Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
How Uncompetitive Elections and Media Consolidation Impoverish the News and Imperil Democracy
Danny Hayes Department of Political Science George Washington University
Jennifer L. Lawless Department of Government
American University [email protected]
Abstract: We investigate the effects of declining electoral competitiveness and rising media consolidation on news coverage of U.S. House campaigns and, subsequently, citizen engagement. Drawing on a detailed analysis of newspaper coverage during the 2010 midterm elections, we first show that competitiveness and a newspaper’s ownership structure affect the volume and substance of coverage. Districts with uncompetitive races and districts served by corporately-owned or large-circulation newspapers see significantly less coverage and less issue coverage than hotly contested districts and those served by smaller, non-corporate outlets. We then merge the news data with survey data from the 2010 Cooperative Congressional Election Study and find that citizens exposed to a lower volume of coverage are less able to evaluate their member of Congress, less likely to express opinions about the House candidates in their districts, and less likely to vote. This is true for people regardless of levels of political attentiveness, indicating that the deleterious consequences of a decline in local coverage are not restricted to the least informed citizens. These analyses suggest that an enrichment of local political news – and an increase in citizen engagement – is likely to come only with a rise in the competitiveness of elections.
1
The most important job in a democracy belongs to citizens: holding their representatives
accountable in elections. By virtually any standard, voters carry out this responsibility most
effectively when they possess an ample supply of substantive information about what their elected
officials are doing (see Althaus 2007). The more they know, the better decisions they will make, and
the more likely they will be to participate. The quality of information at citizens’ disposal, then,
directly affects the quality of the democratic enterprise. In recent years, however, two developments
have raised concerns that the political information voters receive from the news media, already the
subject of much criticism, may be under further threat.1
The first is the decline in the competitiveness of congressional elections. Since the end of
World War II, a growing share of U.S. House seats has become safe for one party; incumbent re-
election rates have increased and district polarization has rendered the outcomes of most contests
predictable before the campaign even begins (e.g., Cox and Katz 1996; Mayhew 1974). This change
is evident in the number of close House elections. In the congressional elections of the 1950s, 39%
of seats each cycle on average were decided by 20 points or less. By the 1970s, that figure was down
to 24%. And after a slight uptick in the 1990s, it had fallen even further by the 2000s.2 The number
of Americans living in competitive House districts has shrunk to a small minority.
The available evidence suggests that this decline in electoral competitiveness should change
the calculus for journalists and ultimately carry negative consequences for the quality of news.3
Because close elections have uncertain outcomes, they generate more drama and are inherently more
newsworthy (e.g., Bennett 2011; Graber 2009). In addition, competitive races produce more
1 For accounts of how well (or poorly) the news media provide citizens with substantive political information,
see, e.g., Bennett 2011; Farnsworth and Lichter 2011; Patterson 1980; 1994. 2 We calculated these figures from Gary Jacobson’s widely used data on congressional elections. Other
measures (e.g., Abramowitz, Alexander, and Gunning 2006, 76) show a similar trend. 3 Defining and setting standards for news quality is challenging (Zaller 1999) and contestable (Bennett 2003;
Zaller 2003). For our purposes, we define news as higher in quality when there is more coverage of politics and when there is more coverage of issues (see Dunaway 2008; Hayes 2010).
2
campaign activity (and more conflict between candidates), so they offer more campaign-trail
developments for reporters to cover (e.g., Bruni 2002). Indeed, numerous studies have found a
strong relationship between competitiveness and news attention (Arnold 2004; Clark and Evans
1983; Gershon 2012a; Goldenberg and Traugott 1984; Kahn and Kenney 1999; Vinson 2003). And
because competition leads to more coverage, competitive contests also produce reporting about
candidates’ issue positions (Kahn and Kenney 1999; Westlye 1991; though see Hayes 2010). By
promoting substantive coverage of elections, then, competitiveness tends to breed a news
environment that more closely approximates democratic ideals. The corollary, of course, is that less
electoral competition produces an information environment with less volume and substance.
