting, party strength in state legislatures now reflects the results of presidential elections much more closely than in earlier decades. In 2012, the correlation be- tween the Democratic share of the presidential vote and the Democratic share of state legislative seats was .85. This was the strongest correlation between the Democratic presidential vote and Democratic state legislative strength for any election year since at least 1956. The average correlation between the Democratic share of the presidential vote and the Democratic share of state legislative seats increased from only .40 between 1972 and 1988 to .58 be- tween 1992 and 2000 and to .73 between 2004 and 2012.

The growing connection between presidential and state legis- lative elections has had a dramatic impact on party control of the nation's state legislatures. In 38 of the 49 states with partisan state legislative elections, the same party that won the 2012 presidential election controlled both chambers of the legislature in 2013. However, there was a substantial Republican advantage in control of state legislatures: Republicans controlled both legislative chambers in 21 of 23 states carried by Mitt Romney while

Fig. 6. Percentage of senate winners consistent with presidential party advantage by decade. Source: Data compiled by authors.

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e22 21

Democrats controlled both chambers in only 17 of 26 states carried by Barack Obama.

Republican success in recent state legislative elections has clearly been aided by the inefficient distribution of Democratic voters across legislative districts in many states and by Republican control of redistricting in a large number of states in the aftermath of the 2010 midterm election including several states that voted for Barack Obama in the 2008 and 2012 presidential elections. How- ever, the main driver of Republican gains in state legislatures in recent years has been the same transformation that has led to Republican gains in the U.S. House of Representativesdthe nationalization of elections as a result of growing partisan consis- tency in voting behavior.

During the 1980s and 1990s, Democrats controlled a large number of legislative chambers in states that were regularly voting for Republican presidential candidates including many states in the South. As recently as 2001, Democrats controlled 24 legislative chambers in the 30 states carried by George W. Bush in the 2000 presidential election. In the past decade, however, the rise of party loyalty and straight-ticket voting has enabled Republicans to take control of almost all of these legislative chambers. In 2013, Dem- ocrats controlled only three chambers in the 24 states carried by Mitt Romney in the 2012 presidential election. As a result of the growing connection between presidential and state legislative elections, the once clear divide between state politics and national politics has largely disappeared in most of the country.

9. Discussion and conclusions

Regardless of the direction or strength of their party identifi- cation, American voters in the 21st century are much more likely to hold strongly negative views of the opposing party than in the past. A growing proportion of Americans dislike the opposing party more than they like their own party. The rise of negative parti- sanship in the American electorate has contributed to the highest rates of party loyalty and straight-ticket voting in the past sixty

years. As a result, to a much greater degree than at any time in recent history, the outcomes of House and Senate elections reflect the relative strength of the presidential parties in the 435 House districts and the 50 states.

As the influence of presidential partisanship has risen, the per- sonal advantage of incumbency has declined. This development has made it much harder for incumbents to survive in districts or states that lean toward the opposing party in presidential elections. The overwhelming majority of House and Senate races are now won by the party that is favored based on the presidential vote in the dis- trict or state. This growing nationalization of elections has impor- tant implications for competition, representation and governmental performance.

First, our results suggest that structural forces are likely to continue to work in favor of Republican candidates in future House elections. The current Republican advantage in House elections is a result of the inefficient distribution of Democratic voters across districts combined with the increasingly partisan behavior of the American electorate. Republicans enjoy a similar advantage in Senate elections due to the extreme overrepresentation of sparsely populated rural states and the fact that Democratic voters are heavily concentrated in more populous states with large urban areas. Until 2014, Democrats were able to overcome this disad- vantage by winning a larger share of Senate contests in Republican- leaning states than Republicans were able to win in Democratic- leaning states. Given the increasing nationalization of Senate elections, though, Democrats may find it difficult to overcome this structural disadvantage in the future.

The rise of negative partisanship has also changed the repre- sentation of constituent preferences in both the House and the Senate. Representation today means almost exclusively partisan representation. There is little or no incentive for most members of the House or Senate to consider the views of supporters of the opposing party because it has become very difficult to persuade these voters to cross party lines and because few members need the votes of opposing partisans in order to secure reelection. For the

A.I. Abramowitz, S. Webster / Electoral Studies 41 (2016) 12e2222

most part, Republican members are only interested in the views of Republican voters and Democratic members are only interested in the views of Democratic voters.

