Television News Repertoires,Exposure Diversity,and Voting Behavior in the 2016 U.S. Election

Abstract

This study explores the television news repertoires and voting behaviors of American citizens in the 2016 U.S. election. The results reveal notably different repertoires, some defined by ideologically driven selective exposure, others defined by political interest selective exposure (preference/disinterest in the news), and others that are seemingly neutral. In turn, both total news exposure and exposure diversity positively predict voter turnout, with exposure diversity demonstrating the stronger effect. Moreover, there was a clear relationship between exposure to partisan news media and voting in that respective party’s primary election.

Keywords

exposure diversity media repertoires news selective exposure set-top box data (STB)big data voting behavior 2016 U.S. election

The contemporary news environment is characterized by an abundance of content options made available across a variety of platforms. The on-demand, anytime-anywhere philosophy embraced by media providers, along with the recognition of the economic viability of specialized niche content (Anderson, 2006), has led to the rapid proliferation of content options competing for one’s attention at any given time (Napoli, 2011). Yet the amount of attention free to devote to media consumption is limited. Thus, despite the abundance of content, individuals are faced with a scarcity of attention—relative to the amount of content available. Some describe this as an “attention economy,” where attention is a scarce resource that audiences work to effectively allocate (Davenport & Beck, 2001; Lanham, 2006). At the same time, the high-choice news environment allows media users to find the content that satisfies their preferences more easily and establish the list of media outlets they purposefully use more efficiently. Together, content abundance and attention scarcity combine to offer unique opportunities and constraints for media consumption.

Much has been made about the potential for these environmental features to facilitate the customization of one’s personal media environment (e.g., Gitlin, 1998; Katz, 1996; Napoli, 2011; Sunstein, 2007; Taneja, Webster, Malthouse, & Ksiazek, 2012). Regarding news and information, some worry that these individual behaviors will scale to aggregate-level patterns like audience fragmentation (Tewksbury, 2005; Webster, 2005) and audience polarization (Ksiazek, Malthouse, & Webster, 2010; Prior, 2007; Stroud, 2010). A recent report from Pew Research Center (2014) showed that there is little overlap in news media usage between liberals and conservatives in the United States, legitimizing concerns about discrepancies in the way people learn about politics and government.

The analysis of personalized media environments can be traced to research on TV channel repertoires (e.g., Heeter & Greenberg, 1985; Yuan & Webster, 2006), and more recently media repertoires (e.g., Dvir-Gvirsman, Tsfati, & Menchen-Trevino, 2016; Edgerly, 2015; Hasebrink & Popp, 2006; Kim, 2016; H. Lee & Yang, 2014; Schrøder, 2015; Taneja et al., 2012; Weeks, Ksiazek, & Holbert, 2016; Wolfsfeld, Yarchi, & Samuel-Azran, 2016; Yuan, 2011), that aims to explore the subsets of content and/or platforms that individuals select among abundant options. However, much of the extant research focuses on which platforms (e.g., TV, Internet, etc.) individuals use, and on the average size of an individual’s repertoire in relation to the total number of available options. Analyzing the actual composition of television news repertoires at the channel/program level would offer insight into the customized information environments that individuals are creating.

The ability to personalize one’s diet of news and information can lead to stark differences in the range of issues and viewpoints available to that individual. For instance, in the realm of selective exposure theory, some research finds selective exposure based on political ideology, where news consumption patterns appear to be defined along partisan lines (e.g., Hollander, 2008; Iyengar & Hahn, 2009; Morris, 2005; Stroud, 2008, 2010, 2011). Others refute this, instead finding a degree of cross-cutting exposure to different partisan news outlets (e.g., Garrett, 2009a, 2009b; Garrett, Carnahan, & Lynch, 2013; Garrett & Stroud, 2014; Gentzkow & Shapiro, 2011; Holbert, Hmielowski, & Weeks, 2012; Ksiazek, 2016; Prior, 2013; Webster & Ksiazek, 2012; Weeks et al., 2016). Together, the current research suggests that individuals do engage in selective exposure to news, but do not actively avoid counter-attitudinal information.

The current study builds on the extant research by employing a unique research design bringing together multiple data sets to explore exposure diversity in customized television news environments and its relationship to voting behaviors, and tests the impact of selective news exposure on voting behaviors in the 2016 U.S. election. Although much has been made of the potential political outcomes predicted by increasing personalization in news exposure, we know less about the relationship between these customized news repertoires and actual voting behaviors.

To analyze news repertoires at the outlet level, this study uses television set-top box data collected from 165,036 cable subscribing households in six states. To our knowledge, this is one of the first studies that uses set-top box data to understand news repertoires and exposure diversity. Set-top boxes monitor second-by-second TV viewing behaviors, thus providing more accurate measures of TV exposure than many studies relying on surveys or diaries. More and more, news is being consumed in digital environments where overt user behaviors can be observed in unobtrusive ways producing big data sets. In addition to the substantive insights on news repertoires, this article offers a concrete example and more general discussion on the strengths and weaknesses of using such data sets for media effects research. This study specifically focuses on TV given its role as one of the primary sources of news and information (Gottfried, Barthel, Shearer, & Mitchell, 2016; Papathanassopoulos et al., 2013). As Graber (2001) noted, TV excels in delivering political knowledge and promoting learning about politics and current issues. It also remains the primary source of news for the American public—even after the rise of the Internet and social media—according to the most recent Pew report on news consumers (Pew Research Center, 2016).

After providing a broad picture of television news repertoires and exposure diversity, this study offers an analysis of the relationship between those exposure patterns and voting behaviors in the 2016 U.S. election. By matching households in the set-top box data with voting records in six states, the analysis combines rich and precise data sources to understand what people are actually watching on TV and doing at the voting booths, rather than relying on self-reports and recall. Given that TV viewing is largely overestimated in surveys compared with meters (Prior, 2009), this study explores news repertoires and tests the relationship between news exposure and voting behaviors using more conservative and accurate measures of news consumption.

Literature Review

News Repertoires

In an environment of content abundance and attention scarcity, individuals cannot possibly consume all that is available, nor do they necessarily desire to do so. Instead, users construct media repertoires—subsets of available content to which they are regularly exposed. Media repertoires are part of a larger family of coping strategies used by audiences to seek out preferred and filter unwanted content to navigate an increasingly complex media environment (Taneja et al., 2012).

Much of the initial research on channel and media repertoires focused on predicting the absolute size of repertoires (e.g., Ferguson, 1992; Ferguson & Perse, 1993; Heeter & Greenberg, 1985; Neuendorf, Atkin, & Jeffres, 2001; Reagan, 1996; Yuan & Webster, 2006). In 2016, the Nielsen Company reported that American TV viewers only use an average of 19.8 channels, despite the average household receiving 205.9 channels (Nielsen, 2016). In fact, this represents one of the lowest proportions (9.6%) of total channels viewed since Nielsen began reporting these figures. This coping strategy of constructing TV channel repertoires is consistent with research by Ji, Ha, and Sypher (2014), who found a significant relationship between exposure to traditional news media (e.g., TV) and perceived information overload.

Building on this, studies began calling for an understanding of the composition of those repertoires, not just the relative size (Ferguson & Perse, 1993; Yuan & Webster, 2006). Yet much research to date on repertoire composition focuses on preferences for news across platforms—for example, TV versus Internet—or makes broad distinctions between traditional and digital media repertoires (Hasebrink & Popp, 2006; Kim, 2016; H. Lee & Yang, 2014; Schrøder, 2015; Wolfsfeld et al., 2016; Yuan, 2011), rather than varying exposure to news content across specific outlets (e.g., TV channels).

