Kick the Bums Out?: A Structural Equation Model Exploring the Degree to which Mainstream and Partisan Sources Influence Polarization and Anti-incumbent Attitudes

lable at ScienceDirect

Electoral Studies 40 (2015) 210e220

Contents lists avai

Electoral Studies

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

Kick the bums out?: A structural equation model exploring the degreeto which mainstream and partisan sources influence polarization andanti-incumbent attitudes

Thomas J. Johnson a, *, Angela M. Lee b

a School of Journalism, University of Texas-Austin, USAb Emerging Media and Communication, School of Arts, Technology, and Emerging Communication, University of Texas at Dallas, USA

a r t i c l e i n f o

Article history:Received 5 July 2014Received in revised form15 July 2015Accepted 24 August 2015Available online 2 September 2015

Keywords:Anti-incumbentPolarizationMedia useMedia effect

* Corresponding author.E-mail address: [email protected] (T

http://dx.doi.org/10.1016/j.electstud.2015.08.0080261-3794/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

The public's anti-incumbent mood was at a then all-time high in the 2010 election with almost two-thirds (65 percent) agreeing that most members of Congress did not deserve re-election. Previousstudies have not directly examined whether and how different types of media influence anti-incumbentattitudes. This study relies on a 2010 state poll to examine how partisan and mainstream news sourcesinfluence anti-incumbent attitudes, controlling for political and demographic variables, and how po-larization mediates such effects. Using a structural equation model, this study finds that whereas relianceon partisan media is positively associated with anti-incumbent attitudes toward most Senators andrepresentatives, reliance on mainstream media is negatively associated with it. On the other hand,reliance on neither media type is statistically associated with anti-incumbent attitudes toward one's ownrepresentatives.

© 2015 Elsevier Ltd. All rights reserved.

When President Barack Obama called the results of the 2010midterm elections a shellacking, he wasn't engaging in hyperbole.The Republicans gained the majority in the House of Representa-tives, winning 63 seats, the Democrats' largest loss since 1948.Republicans won five additional Senate seats and gained five gov-ernors. The Republicans picked up 700 seats in state houses, themost Republicans in state legislatures since 1928 (Kerr, 2010;Murray, 2010). But the shellacking wasn't limited to Democrats asan anti-incumbent mood swept out three moderate RepublicanSenate primary candidates and swept in 90 new Congressionalcandidates, many associated with the Tea Party movement. Moretelling, 41 percent of those who voted in the 2010 primariesdescribed themselves as Tea Party supporters with 21 percentsupporting it strongly (Cook, 2010).

Polls confirmed that the public's anti-incumbent mood was at athen all-time high in 2010 with just under two-thirds (65 percent)agreeing that most members of Congress did not deserve re-election and slightly less than half (49 percent) arguing that theirown Congressional representative deserved to be voted back in

.J. Johnson).

(Saad, 2010), although after the government shutdown anti-incumbent attitudes were even higher (Steinhauser, 2013).Scholars agree public disenchantment with government can beconverted into an anti-incumbent spirit when a troubled economyor a key issue such as an unpopular war can cause peripheral voters,who have little trust in government (Hetherington, 1999), to comeout to vote (Hansford and Gomez, 2010).

Although previous research has explored why people vote for oragainst an incumbent, studies have not directly examined whethermedia may contribute to or reduce anti-incumbent attitudes, andwhether types of media (mainstream vs. partisan sources) havedifferent influences on anti-incumbent attitudes. Content analysesare split on whether national media tend to provide favorable orunfavorable coverage to incumbents (Graber, 2006; Lowry andShidler, 1998). Further, studies are split on the role of media inpromoting or damaging incumbents. Some studies find that,particularly in low-profile campaigns like Congress, media mayhave a pro-incumbent bias as incumbents have better relations andaccess to the media (Fico et al., 2008). Others suggest that themedia tend to provide additional scrutiny to those in office (Graber,2006; Patterson, 1993), while some scholars have found noinfluence.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220 211

This study relies on a 2010 pre-election Texas state poll, spon-sored by The University of Texas at Austin and the Texas Tribune, toexamine the degree to which partisan sources (cable news, talkradio, individual bloggers, and Websites with a mix of news andcommentary such as Drudge Report and Huffington Post) as well asmainstream news sources (broadcast news websites, the websitesof major national newspapers and thewebsites of local newspapersand TV stations) influence anti-incumbent attitudes. Specifically,this study employs a structural equation (SEM) model to examinethe degree to which partisan and traditional (national and local)media influence anti-incumbent attitudes both directly and indi-rectly through their relationship with polarization. This study willalso explore differences between the Republicans, Democrats andTea Party supporters in whether they hold anti-incumbent atti-tudes. Compared to standard regression analysis, SEM is superior intesting theoretical models because it accounts for the interrela-tionship among all modal variables (Iacobucci et al., 2007). How-ever, compared to experiments, cross-sectional survey study suchas this one is ultimately unable to offer definitive causal claims. Assuch, findings from this study should be interpreted with thiscaveat in mind.

This study contends that those who rely on more partisansources of news, which often limits their contact to those withlikeminded views, will be positively associated with polarization. Inturn, such polarizationwill be negatively associatedwith support ofthe government, specifically increased anti-incumbent views. Thisstudy also investigates the relationship among mainstream news,polarization and anti-incumbent attitudes, hypothesizing thatreliance on traditional news sources will be negatively associatedwith polarization, although evidence isn't clear whether main-stream news is associated with more or less support for in-cumbents. This study also hypothesizes that Tea Party members,who are more extreme in their views than either Democrats orRepublicans, will be more polarized than either party in their po-litical views and be more likely to hold anti-incumbent attitudes.

In this age of easily available, and overly abundant news andpolitical information where citizens may more actively select thetype of information they are exposed to (Garrett, 2009), this studytakes on a greater importance in understanding the role of digitaland traditional media on levels of anti-incumbent bias. Wholesalechange in government officials endanger democracy's stability andcan lead to the electorate's increased ideological polarization(Iyengar and Hahn, 2009; Stroud, 2010).

1. Anti-incumbent attitudes

Researchers have long noted that incumbents for Congress andlocal elections have considerable built-in advantages from bettername recognition to a built-in staff. But in the 2010 election in-cumbents seemed as vulnerable as ducks in a shooting gallery. Notonly did approximately two thirds of voters think that most Con-gressmen shouldn't be re-elected, 60 percent said they would prefera Congressional representative that had never served in Congressand three-quarters expected that if most representatives werereplaced, this would change Congress for the better (Saad, 2010).

Few trends in U.S. public opinion have produced more enduringconcern among scholars and commentators than declining levels ofpublic trust in government (Hibbing and Theiss-Morse, 1995),which, according to Gallup, has fallen from near 80 percent in theEisenhower administration to 15 percent in October 2013. Somescholars argue that anti-incumbent attitudes reflect a distrust ofnational institutions that has been engrained in U.S. political cul-ture. This has led to an increased public belief that politicians aremotivated primarily by self-interest rather than the good of thecountry as well as a growing public dissatisfaction with the overall

direction of government (Miller, 1974). Such anti-incumbent atti-tudes pose a direct threat to democratic system. For instance, Millerwarned, “A democratic political system cannot survive for longwithout the support of a majority of its citizens” (1974: 951). Not allresearchers, though, believe that distrust in government, reflectedin anti-incumbent attitudes, is a negative thing. Some scholarsargue that declining trust in government is natural, cyclical and anindicator of a healthy democracy because citizens vote outentrenched politicians who are guided by their own rather than thecountry's interest (Citrin and Luks, 2001).