The second development is consolidation in the media industry, as smaller news outlets have
been acquired by larger, often corporately-owned, firms. From 1960 to 1980, for example, the
Gannett Company alone bought up 57 newspapers. Meanwhile, the total number of newspapers in
the United States has declined, falling by about 14% since 1990. Two-newspaper towns are now
almost non-existent, leaving fewer outlets as the primary local news source for increasingly large
geographic areas. And as the financial margins of the remaining newspapers have dwindled,
reporting resources have been cut: newsroom staffs across the country shrunk by 25% between
2001 and 2009.4
Research suggests that, as is the case with electoral competitiveness, media ownership and
market forces also affect the volume and substance of political news. When multiple congressional
districts share a media market, for instance, coverage of any one district within that market decreases
(Arnold 2004; see also Manheim 1974). As a newspaper’s circulation size grows – thus, increasing
the likelihood of having readers in multiple congressional districts – the amount of campaign
coverage in a district falls (Gershon 2012a; 2012b; Goldenberg and Traugott 1984; Tidmarch and
4 For an overview of, and data pertaining to, these trends in media consolidation, see Edmonds, Guskin, and Rosenstiel 2011; Kirchhoff 2009; Rosenstiel and Mitchell 2011.
3
Karp 1983; Vinson 2003). Corporate ownership can also affect the news. Independently-owned
newspapers are more likely than chain-owned papers to provide regular coverage of their House
incumbents (Schaffner and Sellers 2003), perhaps because “local owners,” as opposed to “out of
market” owners are more invested in the local community (Napoli and Yan 2007). Privately-owned
papers are most likely to produce issue coverage of a campaign, whereas small corporate papers are
least likely to do so (Dunaway 2008). In short, media consolidation can lead to less, and less
substantive, news.
Although these twin developments – a decline in electoral competitiveness and a rise in
media consolidation – have potentially significant implications for the political information at voters’
disposal, our understanding of their consequences has been hampered by two issues. First, we have
little systematic evidence about how these forces affect the information environment in relatively
low-salience elections, such as House races, even though this is where competitiveness and
consolidation are likely to have their strongest influence. Second, changes to the information
environment may have different implications for citizen engagement at the local level than at the
national level, where most research has focused. Previous work has found that the “post-broadcast”
media environment exacerbates inequality in engagement, with the politically aware becoming
“information richer” and the less politically aware becoming “information poorer” (Prior 2007). But
because the choices for political news at the local level are quite limited – in contrast to an ever-
expanding national media environment – declines in the volume and substance of House coverage
may reduce engagement among all citizens, not just those who are less interested in politics.
In this paper, we conduct the first analysis of the effects of competitiveness and media
consolidation on both news content and citizens’ levels of engagement. We do this in two steps.
Drawing on a detailed analysis of newspaper coverage in every House district during the 2010
midterm elections, we first show – in the largest and most comprehensive study of its kind – that
4
competitiveness and a newspaper’s ownership structure affect media coverage in House elections.
Districts with uncompetitive races and districts served by corporately-owned or large-circulation
newspapers see significantly less coverage and less issue coverage than hotly contested districts and
those served by smaller, non-corporate outlets. We then merge the news data with survey data from
the 2010 Cooperative Congressional Election Study and find that citizens exposed to a lower
volume of coverage are less able to evaluate their member of Congress, less likely to express
opinions about the House candidates in their districts, and less likely to vote. This is true for people
regardless of levels of political attentiveness, indicating that the deleterious consequences of a
decline in local coverage are not restricted to the least informed citizens. These analyses suggest that
an enrichment of local political news – and an increase in citizen engagement – is likely to come only
with a rise in the competitiveness of elections.
Uncompetitive Elections, Media Consolidation, and Political Information: Empirical Limitations and Theoretical Expectations
The existing literature suggests that both competitiveness and consolidation should play a
role in shaping the news and citizens’ likelihood of engaging with politics. But our empirical
understanding of these relationships remains under-developed, largely because few studies have
focused on low salience races, such as House elections. Much research examines presidential
(Dalton, Beck, and Huckfeldt 1992; Patterson 1994) or statewide contests, such as Senate or
gubernatorial elections (Dunaway 2008; Kahn and Kenney 1999). These studies, however – by virtue
of the level of office on which they focus – analyze just a small number of contests or news outlets.