Today, only a few members who represent districts or states that tilt toward the opposing party are likely to be concerned about winning over voters outside of their own party. However, these members are finding this task increasingly difficult as the personal advantage of incumbency diminishes. Thus, Republican gains in the 2014 Senate elections reflected the inability of Democratic in- cumbents to overcome their party's unpopularity in states such as Arkansas, Alaska, Louisiana and North Carolina.

Finally, the growth of partisan representation has changed the policy-making process in Washington. Given that most members of Congress now face incentives that discourage “reaching across the aisle” to work with members of the other party, we are unlikely to see major legislation passed with bipartisan support. This fact, combined with the regular occurrence of divided government, explains why confrontation and gridlock have become the “new normal” in Washington.

The rise of negative partisanship in the American electorate appears to be part of a vicious cycle of mutually reinforcing elite and mass behavior. Confrontational politics in Washington and in many state capitols is causing Democratic and Republican voters to develop increasingly negative views of the opposing party and to vote along party lines from the top of the ticket to the bottom. Negative views of the opposing party among voters, in turn, encourage political elites to adopt a confrontational approach to governing. Given these mutually reinforcing patterns of elite and mass behavior, negative partisanship is likely to remain an impor- tant feature of American politics for the foreseeable future.

References

Abramowitz, Alan I., 2013. The Polarized Public: Why American Government is So Dysfunctional. Pearson Longman, New York.

Abramowitz, Alan I., 2015. Partisan Nation: the rise of affective partisanship in the American Electorate. In: Green, John C., Coffey, Daniel J., Cohen, David B. (Eds.), The State of the Parties: The Changing Role of Contemporary American Parties, seventh ed. Rowman and Littlefield, New York, pp. 21e36.

Ansolabehere, Stephen, Snyder, James M., Stewart, Charles, 2001. Candidate posi- tioning in U.S. house elections. Am. J. Polit. Sci. 45 (1), 136e159.

Bafumi, Joseph, Shapiro, Robert Y., 2009. A new partisan voter. J. Polit. 71 (1), 1e24. Bafumi, Joseph, Herron, Michael C., 2010. Leapfrog representation and extremism: a

study of american voters and their members in congress. Am. Polit. Sci. Rev. 104 (3), 519e542.

Bartels, Brandon L., 2015. The sources and consequences of polarization in the U.S. Supreme court. In: Thurber, James, Yoshinaka, Antoine (Eds.), American Grid- lock. Cambridge University Press, New York (in press).

Bartels, Larry M., 2000. Partisanship and voting behavior, 1952e1996. Am. J. Polit. Sci. 44 (1), 35e50.

Campbell, Angus, Philip, E. Converse, Warren, E. Miller, Stokes, Donald E., 1960. The American Voter. John Wiley and Sons, New York.

Caruana, Nicholas J., McGregor, R., Michael, Laura, B. Stephenson, 2014. The power of the dark side: negative partisanship and political behaviour in Canada. Can. J. Polit. Sci. 1e19.

Chen, Joewi, Rodden, Jonathan, 2009. Tobler's Law, Urbanization, and Electoral Bias: Why Compact, Contiguous Districts are Bad for the Democrats. Department of Political Science, Stanford University. Unpublished paper.

Clark, Tom S., 2009. Measuring ideological polarization on the United States su- preme court. Political Res. Q. 62 (1), 146e157.

Cohn, Nate, 2014. Why democrats can't win the house. N. Y. Times. http://www. nytimes.com/2014/09/07/upshot/why-democrats-cant-win.html? abt¼0002&abg¼1 (accessed 09.02.15.).

Frey, William H., 2015. Diversity Explosion: How New Racial Demographics Are Remaking America. Brookings Institution, Washington, DC.

Green, Donald, Palmquist, Bradley, Schickler, Eric, 2002. Partisan Hearts and Minds: Political Parties and the Social Identities of Voters. Yale University Press, New Haven, CT.

Greenberg, Stanley B., 2004. The Two Americas: Our Current Political Deadlock and How to Break it. St. Martin’s Press, New York.

Greene, Steve, 1999. Understanding party identification: a social identity approach. Polit. Psychol. 20 (2), 393e403.

Hetherington, Marc J., 2001. Resurgent mass partisanship: the role of elite polari- zation. Am. Polit. Sci. Rev. 95 (3), 619e631.

Huddy, Leonie, Mason, Lilliana, Aaroe, Lene, 2015. Expressive partisanship: campaign involvement, political emotion and partisan identity. Am. Polit. Sci. Rev. 109 (1), 1e17.