A few recent studies have begun to dig deeper into the composition of news repertoires at the outlet level, and the present study contributes to this new research trajectory. For instance, Edgerly (2015) analyzed survey data to explore news repertoires across TV, online, print, and radio outlets. She found six distinct repertoires, some demonstrating clear ideological leanings (i.e., conservative or liberal) and others embodying preferences for news exposure through particular media (e.g., TV, online, or print). She argues that repertoires can be characterized by both medium and content attributes. Dvir-Gvirsman et al. (2016) tracked Israeli Internet users and found that some users prefer a mix of partisan and nonpartisan media, some prefer partisan media from both sides, others avoid partisan media altogether, and very few prefer ideologically driven selective exposure to partisan news outlets. Weeks et al. (2016) compared news repertoires of Democrats, Republicans, and the general American public. The news diets of these groups, consisting of TV channels, political talk radio, and print and online news outlets, revealed little variation and a good deal of cross-cutting news exposure, challenging the assumption of audience polarization based on party identification.

The current study contributes to this literature by focusing on television news repertoires and the role of exposure diversity in influencing voting behavior across different types of news consumers. The analysis centers on the composition of the repertoires, not the absolute size, and also offers more granular measures of media exposure by focusing on outlets rather than platforms. In addition, the focus on news situates this work within a growing body of research on selective exposure to understand the diversity of news exposure patterns among citizens and the relationship between news exposure and voting behavior.

Selective Exposure to News Sources

Scholars have argued that we should revive the concept of “selective exposure” in the current media environment because an abundance of offerings has given media users the freedom to easily find and consume media content that matches their preferences (Bennett & Iyengar, 2008; Prior, 2007). Regarding patterns of news consumption, two types of selective exposure—selective exposure driven by political interest and political ideology—have become the foci of academic investigation (Skovsgaard, Shehata, & Strömbäck, 2016). Selective exposure based on political interest explains the division between news seekers and avoiders (Prior, 2007; Strömbäck, Djerf-Pierre, & Shehata, 2013). Those who are interested in politics can seek out a wide range of news content, whereas those who are not may choose to opt out from news offerings entirely. Overall, empirical evidence suggests the existence of both newshounds and avoiders, although not all previous research could link people’s level of political interest with their level of news consumption (Kim & Webster, 2012; Ksiazek et al., 2010; Strömbäck et al., 2013).

The second type of selective exposure, guided by one’s political ideology, is concerned with people’s tendency to choose news sources that are consistent with their political predispositions and/or avoid those that contradict their political beliefs (Bennett & Iyengar, 2008; Mutz & Young, 2011). Although previous studies on selective exposure have produced mixed results (Barlett, Drew, Fahle, & Watts, 1974; Chaffee & Miyo, 1983; Chaffee, Saphir, Graf, Sandvig, & Hahn, 2001), there is a growing body of research that has found a relationship between people’s political ideology and partisan selective exposure to news media in the current high-choice media environment (Graf & Aday, 2008; Hollander, 2008; Iyengar & Hahn, 2009; Prior, 2007; Stroud, 2008).

Although most research on partisan selective exposure has focused exclusively on active seeking of media outlets sharing similar political predispositions, an increasing number of studies have sought to determine whether news consumers also actively avoid news media with dissonant viewpoints—in essence accounting for their broader repertoire of news exposure. As Garrett (2013) pointed out, the image of “echo chambers” where people retreat into enclaves of like-minded viewers is an exaggeration from the original version that Jamieson and Cappella (2008) elaborated in their book. Subsequent empirical studies have shown that people in fact consume a more diverse and balanced diet of news media outlets (Garrett, 2009a, 2009b; Garrett et al., 2013; Garrett & Stroud, 2014; Gentzkow & Shapiro, 2011; Holbert et al., 2012; Ksiazek, 2016; Prior, 2013; Trilling & Schoenbach, 2013; Webster & Ksiazek, 2012; Weeks et al., 2016) and pay some level of attention to news that contradicts their beliefs, suggesting that concerns about audience polarization may be overstated.

News Exposure and Voting

Engaging in selective exposure based on political interest and ideology can have real political and civic outcomes. In terms of political interest, research demonstrates a positive relationship between news exposure and both political and civic participation (Kanervo, Zhang, & Sawyer, 2005). As for ideology, selective exposure to partisan news media is related to party affiliation. Although existing studies tend to treat the relationship as circular, many use party affiliation or ideology to predict patterns of news exposure. This study explores the other relational direction, predicting general voter turnout and primary voting based on past news exposure patterns.

Despite the positive association between news exposure and political participation, the relationship between exposure diversity and political participation remains unclear. De Vreese and Boomgaarden (2006) found that exposure to news outlets with political content positively influenced political learning and voting behavior, whereas exposure to news outlets with less political content has a weaker or no effect on political knowledge gains and voting behavior, suggesting the importance of exposure diversity to more “hard” news. Helberger and Wojcieszak (2018), in their review of exposure diversity, pointed out that some studies have shown the negative effect of encountering counter-attitudinal perspectives in news media on political engagement (Dilliplane, 2011; Nir & Druckman, 2008). These inconsistent findings motivate this study’s analysis of the relationship between individuals’ exposure to diverse news sources and their voting behavior beyond the much studied relationship between the amount of news exposure (i.e., total news exposure) and political participation.

Hypotheses and Research Questions

The purpose of this study is to (a) explore television news repertoires focusing on the channel composition and viewer characteristics of each repertoire, (b) examine the influence of exposure diversity on voting behavior, and (c) test the impact of total news exposure and selective exposure on voting. Thus, this study begins by exploring the television news repertoires that characterize contemporary news consumers. In recent decades, communication researchers have demonstrated a renewed interest in selective exposure theory, particularly as it pertains to concerns about patterns of fragmentation and polarization among audiences. Building on this work, the present analysis explores repertoires that exist based on patterns of television news exposure. The first set of research questions addresses the lack of focus on repertoire composition at the outlet level in past empirical research and integrates this work with selective exposure theory to explore exposure diversity within and across repertoires. Consistent with recent research, the first research question asks whether both ideological selective exposure and more diverse exposure patterns exist among the viewing public. The next two research questions explore repertoire composition, both in terms of the outlets that make up each (RQ2), and the people who inhabit these information environments (RQ3).

RQ1: Do both partisan and more diverse, nonpartisan news repertoires exist?

RQ2: Which news outlets are included in each repertoire?

RQ3: What are the characteristics of individuals in each repertoire?

Next, the analysis explores the relationship between news exposure and voting. Given the unique access to behavioral data for both news exposure and voting in this study, the comprehensive approach outlined below teases out the relationships among different types of news exposure (total news exposure, exposure diversity, news repertoires, partisan news) and different types of voting behaviors (general; primary). Although the relationship between news exposure and political participation is often considered a circular relationship, or “virtuous circle” (Norris, 2000), our viewing data were collected prior to the 2016 U.S. election, allowing us to test whether exposure to particular news repertoires predicts voting behaviors. RQ4 asks if the level of diversity in one’s news diet predicts voter turnout and primary voting, addressing the need to explore whether exposure diversity encourages or depresses voting behaviors. RQ5 asks about variations in turnout across news repertoires, offering a more nuanced picture of how each news repertoire is associated with voter turnout.

RQ4: Does exposure diversity predict voter turnout and primary voting?

RQ5: Do voter turnout and primary voting vary across the news repertoires?

Finally, we propose two hypotheses regarding the relationship between news exposure and voting. H1 predicts confirmation of the general positive relationship between news exposure and voting behavior found in extant research. Although our primary analytical contribution to research on the exposure-voting relationship involves exposure diversity, H1 captures a baseline relationship between total news exposure and voting behaviors for comparative purposes. Moreover, while the relationship between news exposure and voting is well established, treating general and primary turnout as different outcomes offers an additional contribution. Although general election turnout is a basic measure of political participation, primary voting offers an additional, inherent indicator of party affiliation. In addition, the disparity between primary and general election turnout in the 2016 election (and most, if not all, modern U.S. elections) further highlights that primary voters are a subset of general election voters. The final hypothesis integrates this study’s focus on selective exposure, exposure diversity, and voting; H2 predicts a positive relationship between partisan news and voting in the Republican and Democratic primaries, where those who have greater exposure to “red” news outlets should be more likely to vote in the Republican primary and those exposed to “blue” outlets should vote in the Democratic one.

H1: Total news exposure predicts voter turnout and primary voting.

H2: Exposure to “red” and “blue” news outlets positively predicts voting in the Republican and Democratic primaries, respectively.