Scholars agree, however, that the public more strongly supportstheir individual Congressional representative than Congress as awhole (Fenno, 1975) because while Congress as a whole is judged(poorly) based on their legislative performance, people grade theirown Congressional representatives based on perceived service totheir constituents and their personality (Parker and Davidson, 1979).

2. Polarization

With the increased political gridlock in Congress, scholars havefocused more attention on whether the public is becoming morepolarized, that is, taking more extreme stands on issues, whichwould create a more fractured society (Stroud, 2010, 2011). Someresearchers have argued that people are increasingly discussingpolitics with likeminded individuals, both in person and online(Mutz and Martin, 2001). Online discussions with likeminded in-dividuals can create an “echo chamber” effect where those whospend more time talking with likeminded individual receive moreinformation that supports their perspective as well as aligns theirviews with more extreme views to be more in line with perceivedgroup norms and opinions (Stroud, 2010).

While there is broad agreement that discussions with likemindedgroups can produce polarization, scholars are split about whetherthe public has indeed become more polarized in its political views(Baldassarri and Bearman, 2007). Indeed, most studies suggest thatwhile the general public had become more ideologically split oncertain “hot button” issues, such as gay marriage, that overall thepublic has not become more ideologically split (Baldassarri andBearman, 2007), although a recent Pew Research study (2014)found the percentage of those who express consistently liberal orconservative views has doubled over the last two decades. However,the public may appear more polarized because the parties them-selves have increasingly abandoned the middle and taken moreextreme stances. Therefore, political elites and strong partisans areincreasingly toeing the party line (Abramowitz, 2010; Baldassarriand Bearman, 2007; Carmines et al., 2012; Stroud, 2010).

3. Polarization and anti-incumbent attitudes

Congressional scholars have clearly documented that the twoparties in Congress have become increasing polarized since the1970s (Stonecash et al., 2002), which appears to run counter to logic:Increased polarization reduces confidence in Congress because thepublic believes more can be achieved when Congressional memberswork across the aisle to pass legislation (Harbridge and Malhotra,2011; Ramirez, 2009), and that Congressmen need to representtheir constituents or face defeat (Canes-Wrone et al., 2002).

However, while it may be true that independents, weak partisansand strong partisans all believe Congress as a whole should act in abipartisan manner, partisan voters are more likely than in-dependents and weak partisans to support individual members ofCongress who vote for policies they support (Grynaviski, 2010;Harbridge and Malhotra, 2011; Moon, 2015). Because the partiesbestow more resources to Congressional members who reliably toethe party line from campaign dollars to plumb committee

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220212

assignments (Smith, 2007), members who come from safe districtswhere the majority of people come from the same party perceive itadvantageous to favor partisanship over bi-partisanship (Ensley,2012). Also, because partisans are more likely to vote than thosewith weak party affiliations (Pew Research Center, 2014), Congres-sional members often have clear incentives to act in a partisanmanner, increasing polarization (Harbridge and Malhotra, 2011).

Polarization may also have an effect on attitudes toward oppo-sitional candidates. Partisan news coverage may lead viewers toperceive the other party more negatively, to trust them less as wellas to be less supportive of bipartisanship (Levendusky, 2013; PewResearch, 2014).

While Tea Party supporters are much more likely to considerthemselves Republicans than either Independents or Democrats,scholars suggest that it would be a mistake to simply considerthemselves the conservativewing of the Republican Party. Tea PartyRepublicans have been found to be muchmore polarized than Non-Tea Party Republicans, holding much more conservative views,having different issue priorities and being much less likely to bewilling to compromise their views (Aldrich et al., 2013). Further-more, Tea Party members are increasingly taking negative views ofthe Republican Party and their leader House Speaker John Boehner.The Tea Party's increasing negative view of the Republican Partyand its leaders is in line with the party's general anti-Washingtonand anti-incumbent views (Aldrich et al., 2013; Carson andPettigrew, 2013).

While scholars are split onwhether the public has becomemorepolarized (Baldassarri and Bearman, 2007), they do agree that po-larization has potentially corrosive effects. Polarization may leadvoters to trust the other party less and perceive them more nega-tively (Levendusky, 2013). Therefore polarization may increaseanti-incumbent attitudes more generally, but it is unclear the de-gree it would influence attitudes toward your own candidates.

H1: Those with more polarized attitudes will be more likely tosay that most Senators and representatives should not be re-electedRQ1: Will those with more polarized attitudes be more likely tosay that their own representatives should not be re-elected?

Tea Party Republicans have been found to be much morepolarized than Non-Tea Party Republicans (Aldrich et al., 2013;Arceneaux and Nicholson, 2012), and express anti-Washingtonand anti-incumbent views (Aldrich et al., 2013; Carson andPettigrew, 2013). Thus Tea Party members should hold more anti-incumbent views than Democrats and Republicans.

H2: Those who identify themselves as Tea Party members willreport a) they believe most senators and representatives shouldnot be re-elected and b) their own representatives should not bere-electedRQ2: Will a) Republicans and b) Democrats report that theybelieve most Senators and representatives should not be re-elected?RQ3: Will a) Republicans and b) Democrats report that theybelieve their own representatives should not be re-elected?

4. Political parties and media use

Democrats, Republicans and Tea Party members vary greatly inwhichmedia they rely upon and perceive as credible. Democrats andTea Party members are mirror opposite in terms of what media theyuse and judge as credible (Kaye and Johnson, 2014; Pew ResearchCenter, 2011). Political blogs and Fox News Channel were given the

statistically highest credibility ratings by Tea Partiers and the lowestratings by Democrats. Conversely, newspapers, broadcast televisionnews, CNN, MSNBC, and parody news are rated most credible byDemocrats but lowest by Tea Partiers (Kaye and Johnson, 2014).Similarly Pew found that while solid liberals were more likely thanother groups to favor news from NPR, the New York Times and TheDaily Show and also more frequent users of daily newspapers andMSNBC, Tea Party members were significantly more likely than anyother groups to prefer news from Fox News, Glenn Beck and RushLimbaugh. But differences have also found between Republicans andTea Party members. Republicans are more likely to rely on morebalanced sources such as network news and daily newspapers thanTea Party members, who are more likely than Republicans to rely onmore partisan sources such as Fox News, Glenn Beck and RushLimbaugh (Pew Research, 2011). Similarly Republicans judge news-papers, broadcast news and CNN significantlymore credible than TeaParty members, who judge political blogs more believable than Re-publicans (Kaye and Johnson, 2014).

Studies suggest, then that while different parties seek out thesources they believe most support their worldview, overall TeaParty supporters favor partisan sources and Democrats and Re-publicans more mainstream ones.

H3: Tea Party supporters will be more likely than both Re-publicans and Democrats to rely on partisan sources.H4: Republicans and Democrats will be more likely than TeaParty supporters to say they rely on mainstream a) national andb) local sources.

5. Anti-incumbent attitudes and media use

Themedia have routinely faced accusations of media bias. Whilepartisan bias has been much debated among scholars and publicobservers, a meta-analysis of 99 studies of presidential electionover six decades indicated while individual stories may show bias,overall there is little evidence the mainstream media are biased infavor of either party (D'Alesso, 2012).