In fact, over the course of the last 40 years, only a dozen studies have investigated media coverage in
U.S. House elections at all (Arnold 2004; Fogarty 2013; Clark and Evans 1983; Gershon 2012; 2013;
Goldenberg and Traugott 1984; Larson 1992; Manheim 1974; Orman 1985; Tidmarch and Karp
1983; Vermeer 1987; Vinson 2003). Moreover, this small literature is limited in scope. None of the
5
studies analyzes coverage in more than 100 districts; some focus exclusively on just one (Larson
1992; Orman 1985).5 Many can say little about House coverage specifically because the analyses
combine House and Senate races. And several restrict their inquiries to specific types of candidates
and contests by focusing only on incumbents, contested races, or members implicated in scandals
(e.g., Fogarty 2013).6
Yet there are several analytical and theoretical reasons for focusing on coverage of U.S.
House elections. First, House races offer significant variation in the level of competitiveness and the
ownership structure of newspapers across the country. While a majority of congressional districts
feature safe seats for one of the two major parties, largely because of party polarization (Fleisher and
Bond 2004; Jacobson 2000; 2012; Theriault 2008; Theriault and Rohde 2011), others are quite
competitive, producing high levels of spending and campaign activity. Similarly, the prevalence of
media consolidation varies across congressional districts; some local newspapers are big, some are
small, some are owned by corporations, and some remain family-owned or independent. This means
that examining a large number of House races allows for an assessment of the relationships among
competitiveness, consolidation, news content, and citizen engagement that investigations into
statewide contests cannot provide.
5 Snyder and Stromberg (2010) offer one of the most comprehensive studies of news at the district level. They
examine the volume of coverage members of Congress received (from 1991 – 2002) across 385 congressional districts in 161 newspapers. But they do not focus on campaigns or campaign coverage. Gentzkow and Shapiro (2010) analyze coverage in 429 newspapers, but they focus exclusively on media “slant,” or the similarity between the rhetoric members of Congress use and the language that appears in news articles about them.
6 A body of work largely from economics has focused on the implications of media structure and ownership in
local markets, but most of it is not well-suited to answer questions about news content and citizen engagement. Some studies, for instance, have examined the effects of cross-ownership of television stations and newspaper outlets in the same market, or minority ownership of media properties (e.g., Erb 2010; Oberholzer-Gee and Waldfogel 2009). These analyses, however, address a different set of questions than ours. Still others do not employ politically-relevant measures of media content, such as the volume of political news or the amount of issue coverage; they rely instead on blunt measures (e.g., the overall amount of programming, political or not) or proxies for content (e.g., audience size) (Rennhoff and Wilbur 2012; Savage and Waldman 2013). Further, the literature offers inconsistent conclusions about market forces. Some scholars find that media consolidation and ownership structure affect news content (Brocas, Carrillo, and Wilkie 2011), while others find tentative or non-existent effects (Erb 2010; George and Oberholzer 2010; Vavreck, Jackman, and Lewis 2010). Together, these factors suggest the need for additional research.
6
Second, the manner and extent to which declining competition and media consolidation
affect news content are most likely to be consequential in local political coverage, which remains the
dominant source of political information in House elections. Despite changes to the media
environment, the vast majority of the information available to voters during congressional
campaigns comes from local print media (Graber 2010; Vinson 2003). Unlike presidential or
statewide contests, there are few alternative outlets, such as blogs or social media, to provide
information about local elections.7 Only with a systematic study at the local level can we test this
Competitiveness and Consolidation Hypothesis and determine how large-scale changes in electoral
competition and the news industry affect the information environment in which voters evaluate and
select their representatives.
Third, the relative paucity of local news sources means that changes to media content may
have different effects on citizen engagement than what previous work at the national level has
found. A long line of research has shown that increased levels of news attention promote citizens’
political knowledge and stimulate their political participation (e.g., Althaus and Trautman 2008;
Barabas and Jerit 2009; Delli Carpini and Keeter 1996; de Vreese and Boomgaarden 2006; Eveland
and Scheufele 2000; Gentzkow 2006; Gentzkow, Shapiro, and Sinkinson 2011; Jerit, Barabas, and
Bolsen 2006; Larcinese 2007; Leighley 1991; Mondak 1995; Nicholson 2003). But dramatic changes
to the media environment in recent years have raised the prospect of an evolving relationship
between political news and citizen engagement.