Method

Data

This study uses a unique data set from a U.S. cable TV multisystem operator (MSO) recording all viewing for 165,036 subscribing households that could be matched with voting and demographic data over the 9-month period from April to December 2015, which allows us to compute the length of time that each household “tuned into” each of 277,000 unique program episodes. The MSO provided a taxonomy identifying the genre of each program. This study focuses on programs in the news genre and grouped them into 1,300 distinct news titles. For example, “ABC World News” is a title, and the show that aired on December 1, 2015, is an episode. The majority of the titles were local news (e.g., “WDBJ7 News at 5:00”), which were further grouped into one local news category. This MSO serves small cities, rural areas, and some suburbs in several U.S. regions: South, Appalachia, Great Plains, and the Mountain West.

The cable TV company matched viewing data with voting data from the 2016 election for households in six states giving (a) the number of people in the household who voted in the general election (i.e., voter turnout) and (b) if household members voted in the Republican or Democratic primary, a strong indicator of party affiliation. In total, 165,036 households could be matched, with 127,637 (77%) voting in the general election. There were 77,390 voting in primaries, with 21,222 (27%) voting Democratic. We do not have personally identifiable information for any households. Households are identified only by an eight-digit anonymous ID.

We should note that set-top box data inherently measure cable TV viewing at the household level, whereas the concept of channel repertoire is pertinent to individual-level media usage patterns. We argue that unlike watching TV on a phone or tablet, watching cable TV is usually a communal activity. Despite the decline of family viewing in the current environment, scholars have argued that family viewing remains a reality, just to a lesser degree or frequency (F. Lee, 2010). If we think about the practice of cable TV viewing, it is done in an open setting; speakers are used rather than headphones, and thus other members of the household will be exposed to the news program as they move about the household. Moreover, while one could view with the door closed, the fact that a certain type of news is being viewed in the household means that the events and opinions covered on that news will be discussed in the household and influence household members’ attitudes on social issues, as shown in previous research on the indirect effect of interpersonal communication between adolescents and parents (Boyd, Zaff, Phelps, Weiner, & Lerner, 2011). In this study, we matched set-top box data with voting data at the household level. We address how this analytic approach may have influenced our results in the discussion. We also note that our universe is restricted to households with cable TV, since channels such as Fox and MSNBC are only available to cable subscribers.

Measures

To test the first set of research questions, factor and cluster analysis were used to identify news repertoires. All 165,036 households that watched TV news were included in the analysis and the log time spent viewing 62 different news programs were factored. The distributions of viewing time were highly skewed, and the logarithm was used to symmetrize the distribution and stabilize the variance. News programs came from the main cable news networks (Fox, CNN, and MSNBC), the three major broadcast networks that offer national news (ABC, NBC, CBS), public TV (PBS and BBC), and local stations, which were aggregated into one local TV variable.

The analysis began with an exploratory factor analysis with a promax rotation.¹ The Kaiser criterion suggested 10 factors. The first three factors consisted of Fox, CNN, and MSNBC programs, respectively, all with large loadings and almost no substantial cross-loadings. The only exceptions were CNN’s HLN Weekend Express and the Robin Meade Show, which formed a separate factor. Likewise, there were three factors for local and network (ABC, CBS, and NBC) programs, although there were cross-loadings between them. There were two other network factors, one consisting of CBS’s 48 Hours and ABC’s 20/20, and the other had CBS’s Morning News, Up-to-the-Minute News, and Overnight News, although both had cross-loadings with the other network factors. The last factor was made up of BBC and PBS, which had no substantial cross-loadings. A scree plot suggested a more parsimonious factor solution, indicating five factors. The first three factors were the same as the 10-factor solution (Fox, CNN, MSNBC), thus there is strong empirical support for these three factors, and they have face validity. The story is less clear-cut for all broadcast and public news programs. Although the five-factor solution suggests a single broadcast factor (including network and local programs) and a public TV factor, the 10-factor solution seems to break these out by individual networks with some cross-loading of network programs and local news programs. As we do not theorize about the different networks separately, and coefficient alpha equals .894 when we group all broadcast channels together, there is empirical justification for a single broadcast factor. Coefficient alpha was computed for each factor, which decreases when any item is dropped. See Table 1 for the five factors that will be used in the subsequent cluster analysis.

Table 1.

Results of a Factor Analysis on 62 News Programs.

Factor	Items (news programs) included in each factor
Fox (α = .975)	Happening Now, Outnumbered, Neil Cavuto, Gretchen Carlson, Shepard Smith, Greta Van Susteren, Bret Baier, America’s Newsroom, Hannity, Fox & Friends, Red Eye, The Kelly File, America’s News Headquarters, Media Buzz, Journal Editorial Report, and Chris Wallace. When these variables were factored separately there was only one eigenvalue greater than one (73% of variance), suggesting unidimensionality.
CNN (α = .941)	Wolf Blitzer, Ashleigh Banfield, Jake Tapper, Berman Bolduan, Erin Burnett, CNN, New Day, Anderson Cooper, Michael Smerconish, and Fareed Zakaria. A factor analysis of only these variables also had a single eigenvector greater than one (67% of variance). Note that Robin Meade and HLN Weekend Express are omitted.
MSNBC (α = .919)	Hardball, Chris Hayes, Andrea Mitchell, Last Word, Rundown, Live, Rachel Maddow, Alex Witt, Harris Perry, Up with Steve Kornacki. There was a single eigenvalue greater than one (59% of variance).
Broadcast (α = .894)	CBS Morning Sunday, Face the Nation, Morning Saturday, This Morning, Evening News, 60 Minutes, 48 Hours, Morning News, Up-to-the-Minute News, Overnight; ABC Nightline, World News, America This Morning, Stephanopoulos, and 20/20; NBC Today, Nightly News, Meet the Press, and Dateline; and all local news programs. The first factor of these variables accounts for 34% of the variance between the variable.
Public (α = .59)	PBS and BBC

Note. Detailed tables giving program-level factor loadings and Cronbach’s α analyses are available in the supplemental material.

We now document how we computed factor scores. For each household, the analysis computed the total minutes watched from each of the factors. The means, standard deviations, and correlations of the log watch times are shown in Table 2. Broadcast news has the largest mean, followed by Fox and CNN. Broadcast also has a relatively small standard deviation, while the cable networks (especially Fox) have larger standard deviations, indicating more variation in viewing time across households. The largest correlation is between CNN and MSNBC (r = .502), and then Fox and CNN (r = .416). All correlations are positive, including MSNBC and Fox (r = .249).

Table 2.

Correlations and Descriptive Statistics for Time Spent Viewing Across the News Consumption Factors (N = 165,036).

	Fox	CNN	MSNBC	Broadcast	Public
CNN	.416	—
MSNBC	.249	.502	—
Broadcast	.284	.364	.238	—
Public	.194	.299	.252	.211	—
Mean (log)	5.64	4.39	2.38	8.59	1.74
SD (log)	3.21	2.96	2.69	1.66	2.16

Note. All correlations significant at p < .0001 level.

To analyze voting behavior, the analysis used measures of voter turnout (voting in the general election) and primary voting (Democratic vs. Republican). To test H1 and RQ4 a logistic regression was fitted using two other news exposure measures. For H1, log total time spent watching news was used. For RQ4, entropy was used,² which measures exposure diversity and is defined as follows. For each household, let p_j equal the proportion of time spent watching a factor j = 1, . . ., 5 (1 = Fox, 2 = CNN, 3 = MSNBC, 4 = broadcast, 5 = public), so the p_j values sum to 1. Then entropy is defined as $- \sum_{j = 1}^{5} p_{j} \log (p_{j})$ . Larger values of entropy indicate more exposure diversity.