While there is little evidence of partisan bias, it is possible thatthere may be other types of structural biases such as incumbency.Several studies looking at media coverage at local, statehouse andCongressional level have found a pro-incumbent bias. In theselower-profile races incumbents have built-in advantages because oftheir routine access to the media and ability to provide service tothe district that attracts media attention (Fico et al., 2008).

However, scholars debate the advantage of incumbency in morehigh-profile races such as governor, senator and president. Graber(2006) advanced an anti-incumbent bias hypothesis, noting thatbecause incumbents have records, they must run on these recordsand their records come under attack from the press. Similarly,studies have suggested that front-runners and incumbents mayreceive negative coverage because the media spent time buildingthese candidates up so that they should receive harsher scrutinythan their challengers (Patterson, 1993). But the anti-incumbenthypothesis has received mixed support. Some studies have foundthat in gubernatorial and presidential races (Fico and Cote, 2002)challengers do receive more and better coverage because, as Graber(2006) suggests, the media recognize that incumbents have a built-in advantage in ability to “make news” and therefore journalistscompensate by paying more attention to challengers. Other studiessuggest structural factors, particularly media norms of trying to befair and balanced, lead to both incumbents and challengersreceiving relatively equal coverage (D'Alesso, 2012). Therefore,studies are split on whether mainstream media coverage leads toanti-incumbent attitudes (Fico and Cote, 2002; Patterson, 1993) or

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220 213

not (D'Alesso, 2012).Researchers argue that how themedia cover the incumbentmay

influence how much support the incumbent earns. How muchcoverage candidates in Congressional races receive is oftencontingent on how much support they are perceived to have, andbecause incumbents typically have a built-in advantage in bothname recognition and financial support, they typically receivemorecoverage than challengers. Voters use amount of coverage as a cueon how serious the candidate is and thereforewhich candidate theyshould support. Thus, several studies have found than mainstreammedia use is connected to voting for the incumbent (Prior, 2006).However, several studies have failed to find a relationship betweenmedia use, particularly television news, and incumbent advantage(Ansolahehere et al., 2006). This may be because if the race iscompetitive, the media will turn attention to both candidates,advantaging challengers who often lack the financial support of theincumbents (Reynolds and Stewart, 1990).

Little attention has been paid specifically to partisan media andanti-incumbent attitudes, but studies of blogs, talk radio and cablenews suggest that the highly caustic, one-sided negative attacksagainst politicians could lead to anti-government attitudes (Prior,2013). Political blogs, because they tend to be one sided and linkto likeminded blogs, polarize the electorate by causing users to takemore extreme political stances (Lawrence et al., 2010). Conse-quently, blogs have been found to foster increased distrust anddislike of government institutions (Lawrence et al., 2010). Talk radiohosts such as Rush Limbaugh consistently attack politicians andblame the government for society's problems (Cappella andJamieson, 1997), which can lead to increased political cynicismand government distrust (Sobieraj and Berry, 2011; Jones, 2002).Blogs, talk radio and pundits on cable news are all criticized fortheir incivilityddiscussion that is rude, hostile, emotional and thatin general violate social standardsdand this incivility has beenblamed for increasing public distrust of government and its officials(Sobieraj and Berry, 2011). Because the leading voices of blogs, cablenews and talk radio are largely conservative and because partisansare more likely to hold anti-government and anti-incumbent atti-tudes when their party is out of office (Barfumi and Shapiro 2009;Grynaviski, 2010), partisan media use should be positively associ-ated with anti-incumbent attitudes among Republicans and TeaParty members in the 2010 electionwhen the Democrats controlledboth houses of Congress.

While studies are mixed onwhether mainstreammedia leads tosupport of incumbents, partisan media should increase generalanti-incumbent attitudes. It is unclear whether partisan media willaffect perceptions of one's own Congressional candidates:

H5: Heavy reliance on partisan sources will be positively asso-ciated with respondents' saying that most Senators and repre-sentatives should not be re-elected.RQ4: Will heavy reliance on partisan sources be positivelyassociated with respondents‘ saying that their representativesshould not be re-elected?RQ5: Will heavy reliance on mainstream national or localsources be positively associatedwith respondents‘ saying that a)most Senators and representatives and b) their representativesshould not be re-elected?

6. Polarization and the media

Scholars may argue about what portion of the public has takenon more extreme political views, but few dispute the claim that themedia appear to play a central role in politically interested in-dividuals becoming more polarized (Frederico and Hunt, 2013;

Stroud, 2010). As Stroud notes (2010), the media are the mainway that elites transmit their opinions to the public. Also, the rise ofmore partisan sources from cable news and talk radio to politicalwebsites, blogs and online forums has exponentially increased theamount of partisan information available as well as individual'sability to seek out supportive information. While a plurality of in-dividuals still say they want a balanced perspective for news andinformation, the percentage of people looking for political newswith a point of view has increased from 26% in 2004 to 34% by 2010(Pew Research, 2010b).

Partisans are more likely to seek out information that supporttheir views than non partisans and are thereforemore likely to seekout more partisan sources such as blogs (Johnson and Kaye, 2013;Lawrence et al., 2010), political websites (Garrett, 2009; Stroud,2011), and cable news (Baum and Goeling, 2008).

Only a few studies have directly explored the relationship be-tween media and polarization (Knobloch-Westerwick, 2012; Prior,2013; Stroud, 2010), with most studies finding that exposure topartisan content leads to increased polarization (Stroud, 2010;Knobloch-Westerwick, 2012). However, Johnson and Kaye (2012)discovered that strength of party ties rather than exposure topartisan content was a more powerful predictor of polarization.

Studies indicate that partisan media appear to play a central rolein politically interested individuals' becoming more polarized(Frederico and Hunt, 2013; Stroud, 2010), with most studies findingthat exposure to partisan content leads to increased polarization(Knobloch-Westerwick, 2012; Stroud, 2010). While studies suggestmainstream sources, because they strive to present different sidesof an issue, may reduced polarization (Wojcieszak and Rojas, 2011),other studies suggest that both Democrats and Republicansperceive that sources that strive to be balanced, such as broadcastnews, national newspapers like The New York Times and cablenetwork like CNN, present a liberal point of view (Kaye andJohnson, 2014; Pew Research, 2011; Stroud, 2010, 2011), thisstudy poses the following hypothesis and research question:

H6: Heavy reliance on partisan sources will be positively asso-ciated with high levels of polarization.RQ6: Will heavy reliance on mainstream national or localsources be positively associatedwith high levels of polarization?

Because this study argues that polarization will lead to anti-incumbent attitudes, it will explore whether the relationship be-tween reliance on partisanship and mainstream sources influenceanti-incumbent attitudes indirectly through polarization:

RQ7: Will heavy reliance on partisan sources be indirectlyrelated to the belief that a) most Senators and representativesand b) their representatives should not be re-elected throughpolarization?RQ 8: Will heavy reliance on mainstream national or localsources be indirectly related to the belief that a) most Senatorsand representatives and b) their representatives should not bere-elected through polarization?

Below is a visual representation of the proposed pathmodel (seeFig. 1):

7. Method

7.1. Data

The Texas Statewide Survey's 2010 pre-election poll was used toexamine the degree to which partisan sources (e. g., talk radio andblogs) and mainstream news sources (e.g. broadcast news and

Fig. 1. Conceptual path model.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220214

newspaper websites) influence anti-incumbent attitudes, control-ling for political and demographic variables. The study will alsoexplore whether media influence anti-incumbent attitudes directlyor more indirectly through their relationship with polarization.While the use of cross-sectional survey data prevents this studyfrom arguing definitive causal claims, we hope this study will paveway for better research designs in the future to examine the causalrelationships proposed in our model.