Prior (2007) argues that in a “high-choice” environment – one characterized by access to an
expanded array of media choices – people interested in politics will consume large amounts of
7 This many seem anachronistic in the digital age. But there is very little coverage of individual congressional campaigns, or other local races, in national outlets like Fox News and the New York Times, and the audiences for political information in many newer venues remain very small. For instance, blog readers constitute just a fraction of the public (Lawrence, Sides, and Farrell 2010), fewer than one in five Americans are on Twitter (Smith and Brenner 2012), and just one-third of social media users say that such sites are “very” or “somewhat” important for learning about politics (Rainie and Smith 2012). Moreover, local print coverage has been found to affect voter attitudes toward members of Congress, while local television has not (Schaffner 2006).
7
political news. With non-stop access to national political coverage on cable television and the
internet, political junkies can indulge their interest in ways never before possible. At the same time,
the less politically interested can more easily avoid political information altogether, spending their
free time not with the news, but with entertainment or sports programs. As a consequence, the
contemporary media environment contributes to inequality in political knowledge and participation
(Prior 2005). The information-rich get richer, and the information-poor get poorer.
Such a dynamic certainly seems plausible at the local level. Declines in the volume and
substance of news content in U.S. House races could prove most harmful to the knowledge and
participation levels of the less politically engaged. Those who are more politically interested may be
more insulated from the effects of these declines in news because they may be able to acquire
relevant information from other sources. Because the effects of an impoverished information
environment would be limited to the least politically aware, we call this the Poor Get Poorer Hypothesis.
But relatively little work has studied coverage of local contests, so it is not empirically
established that news coverage will stimulate knowledge and participation in House elections above
and beyond the effects of electoral context and individual-level characteristics. Indeed, we expect to
find a different pattern. Alternative sources of political information are so limited at the local level
that we anticipate that the knowledge and participation of people at all levels of awareness will be
affected. A reduction in the quality of political news will, of course, lower engagement among people
who are not especially interested in politics. But declines in news content will also affect the
knowledge and participation levels among the more politically aware. In an analysis of 1,074 local
news and information sources in the top 100 U.S. television markets, Hindman (2010) reports that
less than 2% of local news websites are unaffiliated with traditional print or broadcast media. In
other words, virtually all of the local political news available to consumers comes from mainstream
news organizations, like newspapers. Unlike with national politics, there are virtually no other outlets
8
to which people can turn for information about local politics. Thus, variations in local newspaper
coverage are likely to have similar effects among both the most and least politically aware citizens.
We call this the Everyone Gets Poorer Hypothesis.
Taken together, the literatures on congressional elections and political communication
suggest that low levels of electoral competitiveness, coupled with consolidation in the newspaper
industry, will reduce the volume and substance of political news. Because of the dearth of sources of
local news, we expect that the decline in political coverage will reduce knowledge and participation
for everyone, regardless of their levels of political awareness or interest. The extent to which we
uncover support for these expectations underscores the importance of considering how changes to
the media environment in contemporary American politics exert differential effects at the national
and local levels.
Uncompetitive Elections and Media Consolidation: Investigating the Information Environment in U.S. House Races
We rely on a research design that allows us to measure both media content and public
opinion during an election cycle. In doing so, we offer the first nuanced, large-scale assessment of
the extent to which competitiveness and consolidation influence both the news and citizens’ political
engagement. We begin our analysis with an examination of the information environment voters
navigate in House elections and the manner in which it is shaped by electoral competitiveness and
consolidation in the news industry.
To measure the political information environment, we rely on three types of data. First, we
conducted an unusually detailed content analysis of the election coverage in all House districts in the
2010 midterm elections. In each of the 436 congressional districts across the country, we identified
the largest circulation local newspaper that we could access through one of several electronic
databases or the newspaper’s online archives. We collected every newspaper article that mentioned
9
at least one of the two major-party candidates for the House seat and analyzed the content of the
coverage in the month leading up to Election Day (October 2 – November 2, 2010).