Demographic measures are used in the repertoire profiles and as control variables in the regression models. Age is measured in years. Income is on a 9-point scale, where 1 = <15,000, 2 = 15,000 to 20,000, 3 = 20,000 to 30,000, 4 = 40,000 to 50,000, 5 = 50,000 to 75,000, 6 = 50,000 to 75,000, 7 = 75,000 to 100,000, 8 = 100,000 to 125,000, and 9 = 125,000+. Household composition is captured by the percentage of households that are married and have at least one child, respectively. Education of the head of household is given in four categories: high school, college, graduate school, and vocation/technical. College has the same directional results as graduate school and vocational is a small category with few households. Finally, past voting behaviors from 2012 are used as an important baseline control in the regression models, allowing the analysis to isolate the impact of news viewing in 2015 on voting behavior in 2016. The mean age is 55.2 years and its standard deviation is 17.5 years. In our sample, 57% are married and 33% have children. For education, 66% have a high school degree, 21% have college a college degree, and 12% have a graduate degree. The income variable has nine values with a mean of 4.73 and SD = 2.53.

Analysis

To test the first three research questions, K-means cluster analysis was applied to the five news factor variables and all 165,036 households with positive viewing time to identify news repertoires. The analysis estimated the four to nine cluster solutions without standardizing the data. As the units of all variables were commensurate, standardizing was not necessary, and could obscure structure in the data by inflating the variation (and thus importance) of variables with smaller standard deviations (public and broadcast) and reducing the variance of those with larger standard deviations (Fox and CNN). The variables were mean-centered to improve the interpretability of the solution, where the value 0 indicated an average amount of time, positive values revealed above-average viewing times, and negative values indicated below-average viewing. Centering did not affect the variation in the data, and therefore did not affect the cluster solution. The cluster R² values were as follows: 56.2% (four clusters), 60.6% (five clusters), 63.9% (six clusters), 66.6% (seven clusters), 68.4% (eight clusters), and 70.0% (nine clusters). The improvement in R² values slows after six clusters. Cluster means are provided in Table 3 and answer RQ1 and RQ2. The last two columns of Table 3 profile the clusters on total news consumption and entropy, which were not included in the cluster analysis. We also profiled the repertoires on the available demographic variables to answer RQ3 (Table 4). For age and income a one-way analysis of variance (ANOVA) using Duncan groupings was performed. For the other variables a chi-square test of independence and the adjusted standardized residuals were analyzed to understand which repertoires were significantly different from the others.

Table 3.

News Repertoires: Cluster Means Across News Factors Used in Clustering, and Profile Variables Total News Exposure and Exposure Diversity.

News repertoire	n	Fox	CNN	MSNBC	Broadcast	Public	log (total news)	Exposure diversity
Conservative Light	25,776 (15.7%)	0.48	−3.89	−1.64	−0.51	−0.53	8.11	0.26
Liberals	26,621 (16.2%)	−0.83	1.03	2.55	0.30	−0.12	8.55	0.46
Nonpartisans	26,397 (16.0%)	−1.06	1.09	−1.99	0.26	0.04	8.48	0.33
News Avoiders	26,023 (15.8%)	−5.25	−3.23	−1.77	−1.33	−0.79	6.86	0.12
Newshounds	31,984 (19.4%)	2.52	3.15	3.45	0.80	1.24	9.59	0.71
Conservatives	27,711 (16.8%)	3.38	0.99	−1.35	0.25	−0.04	9.44	0.51

Note. For cluster means across news factors, positive values indicate above-average viewing times, and negative values indicate below-average viewing. Total news exposure and exposure diversity are separate profile variables for descriptive and comparative purposes across repertoires. See supplemental material for robustness checks using larger universes of MSO subscribers to profile the clusters. The underlying story is largely the same. MSO = multisystem operator.

Table 4.

Demographic Profile of the Repertoires.

News repertoire	n	Age	Income	Married (%)	Kids (%)	High school (%)	Grad school (%)
Conservative Light	25,776	54.7	5.24	64.6	38.4	62.7	12.9
Liberals	26,621	54.6	5.00	61.9	38.4	64.4	12.2
Nonpartisans	26,397	55.5	5.28	64.7	38.1	61.1	13.8
News Avoiders	26,023	50.2	4.92	56.7	39.6	66.4	11.0
Newshounds	31,984	61.3	5.38	70.5	32.8	57.6	15.9
Conservatives	27,711	61.3	5.73	74.2	33.2	54.8	17.2

Note. See supplemental material for robustness checks using larger universes of MSO subscribers to profile the clusters. The underlying story is largely the same. MSO = multisystem operator.

To test the remaining hypotheses and research questions, the analysis predicted (a) whether or not at least one person in the household voted in the 2016 general election and (b) whether they voted in the Democratic versus Republican primary. Mixed effect logistic regression models were estimated predicting each from news exposure variables, controlling for age, education, income, and corresponding voting behaviors in 2012. For different states, we allow for random intercepts and slopes for total news exposure and exposure diversity. In the case of intercepts, this means that we assume intercepts vary across states (e.g., the base level for voting Democratic could be higher in some states than others) and follow a normal distribution. The mean and standard deviation of this distribution are estimated by the model, where a standard deviation of 0 indicates that all states have the same intercept. The slopes for total news exposure also vary across states and follow a normal distribution, whose mean and standard deviation are estimated. All numerical variables were standardized before estimating the coefficients.

Results

The repertoire (cluster) profiles in Table 3 answer RQ1 and RQ2. The first cluster (Conservative Light) consumed more Fox than the average household in the sample, and less than average CNN, MSNBC, broadcast news, and public TV. The second cluster (Liberals) consumed above average amounts of MSNBC and CNN, and was also above average on broadcast. It was somewhat below average on Fox and Public TV. The third cluster (Nonpartisans) was somewhat above average on CNN and broadcast, roughly average on public TV, and below average on the more partisan Fox and MSNBC outlets. Cluster 4 was below average on all news and indicates the repertoire for News Avoiders. Conversely, Cluster 5 was above average on all variables and indicated the repertoire for Newshounds. Cluster 6 (Conservatives) was similar to cluster one in that it had a Fox-centric news diet, but this cluster also consumed above average amounts of CNN and broadcast TV. It should also be noted that Conservatives are similar to the Liberals with one exception: Conservatives watch a lot of Fox and avoid MSNBC, while Liberals do just the opposite, watching a lot of MSNBC and avoiding Fox.

In summary, there was a spectrum of repertoires from avoiders (Cluster 4) to hounds (Cluster 5). Cluster 1 had low news consumption mainly from Fox. Cluster 3 also had fairly light news consumption, mainly from CNN and broadcast. Clusters 2 and 6 had broader news diets, with two being more liberal and six more conservative. Collectively, these findings suggest that both partisan and more diverse, nonpartisan news repertoires coexist (RQ1).

To answer RQ3, Table 4 offers demographic profiles of the repertoires. Newshounds and Conservatives are more established, with significantly³ greater ages and incomes. They are more likely to have graduate degrees and be married, and less likely to have children. News Avoiders are the youngest, have the lowest income, are the least likely to have a graduate degree, more likely to have high school education only, and are more likely to have kids and be unmarried. The first three clusters (Conservative Light, Liberals, and Nonpartisans) are also fairly likely to have only a high school education and are between the other groups on age and income.

To test H1 and RQ4, logistic regression models were estimated to explain and profile voting behaviors across the repertoires. Parameter estimates are provided in Table 5. Columns 2 to 4 show the results for predicting voting in the general election, controlling for voting in the 2012 general election. Past voting has, by far, the largest z statistic (142.7). After controlling for past voting, the coefficient for both total news exposure (b = 0.1293, p = .0002) and exposure diversity (b = .0674, p < .0001) have positive effects on voting. The standard deviation of slopes for exposure diversity is nearly 0, indicating that it has a consistent positive association with voting across states. The standard deviation for total news exposure is larger (0.080), which indicates that in some states total news consumption has a stronger positive effect on voting than in others. Interestingly, the higher z statistic for entropy (6.8) indicates a stronger effect for exposure diversity than total news exposure (z = 3.7) on voter turnout.

Table 5.

Estimates From Logistic Regression Predicting Voting in the General Election and in the Democratic Versus Republican Primary.