YouGovPolimetrix surveyed 914 respondents in its October 2010statewide study, and used post-stratification weighting to matchthe state's demographic characteristics, bringing the sample downto 800. Several news organizations rated YouGov as one of the mostaccurate polling organizations in the 2012 election (Byers, 2012;Matthews, 2012).

7.2. Variables

The media variables were adapted from the Pew ResearchCenter Biennial Media Consumption Survey (2010a), measured ona four-point scale of never, hardly ever, sometimes and regularly.

7.2.1. National mainstream mediaTo measure national mainstream media, and to give equal

weight to print and TV news media, national nightly network newsand TV station websites were summed and averaged to create a“national TV” variable. Then, national TV and national newspaperwebsites were summed and averaged, with a higher value repre-senting more frequent consumption (range ¼ 1e4, M ¼ 2.33,SD ¼ .85).

7.2.2. Local mainstream mediaTo construct this variable, local news, printed version of local

community newspapers and websites of local newspapers weresummed and averaged, with a higher value denoting more frequentconsumption (range ¼ 1e4, M ¼ 2.77, SD ¼ .76; Cronbach'sAlpha ¼ .55).

7.2.3. Partisan mediaTo assess partisan media consumption, the following four news

sources were summed and averaged, with a higher value indicating

more frequent consumption: 1) Cable news. 2) Talk radio. 3) Indi-vidual bloggers. 4) Websites with a mix of news and commentary(range 1 ¼ 4, M ¼ 2.40, SD ¼ .75, Cronbach's Alpha ¼ .61).

7.2.4. PolarizationTo operationalize polarization, we relied on the questions asking

how well the term “strong leader” described Texas gubernatorialcandidates Republican Rick Perry and Democrat Bill White, sub-tracting White's score from Perry's. Consistent with previousliterature (see Stroud, 2010), absolute values for polarization werecomputed (range ¼ 0e10, M ¼ 2.33, SD ¼ 3.09).

7.2.5. Anti-incumbent sentimentsThis study estimates two anti-incumbent sentiments, taken

from a May 2010 USA Today/Gallup Poll: One toward “most rep-resentatives,” and the other toward “one's own representative.” Thevariables were recoded in a way that a higher value indicates moreanti-incumbent sentiments. For “most representatives,” thefollowing two items were summed and averaged: 1) Most repre-sentatives deserve re-election. 2) Most Senators deserve re-election(range¼ 1e4,M¼ 3.26, SD¼ .77, Cronbach's alpha¼ .92). To assessone's anti-incumbent sentiment toward one's own representative,the following item was used: Own representative deserves re-election (range ¼ 1e4, M ¼ 2.71, SD ¼ 1.10).

7.2.6. Political partiesThree dummy variables are used to assess the effect of political

party on anti-incumbent sentiments. For Republicans, While Re-publicans were coded as 1, Democrats and Independents are codedas 0 (M ¼ .39). For Democrats, while Democrats were coded as 1,Republicans and Independents are coded 0. (M ¼ .32) Anothervariable was also used to identify Tea Party members where TeaParty was coded as 1, and non-Tea Party is coded as 0 (M ¼ .33). Forthese three variables, those self-identified as “Not sure” or “Other”were coded as missing variables.

7.3. Control variables

The following variables were entered into analysis as controlvariables: Political interest: 1. Not at all interested. 2. Not very

1 The significant Chi-square test is expected given the large sample size used inthis study. Conventional standards denote the following criteria desirable forgoodness of fit statistics: RMSEA less than .08; CFI and TLI equal to or greater than.9; and SRMR less than .05.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220 215

interested. 3. Somewhat interested. 4. Extremely interested(range ¼ 1e4, M ¼ 2.51, SD ¼.69). Political ideology: 1. Extremelyliberal e 7. Extremely conservative (range ¼ 1e7, M ¼ 4.91, SD ¼1.71). Political knowledge: 1. One correct answer. 2. Two correctanswers. 3. Three correct answers. Likelihood to vote: 1. Nevervoted. 2. Voted once or twice. 3. Voted some of the times. 4. Votedalmost every time. 5. Voted every time (range ¼ 0e3, M ¼ 1.80,SD ¼ 1.01). State of U.S. economy: 1. A lot better off. 2. Somewhatbetter off. 3. About the same. 4. Somewhat worse off. 5. A lot worseoff (range ¼ 1e5, M ¼ 3.67, SD ¼ 1.23). Age (range ¼ 18e85,M ¼ 50.64, SD ¼ 14.73), gender (male coded as 1, female 0)(range ¼ 0e1,M ¼ .45, SD ¼ .50), race (white coded as 1, non-whiteas 0) (range¼ 0e1,M ¼ .68, SD ¼ .47), and education (range ¼ 1e6,M ¼ 3.39, SD ¼ 1.40).

While survey studies can never account for all spurious factors,which contribute to their relatively low internal validity, we hopethat our inclusion of the nine key control variables will help usaccount for the possible effect of other variables on our proposedpath model. Nonetheless, these variables may not control for thepossible effect of other latent variablesdthose that cannot bedirectly accounted for by a survey questiondsuch as extremism ofideology on the endogenous variables in the proposed model,Therefore, findings from this study should be interpreted with thiscaveat in mind.

7.4. Statistical analysis

To test the hypotheses and research questions in the proposedSEM model, ordinary least square (OLS) multiple regression ana-lyses were conducted, holding constant the aforementioned controlvariables. To assess the direct and indirect effects estimated in therest of the study, simultaneous maximum likelihood regressionanalyses under the structural equation modeling (SEM) frameworkwere performed using MPlus 7.0. Specifically, this study tests forthe direct effect of media use on polarization, direct effect of po-larization on anti-incumbent sentiments, and indirect effects ofmedia use on anti-incumbent sentiments through polarization.SEM allows for a holistic evaluation of the interrelationship amongthe observed variables, and also allows for more accurate estimatesthat simultaneously control for all other variables in the full model.

8. Results

8.1. Sample demographics

The mean age of the sample is 50.64, SD ¼ 14.73; 45.2% of thesample is male; 68.3% of the sample is White; 30.6% of the samplehas a high school degree or less, 37.5% some college, 20.4% a collegedegree, and 9.5% a post-graduate degree. Moreover, 39.3% of thesample is a self-identified Republican, 32.2% a self-identifiedDemocrat, and 33.4% a member of the Tea Party.

8.2. Profile of media users

While the three media use variables (mainstream national,mainstream local and partisan) are not mutually exclusive (i.e., therespondents could be consumers of all three news media), they arepredicted by different demographic and political variables. Specif-ically, those who are white (b ¼ .22, p < .001), believe the economyis getting better (b ¼ �.30, p < .001), more likely to vote (b ¼ .08,p < .05), more liberal in ideology (b ¼ -.13, p < .01), and have highereducation (b ¼ .14, p < .001) use more mainstream national newsmedia. Thosewho havemore political knowledge (b¼ .19, p < .001),white (b ¼ .11, p < .05), believe the economy is getting better (b ¼ -.15, p < .001), and more likely to vote (b ¼ .18, p < .001) use more

mainstream local news media. And those who have less politicalknowledge (b ¼ -.11, p < .01), are white (b ¼ .35, p < .001), havemore political interest (b ¼ .11, p < .01), have more conservativeideology (b ¼ .19, p < .001), and have higher education (b ¼ .14,p < .001) use more partisan media.