Iyengar, Shanto, Hahn, Kyu, 2009. Red media, blue media: evidence of ideological selectivity in media use. J. Commun. 59 (1), 19e39.

Iyengar, Shanto, Sood, Gaurov, Lelkes, Yphtach, 2012. Affect, not ideology: a social identity perspective on polarization. Public Opin. Q. 76 (3), 405e431.

Izadi, Elahe, 2014. Congress sets record for voting along party lines. Natl. J. http:// www.nationaljournal.com/congress/congress-sets-record-for-voting-along- party-lines-20140203 (accessed on 27.03.15.).

Jacobson, Gary C., 2007. A Divided, Not a Uniter: George W. Bush and the American People. Pearson Longman, New York.

Jacobson, Gary C., 2015. It's nothing personal: the decline of the incumbency advantage in U.S. House elections. J. Polit. 77 (3) (in press).

Jones, Jeffrey M., 2015. In U.S., New Record 43% are Political Independents. Gallup Poll. http://www.gallup.com/poll/180440/new-record-political-independents. aspx?utm_source¼independent identification&utm_medium¼search&utm_ campaign¼tiles (accessed on 10.05.15.).

Keith, Bruce E., Magleby, David B., Nelson, Candice J., Orr, Elizabeth A., Westlye, Mark C., 1992. The Myth of the Independent Voter. University of Cal- ifornia Press, Berkeley.

Klar, Samara, 2014. Identity and engagement among political independents in America. Polit. Psychol. 35 (4), 577e591.

Kraushaar, Josh, 2014. The most divided congress ever, at least until next year. Natl. J. http://www.nationaljournal.com/2013-vote-ratings/the-most-divided- congress-ever-at-least-until-next-year-20140206 (accessed 27.03.15.).

Levendusky, Matthew S., 2013. Why do partisan media polarize voters? Am. J. Polit. Sci. 57 (3), 611e623.

Mann, Thomas E., Ornstein, Norman J., 2013. It's Even Worse than it Looks: How the American Constitutional System Collided with the New Politics of Extremism. Basic Books, New York.

Mason, Lilliana, 2013. The rise of uncivil agreement: issue versus behavioral po- larization in the american electorate. Am. Behav. Sci. 57 (1), 140e159.

Mason, Lilliana, 2015. ‘I disrespectfully agree’: the differential effects of partisan sorting on social and issue polarization. Am. J. Polit. Sci. 59 (1), 128e145.

McCarty, Nolan, Poole, Keith T., Rosenthal, Howard, 2008. Polarized America: the Dance of Ideology and Unequal Riches. MIT University Press, Cambridge, MA.

Medeiros, Mike, No€el, Alain, 2014. The forgotten side of partisanship: negative party identification in four anglo-American democracies. Comp. Polit. Stud. 47 (7), 1022e1046.

Mutz, Diana, 2006. How the mass media divide us. In: Brady, David, Nivola, Pietro (Eds.), Red and Blue Nation?, vol. 1. Hoover Institution Press, Stanford, CA, pp. 223e248.

Mutz, Diana, 2007. Effects of ‘in-your-face’ television discourse on perceptions of a legitimate opposition. Am. Polit. Sci. Rev. 101 (4), 621e635.

Prior, Markus, 2007. Post-Broadcast Democracy: How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. Cambridge Univer- sity Press, New York.

Rodden, Jonathan, 2010. The geographic distribution of political preferences. Annu. Rev. Political Sci. 13, 321e340.

Rose, Richard, Mishler, William, 1998. Negative and positive party identification in post-communist countries. Elect. Stud. 17 (2), 217e234.

Shor, Boris, McCarty, Nolan, 2011. The ideological mapping of american legislatures. Am. Polit. Sci. Rev. 105 (3), 530e551.

Sides, John, McGhee, Eric, 2013. Redistricting Didn't Win Republicans the House. The Washington Post. http://www.washingtonpost.com/blogs/wonkblog/wp/ 2013/02/17/redistricting-didnt-win-republicans-the-house (accessed on 07.02.15.).

Sinclair, Barbara, 2006. Party Wars: Polarization and the Politics of National Policy Making. University of Oklahoma Press, Norman, OK.

Stone, Geoffrey R., 2014. Our Politically Polarized Supreme Court? Huffingt. Post. http://www.huffingtonpost.com/geoffrey-r-stone/our-politically-polarized_b_ 5879346.html (accessed on 20.03.15.).

Theriault, Sean M., 2008. Party Polarization in Congress. Cambridge University Press, New York.