Dependent variable: Vote 2016	General election (n = 165,036)			Primary (n = 78,942)
Parameter	Estimate	SE	z value	Estimate	SE	z value
Intercept	0.3295σ = .438	0.1790	1.2	−1.2675 σ = .757	0.3390	−3.74
Vote 2012	2.2713	0.0159	142.7	3.0603	0.0329	93.0
log (total news exposure)	0.1293 σ = .080	0.0348	3.7	−0.0870 σ = .092	0.0436	−2.0
Exposure diversity	0.0674 σ = .007	0.0100	6.8	−0.0260σ = .085	0.0401	−0.65
Age	−0.0332	0.0072	−4.6	−0.1301	0.0109	−11.9
Age²	−0.0866	0.0060	−14.5	−0.0169	0.0098	−1.7
Education = High school	0	—	—	—	—	—
Education = College graduate	0.0606	0.0177	3.4	−0.1378	0.0237	−5.8
Education = Grad school	0.2998	0.0252	11.9	−0.1278	0.0288	−4.4
Income	0.2034	0.0075	27.0	−0.3900	0.0110	−35.6

Note. Estimates for Primary Voting indicate the effect on the odds of voting in the Democratic Primary, where positive values indicate voting in the Democratic Primary and negative values indicate voting in the Republican Primary. Bold values are significant at the .01 level and italic values are significant at the .05 level. See supplemental material for separate models for each of the six states. The results are similar.

The same analysis was repeated for the probability of voting Democratic versus Republican in the primary (columns 5-7; positive values indicate Democratic Primary, while negative values indicate Republican Primary). The coefficient for total news exposure (b = −0.0870, p = .0459) is only borderline significantly different from 0, and exposure diversity (b = −0.0260, p = .5158) is not significant at all. The standard deviations for both are large, suggesting that in some states more news consumption is associated with voting Democratic, while in other states Republican.

In summary, the first hypothesis is partially supported, where total news exposure has a positive relationship with voter turnout but not primary voting. Regarding RQ4, exposure diversity similarly predicts general election voter turnout, but not primary voting. Although both total and diversity of news exposure predict voter turnout, exposure diversity demonstrates a stronger effect. These findings highlight the value of considering both total news exposure and exposure diversity and the differential effects of those measures on voting behaviors.

To answer RQ5, voting profiles of the clusters are provided in Table 6. The results tell a similar story to the regressions, with those consuming more news being more likely to vote in the general election. The Newshounds and Conservatives are the most likely to vote. The lighter news clusters (Conservative Light, Liberals, and Nonpartisans) are somewhat less likely to vote, while the News Avoiders are least likely to vote. The last two columns profile primary voting for Republicans. As one would expect, the viewers of conservative news media (Conservative Light; Conservatives) are most likely to vote Republican, while liberal media viewers (Liberals) are least likely (52.3%). Our sample comes from six states that range from very likely to vote Republican to swing states, and so even the Liberals are likely to vote Republican.

Table 6.

Voting Profiles Across News Repertoires.

News repertoire	Vote general election		Vote primary
News repertoire	n	% yes	n	% Republican
Conservative Light	25,776	76.6	11,910	84.1
Liberals	26,621	76.6	11,681	52.3
Nonpartisans	26,397	79.0	11,852	65.5
News Avoiders	26,023	71.7	9,520	65.0
Newshounds	31,984	83.3	16,987	68.1
Conservatives	27,711	86.0	16,992	92.1
Total	95,001	76.1	45,927	84.8

Note. For the general election all pairwise differences have p < .0001 except for one: conservative light and liberals are not significantly different (p = .9923). For the primary, all differences have p < .0001 except for nonpartisans and avoiders, where p = .4195 > .05.

Table 7 gives the results from a mixed-effect logistic regression predicting if the household voted in the general election and separately in the Democratic versus Republican primary from the news consumption factors after controlling for age, education, income, and 2012 voting behaviors. We include random intercepts for each state but no random slopes. We begin with the general election. As with the other model, voting in 2012 has, by far, the largest z statistic (141.6), indicating that past voting is the best indicator of future voting. After controlling for past voting and demographics, there is a strong positive association between the amount of Fox news and voting (z = 16.0). Public TV also has a strong positive effect (z = 9.5). Conversely, MSNBC has a significant negative association with voting (z = −6.7). The effects for broadcast (z = 3.7) and CNN (z = 2.3) are smaller, but positive. In summary, the analysis suggests viewing Fox and public TV motivated Americans to vote in 2016.

Table 7.

Estimates From Logistic Regression Predicting Voting in the General Election and in the Democratic Versus Republican Primary.

DV: Vote 2016	General election (n = 165,036)			Primary (n = 78,942)
Parameter	Estimate	SE	z value	Estimate	SE	z value
Intercept	0.2481σ = .442	0.1808	1.4	−1.3793 σ = .771	0.3441	−4.0
Vote 2012	2.2587	0.0159	141.6	2.7592	0.0354	77.9
log (Fox)	0.1369	0.0085	16.0	−1.0937	0.0136	−80.3
log (CNN)	0.0211	0.0091	2.3	0.4404	0.1401	31.4
log (MSNBC)	−0.0560	0.0084	−6.7	0.4836	0.0120	40.3
log (public)	0.0745	0.0079	9.5	0.0422	0.0109	3.9
log (broadcast)	0.0290	0.0079	3.7	0.0973	0.0124	7.8
Age	−0.0391	0.0072	−5.4	−0.0669	0.0118	−5.7
Age²	−0.0921	0.0060	−15.4	−0.0011	0.0106	−0.1
Education = High school	0	—	—	0	—	—
Education = College	0.0571	0.0177	3.2	−0.1248	0.0257	−4.9
Education = Grad school	0.2934	0.0253	11.6	−0.1699	0.0316	−5.4
Income	0.1975	0.0076	26.2	−0.3578	0.0118	−30.2

We now discuss the results for Primary voting, shown in the last three columns of Table 7 (positive values = Democratic Primary; negative values = Republican Primary). The signs of the slope estimates are intuitive, with those who consume more Fox being less likely to vote Democratic (z = −80.3) and MSNBC has a positive association with voting Democratic (z = 40.3). These findings support H2. What is less obvious is that the slopes for CNN (z = 31.4), broadcast (z = 7.8), and public (z = 3.9) TV all have significantly positive slopes, although their magnitudes are smaller than for Fox or MSNBC, especially for broadcast and public TV.

Discussion

Individuals create repertoires to cope with an abundant media environment and limited attention. Focusing on television news repertoires, this study explored both composition and diversity in news diets, as well as the link to voting behaviors in the 2016 U.S. election. The results indicate a spectrum of news repertoires, with the existence of both partisan, as well as more diverse repertoires. In fact, the six repertoires seem to suggest evidence for the two dimensions of selective exposure most commonly explored in previous research, where the information environments that individuals create are defined by political ideology and/or political interest. Moreover, profiling the clusters reveals some important distinctions among individuals who inhabit different news environments.

Moving beyond the basic composition of repertoires, the analysis explored the relationship between total and degree of diversity in news exposure and voting behaviors. Both total news exposure and exposure diversity predict voter turnout, with exposure diversity revealing a stronger effect. Finally, the results indicate a strong link between exposure to Fox and MSNBC and voting in the Republican and Democratic primaries, respectively.

The results suggest two dimensions of exposure diversity, one defined by ideology and the other indicating political interest. Although past selective exposure research has tended to focus on either type, this study captures both ideology- and interest-driven selective exposure and demonstrates that they coexist. Regarding ideology, repertoires range from the left (Liberals) to the right (Conservative Light; Conservatives). In addition to the ideologically defined repertoires, the Nonpartisans repertoire is more neutral. These individuals are below average on partisan news outlets (Fox, MSNBC) and above average on CNN and broadcast news (which includes local and network news).

On the surface, this would seem to suggest evidence that individuals may be constructing echo chambers through their news exposure. However, while there is evidence of repertoires defined by ideology, those same repertoires also exhibit at least average levels of viewing for some neutral outlets. Moreover, the Nonpartisans and Newshounds repertoires indicate that many individuals engage in more diverse news exposure. Collectively, the results are consistent with recent research that finds some evidence of selective exposure to partisan news media, but not selective avoidance of counter-attitudinal information (Garrett, 2009a, 2009b; Garrett et al., 2013; Garrett & Stroud, 2014; Gentzkow & Shapiro, 2011; Holbert et al., 2012; Ksiazek, 2016; Prior, 2013; Trilling & Schoenbach, 2013; Webster & Ksiazek, 2012; Weeks et al., 2016). In addition, the results indicate overlapping patterns of general interest news use across many of the repertoires. This is also consistent with recent research (e.g., Webster & Ksiazek, 2012; Weeks et al., 2016) that offers empirical challenges to extreme concerns about audience fragmentation and polarization.