8.3. Polarization and anti-incumbent attitudes

H1 hypothesizes that more polarized attitudes are positivelyassociated with more anti-incumbent attitudes toward most Sen-ators and representatives, but it is not supported by the data (n.s.),controlling for everything else. RQ1 asks whether the same patternis found regarding one's own representatives, and the data suggestpolarization also has no effect on one's own representatives (n.s.).Overall, the data suggest that polarization has no effect on anti-incumbent attitudes.

8.4. Political parties and media use

H2 hypothesizes that Tea Party members will report strong anti-incumbent attitudes, and is supported by the data. Specifically, TeaParty members are more likely to hold anti-incumbent attitudes(b ¼ .10, p < .05) than non-Tea Party members when it comes tomost Senators and representatives. But Tea Party members' anti-incumbent attitudes toward their own representative are not sta-tistically different from that of non-Tea Party members, controllingfor everything else. In response to RQ2, while Republicans are notmore likely than non-Republicans to hold anti-incumbent attitudestoward most Senators and representatives, Democrats are morelikely than non-Democrats to hold less anti-incumbent attitudes inthat they are more likely to believe most Senators and represen-tatives deserve to be re-elected (b ¼ �.22, p < .001), controlling foreverything else. Whereas RQ2 explores Republicans and Demo-crats' anti-incumbent attitudes toward most Senators and repre-sentatives, RQ3 focuses on respondents’ own representative. Thedata suggest that Republicans are more likely than non-Republicans to believe their own representatives should be re-elected (b ¼ �.16, p < .01), whereas Democrats are not statisti-cally different from non-Democrats. See Table 1.

H3 hypothesizes that that Tea Partymembers will bemore likelythan both Republicans and Democrats to rely on partisan sources,and is supported by the data (b ¼ .13, p < .01). H4 hypothesizes thatRepublicans and Democrats will be more likely than Tea Partymembers to say they rely on (a) mainstream national sources(b ¼ .11, p< .05) and (b) mainstream local sources (b ¼ .12, p<. 05),and it is supported by the data. See Table 2.

8.5. Goodness of fit statistics

For hypotheses and research questions below that assess directand indirect effects, simultaneous multiple regression analysesusing maximum likelihood estimation under the structural equa-tion modeling (SEM) framework was conducted. A variety ofgoodness of fit statistics were performed for the estimated model:X2(27) ¼ 193.34, p < .001, RMSEA ¼ .09, CFI ¼ .83, TLI ¼ .58, andSRMR ¼ .05.1 Overall the estimated model has an acceptable fitalthough some of the estimates are better than others. In the dis-cussion we will discuss possible explanations for these goodness offit statistics.

Table

1OLS

regression

analyses

ofpredictorsof

anti-incu

mbe

ntsentimen

ts.

Pred

ictors

H1

RQ1

H2a

H2b

RQ2a

RQ2b

RQ3a

RQ3b

Mostreps

Ownreps

Mostreps

Ownreps

Mostreps

Mostreps

Ownreps

Ownreps

Age

.03

�.02

.04

�.02

.05

.07

.00

.00

Gen

der

(Male)

.11**

.00

.10**

.00

.09*

.08*

�.02

�.01

Race(W

hite)

.08*

.00

.07

.00

.09*

.05

/02

�.01

Political

Interest

.01

.06

.00

.06

.01

.02

.03

.05

Political

Know

ledge

.03

�.07

.03

�.07

.04

.02

�.05

�.06

Stateof

U.S.E

con.

.34***

.01

.31***

.02

.34***

.27***

.04

.01

Like

lihoo

dto

Vote

�.01

.02

�.02

.02

�.02

�.01

.02

.02

Ideo

logy

.14**

�.22

***

.12**

�.22

***

.16**

.05

�.17

**�.25

***

Education

.07

.05

.07

.05

.07

.06

.05

.06

TeaPa

rty

ee

.10*

�.02

ee

ee

Rep

ublican

ee

�.04

e�.16

**e

Dem

ocrat

ee

e�.22

***

e�.02

Polarization

.07*

�.04

ee

ee

ee

Mod

elF(10,

632)

¼22

.57***

F(10,

608)

¼3.76

***

F(10,

832)

¼22

.73***

F(10,

608)

¼3.69

***

F(10,

585)

¼19

.38***

F(10,

585)

¼22

.00***

F(10,

562)

¼4.67

***

F(10,

562)

¼3.68

***

R2

.26

.06

.26

.06

.25

.27

.08

.06

Note:

Celle

ntriesarestan

dardized

coefficien

ts.*p<

.05,

**p<

.01,

***p<.001

.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220216

8.6. Anti-incumbent attitudes and media use

H5 hypothesizes that heavy reliance on partisan sources will bepositively associated with anti-incumbent attitudes toward mostSenators and representatives, and it is supported by the data(b ¼ .29, p < .001). RQ4 asks whether reliance on partisan sourceswill be positively associated with anti-incumbent attitudes towardone's own representatives, and the result is not statistically sig-nificant (n.s.), controlling for everything else. Turning to reliance onmainstream sources, RQ5a asks whether reliance on mainstreamnational or local newsmediawill be positively associated with anti-incumbent attitudes toward most Senators and representatives,and the data suggest that whereas use of mainstream nationalnews media is negatively associated with anti-incumbent attitudestoward most Senators and representatives (b ¼ �.25, p<.001), useof mainstream local news media is not statistically associated withanti-incumbent attitudes toward most Senators and representa-tives (n.s.), controlling for everything else. RQ5b asks whetherreliance on mainstream national or local news media will bepositively associated with anti-incumbent attitudes toward one'sown representatives, and it is supported by the data (b ¼ .17,p < .001), whereas use of mainstream local news media is notstatistically associated with anti-incumbent attitudes toward one'sown representatives (n.s.), controlling for everything else.

Therefore, whereas reliance on partisan media is positivelyassociated with anti-incumbent attitudes toward most Senatorsand representatives, it is not statistically associated with anti-incumbent attitudes toward one's own representatives. On theother hand, reliance on mainstream national media is negativelyassociated with anti-incumbent attitudes toward most Senatorsand representatives but positively associated with anti-incumbentattitudes toward one's own Senators and representatives, with theeffect stronger on most Senators and representatives. Mainstreamlocal news media are not statistically associated with anti-incumbent attitudes toward either most or one's own Senatorsand representatives.

8.7. Polarization and the media

H6 hypothesizes that heavy reliance on partisan sources, con-trolling for political and demographic measures, will be positivelyassociated with polarization, and it is supported by the data(b ¼ .13, p < .001). RQ6 asks whether polarization is observedamong thosewho rely onmainstream national or local sources, andthe data suggest that consumption of mainstream sources is notstatistically associated with polarization (n.s.), controlling foreverything else. Neither partisan media (RQ7) nor mainstreamnational or local media (RQ8) was statistically associated with anti-incumbent attitudes either toward Congress in general or their ownrepresentatives.

In sum, partisan media use is positively associated with partisanpolarization, but mainstream media use, at both national and locallevels, are not statistically associated with partisan polarization.