We focus on three measures that allow us to assess the substance of political coverage: (1)
the number of articles published about each House race, (2) the share of stories that mentioned both
candidates (in contested races), and (3) the number of mentions of issues in news coverage.8 We
assume that more coverage, coverage that provides information about both candidates, and coverage
that includes attention to issues (as opposed to topics like fundraising or campaign strategy) is likely
to give voters more useful information about their electoral choices, and breed knowledge and
participation. In all, we coded 6,004 news stories, editorials, and op-ed columns. Appendix A
provides a detailed description of the content analysis project and coding procedures, including data
on the newspapers and issues we measured (see Tables A1 and A2).
Second, we collected data that allow us to gauge the effects of electoral competition and
media consolidation on news coverage. To measure competitiveness, we rely on the Cook Political
Report’s classification of each district as of October 5, 2010. Cook rates races on a four-category scale:
safe for one party, likely to be won by one party, leaning toward one party, or toss-up. We expect
that the more competitive a race is, the more likely news outlets will be to publish stories and
provide substantive coverage of the contest. Since the Cook competitiveness variable alone doesn’t
capture all of the features of a race that may make it more or less newsworthy, we also collected data
on whether the race was contested, whether there was an open seat, whether the race featured a
quality candidate, and the total amount of candidate spending; each of these variables taps a
dimension of the electoral environment that might affect coverage.9 We treat these variables largely
8 More specifically, we recorded references to more than 150 issues associated with a candidate. Rather than
conduct the analysis at the story or paragraph level, which is typically done, we carried out our coding at the level of the individual reference. In other words, we account for every time any particular issue was mentioned.
9 We thank Gary Jacobson for providing the candidate quality and campaign spending data.
10
as controls, but they also have the benefit of accounting for other potentially relevant aspects of a
campaign that our main competitiveness variable, the Cook rating, might not.
Turning to media consolidation, we gathered information pertaining to two features of
newspapers that previous research suggests affect the volume and substance of election coverage.
We gathered the newspaper’s circulation size because larger newspapers are more likely than smaller
papers to span more than one district. Accordingly, they are responsible for covering multiple
campaigns and should be less likely to devote coverage to any single race. We also determined
whether the paper was corporately-owned; existing work shows that corporate papers are less likely
than family-owned or independent papers to cover local politics, and to do so less substantively.
Coupled with the Cook rating, these two variables allow us to examine the Competitiveness and
Consolidation Hypothesis.
Finally, we included in the data set contextual information about each congressional district.
We tracked median income, the percentage of college graduates, racial composition, and the
district’s “market convergence.” Market convergence accounts for the overlap between a
congressional district and a media market, which is likely to affect the amount of political coverage
devoted to any one House race.10 In the face of such controls, we can be confident that any
competitiveness or consolidation effects we uncover are not an artifact of district characteristics.
(See Appendix B for the coding and descriptive statistics of the variables included in the newspaper
content analysis.)
We begin by presenting bivariate relationships among our key variables. The data suggest
that both electoral competitiveness and media consolidation do, in fact, affect the political
information environment, and do so in expected ways. The leftmost column of Figure 1 presents the
relationship between the competitiveness of a House election – as gauged by its Cook classification –
10 We thank Hans Noel for the media market data.
11
and the number of stories written about the race, the percentage of articles mentioning both
candidates, and the total number of times issues were mentioned in the coverage. Overall, the
average number of stories per race was 14.4. In districts with contested races, an average of 44.5%
of stories mentioned both candidates. And the average number of issue mentions over a month’s
worth of campaign coverage was 48.6.
In each case, lower levels of electoral competitiveness are strongly associated with a lower
volume of coverage. For example, congressional races rated toss-up saw an average of 26 stories,
districts rated leaning saw 23, and districts rated likely to go for one party received on average 20
stories. But in the 72% of districts rated as safe for one party, the average number of news stories
was just 10. The relationship is similarly robust when we consider the percentage of articles about a
race that mentioned both candidates and the amount of issue content. Taken together, these data
show that competitive races produce more coverage and coverage that is more likely to inform
voters about both candidates and the candidates’ issue positions.