As for political interest, the repertoires suggest the existence of both Newshounds (above average on all news factors) and News Avoiders (below average on all news factors). Together with the ideology-defined repertoires, the findings confirm recent research based on self-report data. For instance, Edgerly (2015) also found repertoires of news avoiders and newshounds (what she calls “omnivores”), as well as repertoires that lean conservative and liberal. Complementing that work, the present study also offers a more nuanced picture of both avoiders and conservative news consumers. For example, we find evidence of traditional News Avoiders across all types of news, as well as other kinds of avoiders (e.g., Conservative Light, Nonpartisans). These groups generally avoid much of the news, but have clear (dis)interests in specific types of news. At the same time, our analysis revealed two groups of conservatives, one that is only interested in conservative news and nothing else (Conservative Light) and another that consumes a lot of news overall, but especially high levels of Fox (Conservatives).

The audience profiles offer some interesting differences across the repertoires, particularly when comparing Newshounds/Conservatives and News Avoiders. Newshounds and Conservatives are older with higher incomes, as well as more likely to have advanced degrees and be married without children. News Avoiders are the youngest and have the lowest income and education. This group is also more likely to have children than some of the other groups, but less likely to be married.

Moving to the analysis of the relationship between exposure and voting behaviors, the results offer partial support for H1. The logistic regression predicting voter turnout from total news exposure shows a hockey-stick pattern: Up to a certain point, news consumption has little effect on voting, but as consumption reaches a certain point voting increases quickly. This is consistent with the news repertoire profiles in Table 6, where Newshounds were much more likely to vote than News Avoiders. The results also show a positive relationship with primary voting, but the significance is marginal and the standard deviation is large, suggesting the effect on voting in the Democratic Primary varies widely from state to state.

Turning to the relationship between exposure diversity and voting, entropy was used to capture the degree of exposure diversity in news diets. Entropy had a positive effect on general voter turnout, where voting increased for higher levels of entropy (i.e., more diversity). This adds new evidence to extant research that has shown inconsistent results regarding the relationship between exposure diversity and political participation (Helberger & Wojcieszak, 2018). Given evidence of a circular relationship between news exposure and political participation, perhaps greater exposure diversity is a manifestation of higher levels of political interest and engagement, which would explain the positive relationship between exposure diversity and voter turnout. Unlike general turnout, exposure diversity did not predict primary voting. Related, the findings indicate that those who consume more Fox News are less likely to vote in the Democratic primary, while those who watch more MSNBC are more likely to vote in the Democratic primary. What is interesting is that all other news factors (CNN, broadcast, public) positively predicted voting in the Democratic primary. Collectively, the findings suggest that total and diversity of exposure have differential effects on voting behaviors, while individuals from both parties are more likely to vote in their primaries if they watch partisan news that matches their beliefs.

Big Data: Benefits and Limitations

As with all research designs, this study has strengths and weaknesses. The main strength is the digital records of overt behaviors for a very large number of subscribers in six states over a long period of time, which do not rely on self-reporting of media use behaviors. For example, most sources of common method bias (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003) are avoided with our unobtrusive measures of media consumption and voting, including social desirability, mood-state, priming, context-induced, question wording, and implicit theory biases. It is not uncommon for respondents to say they do one thing, but actually do another, especially in this case where there is a social stigma attached to not voting. Similarly, as we note above, Prior (2009) found that self-reported news exposure is largely overestimated (on the order of 3-8 times) when compared with the type of metered exposure data we use in this study.

There are also some limitations. First, the data are recorded at the device level (i.e., cable box), and aggregated to the household, and thus do not provide information on exactly who was watching, if anyone. This is a commonly cited disadvantage of big data techniques for capturing media use (Kim, 2018). One possible way to address this is to replicate the analysis on households comprised of only one individual, essentially creating a subsample of individual-level data. The appendix reports the results of a robustness check comparing our household-level results against single-person-only households, with minor differences (e.g., smaller effect estimates) but roughly the same findings. Second, while the data set includes a complete census of all subscribers to this MSO, it does not provide information on its nonsubscribers, which include those who (a) do not subscribe to cable, (b) use an alternative cable source such as satellite TV or another MSO, or (c) do not live within geographic regions served by this MSO. Third, while the data set is very large with more than 12 billion recorded channel changes over the 9-month period, it is a convenience sample, albeit a large one with broad coverage. In particular, we have a disproportionately large fraction of Republicans in our sample, which will bias the estimates of the intercepts in our regression models; however, we note that the slope estimates, which are used to test our hypotheses, remain unbiased (we developed a simulation program in SAS to test this claim and our code is available). Finally, despite increasing calls for capturing media use across multiple platforms, the current study is limited to TV news viewing.

It is noteworthy that the biases of our behavioral-data research design complement those from self-reported surveys, and thus when we corroborate findings it provides an additional robustness test. We urge future researchers to weigh the benefits and limitations of device-based behavioral data and explore the possibility of merging server-centric (i.e., user log data) and user-centric (panel-based) data.

Future Research

Our findings also suggest future research opportunities. It is possible that our results are indicative of a red-states phenomenon given the fact that our sample was more likely to vote in the Republican (73%) than Democratic (27%) primary, and future research could test this by offering a red-states/blue-states comparison of news repertoires. Related, to complement the increasingly common focus on partisan news exposure, we have seen a recent desire to know more about neutral news consumers (e.g., Bou-Hamad & Yehya, 2016). Although the current study contributes to that understanding by profiling those with a neutral news repertoire, there is still much to learn about these types of individuals. In addition, public opinion polling shows a declining trust in the media and a growing preference for soft news. Future research could explore the implications of these trends in news repertoires. With an influx of soft news programs (e.g., sports, business, entertainment, etc.), future research could integrate other types of news into our understanding of repertoires. Finally, the nature of the positive relationship between total news exposure and primary voting (marginal significance, large standard deviation) suggests the effect on voting in the Democratic and Republican primaries varies widely from state to state. Although we were able to match data for six states, we encourage future research to explore variations in news exposure and voting behavior across a larger sample of states, especially one that allows for a red-states/blue-states comparison.

Conclusion

The tendency to construct news repertoires can lead to stark differences in the public’s understanding of the world. This study found the existence of notably different repertoires, some defined by ideologically driven selective exposure, others defined by political interest (i.e., a preference or disinterest in the news), and others that were seemingly neutral. In turn, exposure diversity and total news exposure differentially predict voting behavior. Although the impact on 2016 U.S. general election turnout was positive for both measures of news exposure, exposure diversity demonstrated a stronger effect. In other words, more news exposure and more diverse news exposure matter for voting, but diversity seems to have a greater impact on general election turnout. Moreover, there was a clear relationship between exposure to partisan news media and voting in that respective party’s primary election, confirming the effect of partisan selective exposure.

Supplemental Material

DS_10.11771077699018815892 – Supplemental material for Television News Repertoires, Exposure Diversity, and Voting Behavior in the 2016 U.S. Election

Supplemental material, DS_10.11771077699018815892 for Television News Repertoires, Exposure Diversity, and Voting Behavior in the 2016 U.S. Election by Thomas B. Ksiazek, Su Jung Kim and Edward C. Malthouse in Journalism & Mass Communication Quarterly

Footnotes

Appendix

Robustness Check Comparing Original Household-Level Estimates With Single-Person-Only Households.