8.8. Political variables and demographics

When it comes to polarization, increase in political interest ispositively associated with polarization (b ¼ .12, p<.05). When itcomes to anti-incumbent attitudes, those who are more conser-vative (b ¼ �.23, p < .001) tend to have more anti-incumbent at-titudes toward their own representatives. On the other hand, men(b ¼ .11, p<.01), those who believe the economy is worse off(b ¼ .34, p<.001), and those who are more conservative (b ¼ .15,p<.001) are more likely to have more anti-incumbent attitudestoward most Senators and representatives. As the path model

Table 2OLS regression analyses of predictors of media use.

Predictors H3 H4a H4b

Partisan sources Mainstream national Mainstream local

Age �.11** �.04 .15**Gender (Male) .04 �.03 �.12**Race (White) �.03 �.05 �.10*Political Interest .33*** .21*** .13*Political Knowledge .11* .02 .01State of U.S. Econ. .02 �.25*** �.09Likelihood to Vote .08* .06 .20***Ideology .16*** �.14** �.05Education .14*** .16*** �.07Tea Party .13** e e

Non-Tea Party e .11* .12*Model F (10, 647) ¼ 32.88*** F (10, 537) ¼ 17.24*** F (10, 537) ¼ 10.37***R2 .34 .24 .16

Note: Cell entries are standardized coefficients. *p<.05, **p<.01, ***p<.001.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220 217

indicates (see Fig. 2), those who have more anti-incumbent atti-tudes toward most representatives also tend to have more anti-incumbent attitudes toward their own representatives (r ¼ .21,p < .001). Also, users of one media tested in this model are alsomore likely to use the other two media, which the strongest cor-relation between use of local mainstream and national mainstreammedia (r ¼ .44, p < .001), and the weakest between use of localmainstream media and partisan media (r ¼ .14, p < .001).

9. Discussion

The 2010 midterm election was marked by a high level of anti-incumbent attitudes, with almost two-thirds of the public arguingthat most Congressmen did not deserve to get re-elected. Manydidn't. Republicans gained control of the House of Representatives,gained five Senate seats and five governorships and made sub-stantial gains at the state levels. Numerous researchers haveexamined why people vote for or against an incumbent, but studieshave not directly examined how media are associated with anti-incumbent attitudes, and whether types of media (mainstreamvs. partisan sources) moderate anti-incumbent attitudes

Fig. 2. SEM model predicting anteced

differently.The Tea Party arose out of anger that the government was acting

fiscally irresponsibly and the belief that the role of the governmentshould be limited. Not surprisingly, then, Tea Party members heldstrong anti-incumbent attitudes, believing most Senators andCongressional members should not be re-elected. Democrats, onthe other hand, thought most representatives should be re-elected.

Americans have traditionally held Congress in low esteem buthave still believed their own representative deserves to be re-elected (Fenno, 1975). Indeed, members of the Tea Party andDemocrats were no more likely than nonmembers to believe theirown representative deserves re-election. Republicans were morelikely to believe their representative should be re-elected. This mayindeed show that while people may hold Congress in contempt,they are more likely to support their own representative because ofperceived constituent service (Parker and Davidson, 1979). Thismay also be an artifact of the Texas sample, as both Senators and 20of 32 members of the House of Representatives were Republicans.Republicans were more pleased with their representatives thanDemocrats because they were more likely to be represented by amember of their own party.

ents of anti-incumbent attitudes.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220218

Few studies have directly examined whether media in generaland partisan and mainstream media in particular, influence anti-incumbent attitudes. Studies are split on whether media providemore favorable coverage to incumbents (Fico et al., 2008), whetherthey favor the challengers (Graber, 2006; Fico and Cote, 2002) orwhether they provide balanced coverage.

Partisan news sources such as Fox News, talk radio, politicalbloggers and partisan news sites were thought to fuel the anti-incumbent attitudes in the 2010 elections because they allowpeople to only seek out information from sources that support theirpolitical views (Iyengar and Hahn, 2009; Lawrence et al., 2010) andthus lead to a more polarized society (Stroud, 2010).

This study found that use of partisan media is positively asso-ciated with polarized atmosphere in society and toward anti-incumbent attitudes. Specifically, those who use partisan mediaare more polarized than those who don't. Also, partisan mediausers have more anti-incumbent attitudes toward Congress ingeneral. Partisan media, particularly conservative sources such asFox News and talk radio, have been highly critical of theDemocratically-controlled Congress, and therefore use of partisanmedia is positively associated with the belief that most members ofCongress should not be re-elected.

Use of national mainstream media is negatively associated withanti-incumbent attitudes toward Congress in general. This supportsstudies that find that users of traditional media, particularlynewspapers, trust the government, and thus should be supportiveof those running it (Avery, 2009). National media rarely cover localraces unless they have national significance so it is surprising na-tional media coverage led to support for local representatives. Localmedia had no influence on attitudes toward national or local in-cumbents, which may reflect how little attention local media,particularly television, play to more local political races.

This study also found that those who thought the economy wasgetting worse were less supportive of most Senators and repre-sentatives, which reinforces studies that stress the importance ofthe state of the economy on anti-incumbent attitudes (Cintrin andLuks, 2001). When the economy worsens, public unhappiness withgovernment increases. A sour economy can create anti-incumbentattitudes because it brings citizens who rarely vote to the polls.Infrequent voters have less trust in government than frequentvoters (Hetherington, 1999) and therefore vote against theincumbent (Hansford and Gomez, 2010).

On the other hand, mainstream newsmedia at both the nationaland local levels are not statistically associated with polarization,although this is not unexpected. While negative coverage of gov-ernment by traditional media has been blamed for declines inconfidence in government and the officials that run them(Patterson, 1993), because one of the main tenets of mainstreamjournalism is to provide balanced coverage in political arguments(Wojcieszak and Rojas, 2011), it is not surprising that it is notassociated with polarization. While conservatives may perceive thatnational newspapers and broadcast news are slanted toward theleft (Pew Research, 2011), their more moderate coverage is notstatistically associated with anti-incumbent attitudes.

While scholars have debated whether the public has becomemore polarized, they are more united about the potential corrosiveeffects of polarization. Polarization leads to increased hatred to-ward the other political party and less tolerance toward opposi-tional views (Stroud, 2010, 2011). Polarization has been blamed forincreased reliance on partisan media for information and to the riseof negative political campaigns. While this study found thatpartisan media users did indeed become more polarized in theirviews, polarization is not statistically associated with anti-incumbent attitudes. This may reflect partisans’ conflicting viewstoward the workings of Congress. Studies indicate polarization can

lead to less confidence in Congress because the public believesmore can be achieved when Congressional members act in abipartisan manner to pass legislation (Harbridge and Malhotra,2011; Ramirez, 2009). On the other hand, strong partisans arealso more likely than independents and weak partisans to supportCongressional members who vote for policies they support(Grynaviski, 2010; Harbridge and Malhotra, 2011).

However, polarization had little influence on attitudes towardlocal politicians. This may be because attitudes toward local offi-cials are typically based on their actual service to the district (Ficoet al. 2008). Results also may reflect the Texas sample as the ma-jority of polarized individuals would have been Republicans or TeaParty supporters who would likely have been represented by con-servative Republicans.