Figure 1 about here
The middle and rightmost panels of Figure 1 illustrate that our key measures of
consolidation are also associated with the volume and substance of political news. Districts served
by newspapers with larger circulations see fewer stories overall, a smaller proportion of articles
mentioning both candidates, and less attention to substantive issues than do those served by small-
circulation papers. For instance, papers with circulations smaller than 100,000 published on average
16 stories about the House race, and newspapers in the 100,000-200,000 range published 15. But the
biggest papers published on average just 10 articles.11 Corporate outlets also provide less coverage of
House races than do newspapers that are independent or family-owned. It is clear, however, that
11 We present the circulation data this way for ease of exposition. The results are similar when we split the
circulation numbers into quartiles, as opposed to increments of 100,000. In the regression models, we rely on raw circulation, not this trichotomized measure.
12
corporate ownership has a smaller effect on news volume and substance than do competitiveness
and circulation size. Even in these bivariate analyses, ownership structure appears to be the weakest
predictor of news content among our variables of interest, while competitiveness has the strongest
effects.
Moving into a multivariate context, we can provide a more refined test of the extent to
which the information environment in House races is shaped by electoral competitiveness and
media consolidation. After all, it is important to determine the independent effects of these twin
forces controlling not only for one another, but also for politically-relevant characteristics of the
congressional race and demographics of the district. Thus, we conducted a series of regression
equations that include such variables. Table 1 presents four models, each of which predicts a
measure of coverage in the House race.12 We restrict our analyses to the 93% of races that saw at
least one news article written about them.13
Consider, first, the effects of competitiveness. In every model, the more competitive the race
is, the more coverage (or substantive coverage) there is. This is true whether we examine the total
number of stories published about a race, the number of stories that mention both candidates, the
number of issue mentions across the campaign’s coverage, or the number of different issue topics
discussed in the coverage. All are indicators of an information environment that could help inform
citizens’ choices, and all are strongly related to the competitiveness of the House race, controlling
12 Models do not include the District of Columbia. Beyond our substantive variables of interest and gauges of
electoral context and district demographics, the models that predict whether both candidates are mentioned, the number of issues mentioned, and the number of different issue domains mentioned each also control for the total number of stories written about the race. Races with more coverage overall provide more opportunities for reporters to mention both candidates and their issue positions.
13 Competitiveness, not surprisingly, affects whether a race received coverage; every contest without coverage
was designated safe by the Cook Report. We find, however, that neither circulation size nor corporate ownership influences whether a paper published at least one story about the race. But with 93% of districts seeing at least some coverage, there is little variation to explain.
13
for a host of other factors that might also plausibly be related to media attention. It is clear that as
competitiveness declines, so does the volume and substance of House campaign coverage.
Table 1 about here
The regression results also indicate that consolidation, measured by circulation size and
corporate ownership, affects news content. Although previous research uncovers competing
findings regarding the consequences of changes in the news industry, our data reveal generally
consistent effects. Among races that generated at least one story, circulation size is inversely related
to the volume and substantive content of coverage. The dearth of newspapers around the country,
in other words, seems to have left many of those that remain with a larger geographic area to serve,
and thus more congressional districts to cover. As a result, the quantity and issue coverage any single
district receives declines. And just as in Figure 1, we find a weaker relationship between corporate
ownership and news content than we do for competitiveness and circulation size. Still, on two
measures – the number of stories mentioning both candidates and the amount of issue content –
corporate ownership does reduce the quality of coverage. Its effects are more uncertain, but
detectable nonetheless.14
In sum, the results – which emerge from more than 6,000 articles and hundreds of local
newspapers, House races, and candidates – could hardly be clearer: The competitive context and the
characteristics of the newspaper serving the district are central drivers of election coverage. This is
the case across a variety of measures of news content. Given this evidence of the extent to which
14 The control variables are informative as well. As is the case with statewide and presidential elections, journalists tend to devote more attention to low-salience races when the contest features quality challengers and high levels of spending. We also confirm previous findings (e.g., Arnold 2004; Cohen, Noel, and Zaller 2004) that the more overlap there is between a media market and a congressional district, the more coverage there is. On the other hand, whether the race is an open seat is consistently insignificant. This suggests that congressional turnover does not automatically generate heightened coverage; that only happens when a race is competitive. Given the number of lopsided districts, congressional retirements do little to invigorate the information environment.