	General election		Primary election
	Original estimaten = 94,075	Single person householdsn = 12,435	Original estimaten = 23,143	Single person householdsn = 1,717
Intercept	−1.1848	−0.9719	−2.5196	−2.7286
Log (total)	0.071	.0540	0.012	.0522
Diversity (entropy)	0.1913 z = 7.31	0.0976 z = 1.57	0.372 z = 3.55	0.2724
Age	−0.00111	0.000742	−0.0141	−0.0166
Education = High school	−0.1124	−0.2274	0.0503	−0.0545
Education = College	−0.0774	0.0470	0.0142	0.3203
Income	0.0803	0.00802	−0.1008	−0.0209
Vote 12	2.3928	0.0569	5.1275	4.9570

Acknowledgements

The authors thank Wei-Lin Wang for helpful modeling advice; the Spiegel Research Center for access to the data; and Jake Atlas and Jeff Smith for help with data preparation.

Declaration of Conflicting Interests

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

Funding

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

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Thomas B. Ksiazek (PhD, Northwestern University) is an associate professor in the Department of Communication at Villanova University, USA. His research interests include new forms of user engagement with the news, implications of audience behavior for society and the field of journalism, patterns of cross-platform media use, and the application of network analysis to the consumption and production of media.

Su Jung Kim (PhD, Northwestern University) is an assistant professor in the Annenberg School for Communication and Journalism at the University of Southern California. Her research interests include cross-platform media consumption, electronic word-of-mouth (e-WOM), and social/mobile media usage. Her current projects examine the patterns and implications of digital media consumption and engagement.

Edward C. Malthouse is the Erastus Otis Haven Professor of Integrated Marketing Communications and Professor of Industrial Engineering and Management Science at Northwestern University and the Research Director for the Spiegel Center for Digital and Database Marketing. He received his PhD in statistics from Northwestern University, and his research interests focus on customer engagement and experiences; digital, social and mobile media; big data; and media management.

References

Anderson

(2006). The long tail: Why the future of business is selling less of more. New York, NY: Hyperion.

Barlett

D. L.

Drew

P. B.

Fahle

E. G.

Watts

W. A.

(1974). Selective exposure to a presidential campaign appeal. Public Opinion Quarterly, 38, 264-270. Retrieved from http://www.jstor.org/stable/2748054

Bennett

W. L.

Iyengar

(2008). A new era of minimal effects? The changing foundations of political communication. Journal of Communication, 58, 707-731. doi:10.1111/j.1460-2466.2008.00410.x

Bou-Hamad

Yehya

N. A.

(2016). Partisan selective exposure in TV consumption patterns: A polarized developing country context. Communication Research. Advance online publication. doi:10.1177/0093650216681896

Boyd

Zaff

Phelps

Weiner

Lerner

(2011). The relationship between adolescents’ news media use and civic engagement: The indirect effect of interpersonal communication with parents. Journal of Adolescence, 34, 1167-1179.

Chaffee

Miyo

(1983). Selective exposure and the reinforcement hypothesis: An intergenerational panel study of the 1980 presidential campaign. Communication Research, 10(1), 3-36. doi:10.1177/009365083010001001

Chaffee

Saphir

M. N.

Graf

Sandvig

Hahn

(2001). Attention to counter-attitudinal messages in a state election campaign. Political Communication, 18, 247-272. doi:10.1080/10584600152400338

Davenport

T. H.

Beck

J. C.

(2001). The attention economy: Understanding the new currency of business. Boston, MA: Harvard Business School Press.

De Vreese

Boomgaarden

. (2006). News, political knowledge and participation: The differential effects of news media exposure on political knowledge and participation. Acta Politica, 41, 317-341. doi:10.1057/palgrave.ap.5500164

10.

Dilliplane

(2011). All the news you want to hear: The impact of partisan news exposure on political participation. Public Opinion Quarterly, 75, 287-316. doi:10.1093/poq/nfr006

11.

Dvir-Gvirsman

Tsfati

Menchen-Trevino

(2016). The extent and nature of ideological selective exposure online: Combining survey responses with actual web log data from the 2013 Israeli elections. New Media & Society, 18, 857-877. doi:10.1177/1461444814549041

12.

Edgerly

(2015). Red media, blue media, and purple media: News repertoires in the colorful media landscape. Journal of Broadcasting & Electronic Media, 59(1), 1-21. doi:10.1080/08838151.2014.998220

13.

Ferguson

D. A.

(1992). Channel repertoire in the presence of remote control devices, VCRs and cable television. Journal of Broadcasting & Electronic Media, 36, 83-91. doi:10.1080/08838159209364156

14.

Ferguson

D. A.

Perse

E. M.

(1993). Media and audience influences on channel repertoire. Journal of Broadcasting & Electronic Media, 37, 31-47. doi:10.1080/08838159309364202

15.

Garrett

R. K.

(2009a). Echo chambers online? Politically motivated selective exposure among Internet news users. Journal of Computer-Mediated Communication, 14, 265-285. doi:10.1111/j.1083-6101.2009.01440.x

16.

Garrett

R. K.

(2009b). Politically motivated reinforcement seeking: Reframing the selective exposure debate. Journal of Communication, 59, 676-699. doi:10.1111/j.1460-2466.2009.01452.x

17.

Garrett

R. K.

(2013). Selective exposure: New methods and new directions. Communication Methods and Measures, 7, 247-256. doi:10.1080/19312458.2013.835796

18.

Garrett

R. K.

Carnahan

Lynch

E. K.

(2013). A turn toward avoidance? Selective exposure to online political information. Political Behavior, 35(1), 113-134. doi:10.1007/s11109-011-9185-6

19.

Garrett

R. K.

Stroud

N. J.

(2014). Partisan paths to exposure diversity: Differences in pro- and counterattitudinal news consumption. Journal of Communication, 64, 680-701. doi:10.1111/jcom.12105

20.

Gentzkow

Shapiro

J. M.

(2011). Ideological segregation online and offline. The Quarterly Journal of Economics, 126, 1799-1839. doi:10.1093/qje/qjr044

21.

Gitlin

(1998). Public sphere or public sphericules? In Leibes

Curran

(Eds.), Media, ritual and identity (pp. 43-67). London, England: Routledge.

22.

Gottfried

Barthel

Shearer

Mitchell

(2016, February 4). The 2016 presidential campaign—A news event that’s hard to miss. Pew Research Center. Retrieved from http://www.journalism.org/2016/02/04/the-2016-presidential-campaign-a-news-event-thats-hard-to-miss/

23.

Graber

D. A.

(2001). Processing politics: Learning from television in the Internet age. Chicago, IL: The University of Chicago Press.

24.

Graf

Aday

(2008). Selective attention to online political information. Journal of Broadcasting & Electronic Media, 52(1), 86-100. doi:10.1080/08838150701820874

25.

Hasebrink

Popp

(2006). Media repertoires as a result of selective media use. A conceptual approach to the analysis of patterns of exposure. Communications, 31, 369-387. doi:10.1515/COMMUN.2006.023

26.

Heeter

Greenberg

(1985). Cable and program choice. In Zillman

Bryant

(Eds.), Selective exposure to communication (pp. 203-224). Hillsdale, NJ: Lawrence Erlbaum.

27.

Helberger

Wojcieszak

(2018). Exposure diversity. In Napoli

(Ed.), Handbooks of Communication Science Series. Mediated communication (pp. 535-560). Berlin, Boston: de Gruyter Mouton.

28.

Holbert

R. L.

Hmielowski

J. D.

Weeks

B. E.

(2012). Clarifying relationships between ideology and ideologically oriented cable TV news use: A case of suppression. Communication Research, 39, 194-216. doi:10.1177/0093650211405650

29.

Hollander

B. A.

(2008). Tuning out or tuning elsewhere? Partisanship, polarization, and media migration from 1998 to 2006. Journalism & Mass Communication Quarterly, 85(1), 23-40. doi:10.1177/107769900808500103

30.

Iyengar

Hahn

K. S.

(2009). Red media, blue media: Evidence of ideological selectivity in media use. Journal of Communication, 59(1), 19-39. doi:10.1111/j.1460-2466.2008.01402.x

31.

Jamieson

K. H.

Cappella

J. N.

(2008). Echo chamber: Rush Limbaugh and the conservative media establishment. New York, NY: Oxford University Press.

32.