This study indicated that in the 2010 campaign that anti-incumbent attitudes were high (more than 3 on a 4-point scale).Scholars have warned that anti-incumbent attitudes reflect a deepdistrust with government and that democracy is weakened whenits leaders do not have the support of a majority of its citizens(Miller, 1974). This study suggests that partisan media use is posi-tively associated with polarization and anti-incumbent attitudestowards Congress in general. Heavy use of partisan sources such asblogs, talk radio, cable news and political websites is positivelyassociated with selective exposure. Also, scholars have warnedabout the uncivil tone from partisan sources that routinely attackpoliticians and blame the government for society's problems(Cappella and Jamieson, 1997). Those citizens who rely heavily onpartisan sources, then, often receive a constant drumbeat ofnegative coverage against those in power, which reduces confi-dence and can fuel beliefs that those in office should be removedfrom power.

9.1. Limitations and suggestions for future studies

The sample of 2010 Texas voters had advantages that madeappropriate for studying the influence of media and different partysupport on anti-incumbent attitudes. Almost a third of respondents(32%) identified themselves as Tea Party members and more than 7in 10 (71%) disagreed or strongly disagreed that most representa-tives deserved re-election, demonstrating a strong anti-incumbentattitude. The study is also the official state poll, conducted with arandom sample with results weighted to represent U.S. censusfigures for Texas. But Texas is a heavily Republican state whoseviews may not be representative of the country as a whole. Futurestudies should expand this to a national population.

This study examined the association among different types ofmedia use (mainstream national, mainstream local and partisanmedia), polarization, and anti-incumbent attitudes. But for partisanmedia it is possible that liberal media may have a different influ-ence on polarization than conservative ones. Future studies maywant to explore the difference of specific networks (e.g. Fox, ESPN)and specific shows (e.g. The Rachel Maddow Show and The 0’ReillyFactor) on polarization and anti-incumbent attitudes. Also thisstudy assumed that mainstream media that strive for balancewould be perceived as political neutral, although previous studiesdemonstrate that the public do perceive network news, nationalnewspapers and CNN to have a political slant (Pew, 2011; Stroud,2011). Rather than have the researcher determine which mediawere partisan and which ones were neutral, future studies maychoose to ask respondents to judge the perceived ideological slantof the media.

From a theoretical perspective, we are aware that the causalrelationships among these variables are most likely reciprocal innature. For example, just as media exposure is likely to influencepolarization as the literature shows, polarization also influences

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220 219

subsequent media exposure, and just as political interests andpolitical knowledge are likely to influence media use, media usemay also contribute to political interests and political knowledge. Infact, we believe our model fitness suffers because of this. None-theless, we still think the proposed model is worth testing becausewe are specifically interested in examining how exposure todifferent types of media affect polarization and anti-incumbentsentiments, as this is something that the field has not yet tackledbut necessitates answers. Future studies are encouraged to incor-porate time-series analysis using longitudinal survey to parse outthe full-on reciprocal relationships among these key variables, butthis is beyond the scope of this cross-sectional survey study.Nonetheless, we have tried, using structural equation modeling, toestimate relationships among these variables through novel theo-retical advancement.

There are no “perfect” research designs. For example, while ex-periments are great for causal inferences, they often lack generaliz-ability. On the other hand, while surveys are great forgeneralizability, they lack the ability to make causal claims, partic-ularly ones that are cross sectional. In survey research, with the in-clusion of control variables, regression analyses are often used tomake pseudo-causal claims. However, one of the shortcomings ofregression analyses is that it does not allow for a holistic under-standing of how different variables in a theoretical model affect eachotherdthat is, regressions may be able to demonstrate the associa-tion between A- > B and C- > D, but it does not account for thesystematic relationship between the two links or among the fourvariables (i.e., does the association between A- > B precede C- > D orvice versa? Does A indirectly affect D? If A indirectly affects D, doesthis moderate the association between A and B in some way?).

In other words, while existing research has been able todemonstrate that media use affects polarization, and polarizationaffect anti-incumbent attitudes, the literature does not offer a ho-listic view of the relationship between the two aforementionedcausal linkages. To advance empirical understanding of this puzzle,this study adopted structural equation modeling (SEM), a family ofstatistical methods designed to test theoretical models. In this case,SEM was used to examine the mediating effects of polarization onmedia use and anti-incumbent attitudes. While the use of surveyhinders this study's ability to make causal claims, its findingssuggest some ways in which to conduct experimental studies tofurther validate the causal relationship among media use, polari-zation and anti-incumbent attitudes, thus serves as one buildingblock for future studies in this area of research to flourish.

Finally, the authors were attracted to this subject because of theanti-incumbent sentiment in the electorate. Future studies shouldexamine elections where there is not such a strong kick-out-the-bums out mentality.

Acknowledgement

The lead author would like to thank I Jerrold Johnson for hislifetime of support, without which this and his other researchwould not be possible.

References

Abramowitz, A.I., 2010. Transformation and polarization: the 2008 presidentialelection and the new American electorate. Elect. Stud. 29, 594e603.

Aldrich, J.H., Bishop, B.H., Hatch, R.S., Hillygus, D.S., Rohde, D.W., 2013. Blame, re-sponsibility, and the tea party in the 2010 midterm elections. Polit. Behav. 35,1e21.

Arceneaux, K., Nicholson, S.P., 2012. Who wants to have a Tea Party? The who, what,and why of the tea party movement. PS Polit. Sci. Polit. 45 (4), 700e710.

Avery, J.M., 2009. Videomalaise or virtuous circle? The influence of the news mediaon political trust. Int. J. Press/Polit. 14, 410e433.

Baldassarri, D., Bearman, P., 2007. Dynamics of political polarization. Am. Sociol.

Rev. 72, 784e811.Barfumi, J., Shapiro, R.Y., 2009. A new partisan voter. J. Polit. 71, 1e24.Baum, M.A., Groeling, T., 2008. New media and the polarization of American po-

litical discourse. Polit. Commun. 25, 345e365.Byers, D., 2012. The Most Accurate Polls of 2012. Politico. Retrieved from. http://

www.politico.com/blogs/media/2012/11/the-most-accurate-polls-of-148876.html.

Canes-Wrone, B., Brady, D.W., Cogan, J.F., 2002. Out of step, out of office: electoralaccountability and house members' voting. Am. Polit. Sci. Rev. 96 (1), 127e140.

Cappella, J., Jamieson, K., 1997. Spiral of Cynicism: the Press and the Public Good.Oxford University Press, New York, NY.

Carmines, E.G., Ensley, M.J., Wagner, M.W., 2012. Who fits the left-right divide?Partisan polarization in the American electorate. Am. Behav. Sci. 56 (12),1631e1653.

Carson, J.L., Pettigrew, S., 2013. Strategic politicians, partisan roll calls, and the TeaParty: evaluating the 2010 midterm elections. Elect. Stud. 32, 26e36.

Citrin, J., Luks, S., 2001. Political trust revisited: D�ej�a vu all over again. In:Hibbing, J.R., Theiss-Morse, E. (Eds.), What is it about Government that Amer-icans Dislike? Cambridge University Press, Cambridge.

Cook, R., 2010. The 2010 Primaries: Some Parting Thoughts. Retrieved from. http://blogs.wsj.com/capitaljournal/2010/09/23/the-2010-primaries-some-parting-thoughts.

D'Alesso, D., 2012. Media bias in Presidential Election Coverage 1948e2008. Lex-ington Books, Lanham, MD.

Ensley, M.J., 2012. Incumbent positioning, ideological heterogeneity and mobiliza-tion in U.S. House elections. Public Choice 151, 43e61.

Fenno Jr., R.F., 1975. If, as Ralph Nader says, Congress is “The broken branch,” howcome we love our Congressmen so much? (pp. 277e287. In: Ornstein, N.J. (Ed.),Congress in Change: Evolution and Reform. Praeger, New York.