We also performed the analyses controlling for the year the incumbent was elected, since long-serving incumbents could plausibly receive more coverage. The variable, however, does not generate consistent effects, and the number of cases drops considerably since all open seats are omitted. The age of the newspaper also does not affect the volume or substance of coverage, nor does it matter whether there had been a contested primary before the general election. The effects of competitiveness and media consolidation are generally unchanged in these models.
14
electoral competitiveness and consolidation in the newspaper industry shape the volume and
substance of political news in House races, we can now turn to an examination of whether and how
this coverage affects voters’ political knowledge and engagement.
The Effects of the Information Environment on Civic Engagement
Does a reduction in campaign coverage lead to lower levels of political knowledge and
participation? And are any effects concentrated among the least politically attentive citizens, or do
they also emerge even among people who are more politically-interested? Only by examining the
effects of news content on civic engagement can we discern how electoral competiveness and media
consolidation shape the prospects for democratic accountability in contemporary American politics.
To do so, we analyze data from the 2010 Cooperative Congressional Election Study.15 We
rely on several measures of political knowledge and participation: whether respondents (1) could
offer a rating of their incumbent House member, (2) place the Democratic House candidate in their
district on the ideological scale, (3) place the Republican candidate in their district on the ideological
scale, and (4) offer a vote intention in the pre-election survey. We expect that, all else equal,
individuals living in congressional districts with more news coverage will be more informed about
their incumbent and the House candidates in their district, and more likely to participate than will
people living in districts where there is less news coverage. Any support we uncover for this
expectation is evidence that the decline of competitiveness and the rise of media consolidation
diminish civic engagement indirectly – that is, by reducing the political information to which citizens
have access in the local media.
We employ two key independent variables to test our hypotheses. First, we model the news
environment with a measure of the overall volume of coverage (N of stories) devoted to the House
15 The CCES is a collaborative survey among dozens of academic institutions, conducted by
YouGov/Polimetrix. Details about the survey design, sampling, and other technical information are available at http://projects.iq.harvard.edu/cces/.
15
race in each district. This is not only the most comprehensive measure of the news environment
included in our content analysis, but it also captures multiple aspects of the information to which
citizens have access. As we showed in Table 1, for instance, districts with more news coverage
overall also saw more substantive coverage.16 Second, we measure respondents’ news consumption
patterns. People who consume a lot of media are likely to have higher levels of civic engagement
than those who consume less. Thus, we created from a battery of CCES questions a High Media
Use variable, which indicates whether the respondent consumed news from three or more different
sources.17 This variable’s key function however, is more than that of a control; it allows us to
determine whether the effects of news coverage are different for people with different levels of
media use. Interacting High Media Use with the news coverage measure provides an opportunity to
test the Poor Get Poorer and Everyone Gets Poorer hypotheses.
Table 2 presents the results of a series of logistic regression equations that predict
respondents’ political knowledge and participation. In addition to the volume of campaign coverage
in the district and the respondents’ reported levels of media use, the models include variables that
capture several features of the electoral environment that could affect engagement levels, as well as
well-known individual-level correlates of civic engagement, such as strength of partisanship and
education. With this extensive battery of controls, we are stacking the deck against finding news
effects. (See Appendix C for a description of all of the variables included in the CCES analysis.)
16 Using more specific variables, such as the amount of issue coverage, would allow us to discern whether
certain types of news content have differential effects on engagement. But given that our present interest is to determine whether there is a general link between news content and engagement, we set aside those more specific questions for future work.
17 While self-reported media consumption variables are less than ideal proxies for political attentiveness (e.g.,
Price and Zaller 1993), we use these measures because they give us the ability to measure political attentiveness in a way that can be connected to local news coverage. People who are attentive to politics are likely to be heavier news consumers than those who are less attentive. In addition, we obtain similar, although not identical, results when we use alternative measures, such as measures constructed from knowledge scales.
16
Before turning to our main results, we need to deal with the possibility of endogeneity; a
correlation between news coverage and civic engagement does not necessarily mean that the former
causes the latter. It could, in fact, be that news outlets in districts where people are highly politically
engaged will devote more coverage to public affairs. In other words, a relationship between the
engagement variables and the volume of campaign news coverage could simply reflect the fact that
newspapers cater to their market, not actually cause engagement to rise or fall. To address this
possibility, we begin with placebo tests in which we regress on our independent variables two
measures of general political knowledge unrelated to the campaign: whether a respondent knows
which party controls the House of Representatives and the respondent’s ability to provide a rating
for Congress. If our news stories variable is not significantly correlated with these two measures of
general knowledge, then we can have confidence that news volume is not merely a consequence of
district-level political knowledge. These two items are also useful because they vary in their difficulty
– 90% of respondents offered a rating of Congress, but just 70% knew that the Republicans
controlled the House of Representatives.