Sypher

(2014). The role of news media use and demographic characteristics in the possibility of information overload prediction. International Journal of Communication, 8, 699-714.

33.

Kanervo

Zhang

Sawyer

(2005). Communication and democratic participation: A critical review and synthesis. Review of Communication, 5, 193-236. doi:10.1080/15358590500422585

34.

Katz

(1996). And deliver us from segmentation. The Annals of the American Academy of Political and Social Science, 546, 22-33. doi:10.1177/0002716296546001003

35.

Kim

S. J.

(2016). A repertoire approach to cross-platform media use behavior. New Media & Society, 18, 353-372. doi:10.1177/1461444814543162

36.

Kim

S. J.

(2018). Audience measurement and analysis. In Albarran

A. B.

Mierzejewska

Jung

(Eds.), The handbook of media management and economics (2nd ed., pp. 379-393). New York, NY: Routledge.

37.

Kim

S. J.

Webster

J. G.

(2012). The impact of a multichannel environment on television news viewing: A longitudinal study of news audience polarization in South Korea. International Journal of Communication, 6, 838-856. Retrieved from http://ijoc.org/index.php/ijoc/article/view/1054

38.

Ksiazek

T. B.

(2016). Partisan audience polarization: Beyond selective exposure. Atlantic Journal of Communication, 24, 216-227. doi:10.1080/15456870.2016.1208658

39.

Ksiazek

T. B.

Malthouse

E. C.

Webster

J. G.

(2010). News-seekers and avoiders: Exploring patterns of total news consumption across media and the relationship to civic participation. Journal of Broadcasting & Electronic Media, 54, 551-568. doi:10.1080/08838151.2010.519808

40.

Lanham

R. A.

(2006). The economics of attention: Style and substance in the age of information. Chicago, IL: The University of Chicago Press.

41.

Lee

(2010). The influence of family viewing preferences on television consumption in the era of multichannel services. Asian Journal of Communication, 20, 281-298.

42.

Lee

Yang

(2014). Political knowledge gaps among news consumers with different news media repertoires across multiple platforms. International Journal of Communication, 8, 597-617.

43.

Morris

J. S.

(2005). The fox news factor. The International Journal of Press/Politics, 10, 56-79. doi:10.1177/1081180X05279264

44.

Mutz

D. C.

Young

(2011). Communication and public opinion: Plus ça change? Public Opinion Quarterly, 75, 1018-1044. doi:10.1093/poq/nfr052

45.

Napoli

(2011). Audience evolution: New technologies and the transformation of media audiences. New York, NY: Columbia University Press.

46.

Neuendorf

K. A.

Atkin

D. J.

Jeffres

L. W.

(2001). Reconceptualizing channel repertoire in the urban cable environment. Journal of Broadcasting & Electronic Media, 45, 464-482. doi:10.1207/s15506878jobem4503_6

47.

Nielsen. (2016, September 26). The Nielsen total audience report: Q2 2016. Author. Retrieved from http://www.nielsen.com/us/en/insights/reports/2016/the-nielsen-total-audience-report-q2-2016.html

48.

Nir

Druckman

(2008). Campaign mixed-message flows and timing of vote decision. International Journal of Public Opinion Research, 20, 326-346. doi:10.1093/ijpor/edn035

49.

Norris

(2000). A virtuous circle: Political communications in postindustrial societies. New York, NY: Cambridge University Press.

50.

Papathanassopoulos

Coen

Curran

Aalberg

Rowe

Jones

. . . Tiffen

(2013). Online threat, but television is still dominant. A comparative study of 11 nations’ news consumption. Journalism Practice, 7, 690-704. doi:10.1080/17512786.2012.761324

51.

Pew Research Center. (2014). Political polarization & media habits: From Fox News to Facebook, how liberals and conservatives keep up with politics. Retrieved from http://www.journalism.org/2014/10/21/political-polarization-media-habits/

52.

Pew Research Center. (2016). The modern news consumer: News attitudes and practices in the digital era. Retrieved from http://www.journalism.org/2016/07/07/pathways-to-news/

53.

Podsakoff

MacKenzie

Lee

Podsakoff

(2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879-903. doi:10.1037/0021-9010.88.5.879

54.

Prior

(2007). Post-broadcast democracy: How media choice increases inequality in political involvement and polarizes elections. New York, NY: Cambridge University Press.

55.

Prior

(2009). The immensely inflated news audience: Assessing bias in self-reported news exposure. Public Opinion Quarterly, 73(1), 130-143. doi:10.1093/poq/nfp002

56.

Prior

(2013). Media and political polarization. Annual Review of Political Science, 16, 101-127. doi:10.1146/annurev-polisci-100711-135242

57.

Reagan

(1996). The “repertoire”; of information sources. Journal of Broadcasting & Electronic Media, 40, 112-119. doi:10.1080/08838159609364336

58.

Schrøder

K. C.

(2015). News media old and new: Fluctuating audiences, news repertoires and locations of consumption. Journalism Studies, 16(1), 60-78. doi:10.1080/1461670X.2014.890332

59.

Skovsgaard

Shehata

Strömbäck

(2016). Opportunity structures for selective exposure: Investigating selective exposure and learning in Swedish election campaigns using panel survey data. The International Journal of Press/Politics, 21, 527-546. doi:10.1177/1940161216658157

60.

Strömbäck

Djerf-Pierre

Shehata

(2013). The dynamics of political interest and news media consumption: A longitudinal perspective. International Journal of Public Opinion Research, 25, 414-435. doi:10.1093/ijpor/eds018

61.

Stroud

N. J.

(2008). Media use and political predispositions: Revisiting the concept of selective exposure. Political Behavior, 30, 341-366. doi:10.1007/s11109-007-9050-9

62.

Stroud

N. J.

(2010). Polarization and partisan selective exposure. Journal of Communication, 60, 556-576. doi:10.1111/j.1460-2466.2010.01497.x

63.

Stroud

N. J.

(2011). Niche news: The politics of news choice. New York, NY: Oxford University Press.

64.

Sunstein

(2007). Republic.com 2.0. Princeton, NJ: Princeton University Press.

65.

Taneja

Webster

J. G.

Malthouse

E. C.

Ksiazek

T. B.

(2012). Media consumption across platforms: Identifying user-defined repertoires. New Media & Society, 14, 951-968. doi:10.1177/1461444811436146

66.

Tewksbury

(2005). The seeds of audience fragmentation: Specialization in the use of online news sites. Journal of Broadcasting & Electronic Media, 49, 332-348. doi:10.1207/s15506878jobem4903_5

67.

Trilling

Schoenbach

(2013). Patterns of news consumption in Austria: How fragmented are they? International Journal of Communication, 7, 929-953. Retrieved from http://ijoc.org/index.php/ijoc/article/view/1769

68.

Webster

J. G.

(2005). Beneath the veneer of fragmentation: Television audience polarization in a multichannel world. Journal of Communication, 55, 366-382. doi:10.1111/j.1460-2466.2005.tb02677.x

69.

Webster

J. G.

Ksiazek

T. B.

(2012). The dynamics of audience fragmentation: Public attention in an age of digital media. Journal of Communication, 62(1), 39-56. doi:10.1111/j.1460-2466.2011.01616.x

70.

Weeks

Ksiazek

T. B.

Holbert

R. L.

(2016). Partisan enclaves or shared media experiences? A network approach to understanding citizens’ political news environments. Journal of Broadcasting & Electronic Media, 60, 248-268. doi:10.1080/08838151.2016.1164170

71.

Wolfsfeld

Yarchi

Samuel-Azran

(2016). Political information repertoires and political participation. New Media & Society, 18, 2096-2115. doi:10.1177/1461444815580413

72.

Yuan

(2011). News consumption across multiple media platforms. Information, Communication & Society, 14, 998-1016. doi:10.1080/1369118X.2010.549235

73.

Yuan

Webster

J. G.

(2006). Channel repertoires: Using peoplemeter data in Beijing. Journal of Broadcasting & Electronic Media, 50, 524-536. doi:10.1207/s15506878jobem5003_10

Supplementary Material

Please find the following supplemental material available below.

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

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

0.00 MB