Fico, F., Cote, W., 2002. Partisan and structural balance of election stories on the1988 governor's race in Michigan. Mass Commun. Soc. 5, 165e182.

Fico, F., Zeldes, G.A., Carpenter, S., Diddi, A., 2008. Broadcast and cable networknews coverage of the 2004 presidential election: an assessment of partisan andstructural imbalance. Mass Commun. Soc. 11 (3), 319e339.

Frederico, C.M., Hunt, C.V., 2013. Political information, political involvement, andreliance on ideology in political evaluation. Polit. Behav. 35 (1), 89e112.

Garrett, R.K., 2009. Echo chambers online? Politically motivated selective exposureamong Internet news users. J. Comput.-Mediat. Commun. 14, 265e285.

Graber, D.A., 2006. Mass Media and American Politics. CQ Press, Washington, DC.Grynaviski, J.D., 2010. Partisan Bonds. Cambridge University Press, New York.Hansford, T.G., Gomez, B.T., 2010. Estimating the electoral effects of voter turnout.

Am. Political Sci. Rev. 104, 268e288.Harbridge, L., Malhotra, N., 2011. Electoral incentives and partisan conflict in

Congress: evidence from survey experiments. Am. J. Polit. Sci. 55, 494e510.Hetherington, M.J., 1999. The effect of political trust on the presidential vote,

1968e96. Am. Polit. Sci. Rev. 93, 311e326.Hibbing, J.R., Theiss-Morse, E., 1995. Congress as Public Enemy: Public Attitudes

toward American Political Institutions. Cambridge University Press, New York.Iacobucci, D., Saldanha, N., Deng, X., 2007. A meditation on mediation: evidence

that structural equations models perform better than regressions. J. Consum.Psychol. 17 (2), 139e153.

Iyengar, S., Hahn, K.S., 2009. Red media, blue media: evidence of ideologicalselectivity in media use. J. Commun. 59, 19e39.

Johnson, T.J., Kaye, B.K., 2012, August. Blinded by the spite? Path model of politicalattitudes, selectivity, and social media. In: Paper Presented at the Associationfor Education in Journalism and Mass Communication Annual Conference,Chicago, IL.

Johnson, T.J., Kaye, B.K., 2013. Putting out fire with gasoline: testing the Gamsonhypothesis on media reliance and political activity. J. Broadcast. Electron. Media57 (4), 456e481.

Jones, D.A., 2002. The polarizing effect of new media messages. Int. J. Public Opin.Res. 14, 158e174.

Kaye, B.K., Johnson, T.J., 2014. The shot heard around the World Wide Web: whoheard what where about Osama bin Laden's death. J. Comput. Mediat. Commun.19 (3), 643e662.

Kerr, J.C., 2010. State Elections: Republicans Gain Control of Key State Legislatures.Retrieved from. http://www.huffingtonpost.com/2010/11/03/state-elections.

Knobloch-Westerwick, S., 2012. Selective exposure and reinforcement of attitudesand partisanship before a presidential election. J. Commun. 62, 628e642.

Lawrence, E., Sides, J., Farrell, H., 2010. Self-segregation or deliberation? Blogreadership, participation, and polarization in American politics. Perspect. Polit.8 (1), 141e157.

Levendusky, M., 2013. Partisan media exposure and attitudes toward the opposi-tion. Polit. Commun. 30 (4), 565e581.

Lowry, D.T., Shidler, J.A., 1998. The sound bites, the biters, and the bitten: a two-campaign test of the anti-incumbent bias hypothesis in network TV news.Journal. Mass Commun. Q. 75 (4), 719e729.

Matthews, D., 2012. Which Pollsters to Trust, in One Chart. Washington PostWonkBlog Retrieved from. http://www.washingtonpost.com/blogs/wonkblog/wp/2012/11/07/which-pollsters-to-trust-in-one-chart/.

Miller, A.H., 1974. Political issues and trust in government, 1964e1970. Am. Polit.Sci. Rev. 6, 951e972.

Moon, W., 2015. Positional effects of partisan attachments on candidate positiontaking. Elect. Stud. 37, 99e108.

Murray, M., 2010. ‘Hurricane’ Ends Democrats' Control of House. Retrieved from.

T.J. Johnson, A.M. Lee / Electoral Studies 40 (2015) 210e220220

http://www.msnbc.msn.com/id/39982780/ns/politics-decision_2010/.Mutz, D.C., Martin, P.S., 2001. Facilitating communication across lines of political

difference: the role of mass media. Am. Polit. Sci. Rev. 95, 97e114.Parker, G.R., Davidson, R.H., 1979. Why do Americans love their Congressmen so

much more than their Congress? Legis. Stud. Q. 4, 43e61.Patterson, T.E., 1993. Out of Order. Knopf, New York.Pew Research Center, 2010a. Biennial Media Consumption Survey 2010: Final

Topline. Retrieved from. http://www.people-press.org/files/legacy-questionnaires/652.pdf.

Pew Research Center, 2010b. Distrust, Discontent, Anger and Partisan Rancor:People and Their Government. Retrieved from. http://www.people-press.org/2010/04/18/distrust-discontent-anger-and-partisan-rancor/.

Pew Research Center, 2011. Beyond Red vs. Blue: Political Typology. Retrieved from.http://people-press.org/files/legacy-pdf/Beyond-Red-vs-Blue-The-Political-Typology.pdf.

Pew Research Center, 2014. Political Polarization in the American Public. Retrievedfrom. http://www.people-press.org/2014/06/12/political-polarization-in-the-american-public/.

Prior, M., 2006. The incumbent in the living room e the rise of television and theincumbency advantage in U.S. house elections. J. Polit. 68, 657e673.

Prior, M., 2013. Media and polarization. Annu. Rev. Polit. Sci. 16, 101e127.Ramirez, M.D., 2009. The dynamics of partisan conflict on Congressional approval.

Am. J. Polit. Sci. 53 (3), 681e694.Reynolds, M., Stewart, C., 1990. Television markets and U.S. senate elections. Legis.

Stud. Q. 15, 495e523.Saad, L., 2010. Voters Issue Strong Rebuke of Incumbents in Congress. Retrieved

from. http://www.gallup.com/poll/127241/voters-issue-strong-rebuke-Incumbents-congress.aspx.

Smith, S.S., 2007. Party Influence in Congress. Cambridge University Press, NewYork.

Sobieraj, S., Berry, J.M., 2011. From incivility to outrage: political discourse in blogs,talk radio, and cable news. Polit. Commun. 28, 19e41.

Steinhauser, P., 2013. CNN Poll: 75% Say Most Republicans in Congress Don'tDeserve Relection. Retrieved from. http://www.cnn.com/2013/10/21/politics/cnn-poll-shutdown-re-election/.

Stonecash, J.M., Brewer, M.D., Mariani, M.D., 2002. Diverging Parties: Realignment,Social Change, and Political Polarization. Westview Press, Boulder, CO.

Stroud, N.J., 2010. Polarization and partisan selective exposure. J. Commun. 60,556e576.

Stroud, N.J., 2011. Niche News: the Politics of News Choice. Oxford University Press,New York.

Wojcieszak, M., Rojas, H., 2011. Correlates of party, ideology and issue based ex-tremity in an era of egocentric publics. Int. J. Press/Polit. 16, 488e507.