The first two columns of Table 2 present the results of the placebo tests. In both cases,
Number of Stories fails to predict correct answers, so it appears that news coverage is not merely a
product of district-level political knowledge. Not surprisingly, media use – which taps attentiveness
to public affairs information generally – and the standard demographic variables are strong
predictors. Thus, these results validate our measures of broad political knowledge.
Table 2 about here
Having ruled out reverse causation, we can move on to the specific campaign-related
measures of engagement. The models presented in the final four columns of Table 2 reveal that the
volume of coverage devoted to a race affects a respondent’s ability to rate the House incumbent, as
well as place the Democratic and Republican candidates on the ideological spectrum. Further, the
17
number of stories written about a race affects a respondent’s likelihood of expressing a pre-election
vote intention.18
How large is the effect of campaign coverage? We would characterize it as meaningful, if not
dramatic. This is in part because of relatively high levels of engagement among the samples in the
CCES. For instance, 81% of respondents offered an assessment of their House incumbent. Holding
the continuous variables in the model at their means and ordinal variables at their modes, we find
that a shift of two standard deviations in coverage (about 26 stories) increases the likelihood of
rating the incumbent by about 1.5 points. Larger shifts in news coverage, of course, produce larger
increases. The strongest effects come on the likelihood of placing the Democratic and Republican
candidates on the ideological scale (which about two-thirds of respondents could do). A positive
shift of two standard deviations in coverage increases the likelihood of rating the Democratic
candidate by about 2 points, and the likelihood of rating the Republican candidate by about 2.5
points. Finally, a shift of two standard deviations in coverage produces a 1.5 point change in a
respondent’s probability of expressing a pre-election vote intention.
To be sure, these effects are not enormous. But we would not expect them to be. News
coverage is just one of many forces that promote or inhibit the acquisition of political knowledge
and political participation. The individual-level attributes and the various measures of electoral
context clearly affect engagement. But in the face of these stringent controls and relatively high
knowledge and participation levels among CCES respondents – which leave even less room for the
news to matter – the presence of media effects is remarkable. The finding underscores that the
media environment has an important, independent effect on citizen engagement.
18 The results are the same when we restrict the analysis to districts with an incumbent seeking re-election, and
when we control for the length of time the incumbent has served The only exception is that we lose the statistical significance on volume of coverage in the vote intention model.
18
The second part of our analysis examines the relationship between news volume and media
use to test how the information environment affects the engagement of people with different levels
of attentiveness to politics. Table 3 supplements each of the models we presented in Table 2 with an
interaction between the number of news stories and a dummy for whether a respondent was high or
low in media use. If the Poor Gets Poorer Hypothesis is correct, then we would expect the interaction to
be negatively signed. That would indicate that as the amount of news coverage changes, the people
most affected by those variations are those who are least attentive to politics. Thus, if news coverage
falls, it would have the most deleterious effects on those who are less attentive. On the other hand,
if the interaction terms are insignificant, then that would suggest that the effects of variations in
news coverage are similar for people at different levels of attentiveness. Such a pattern would
support the Everyone Gets Poorer Hypothesis, because it would show that declines in news coverage
lower engagement for citizens across the board.19
Table 3 provides no support for the Poor Get Poorer Hypothesis and considerable support for
the Everyone Gets Poorer Hypothesis.20 In none of the four models is the interaction between news
coverage and high media use negative. The non-significant interaction terms in the models for rating
the House incumbent and rating the Republican candidate’s ideology suggest that declines in news
coverage affect both high and low media use citizens identically. In other words, when news
coverage declines, everyone gets information-poorer, not just the people least attentive to politics. In
the Democratic ideology and vote intention models, the interaction terms are positive and
significant. This suggests that variations in news volume have stronger effects on citizens who are
19 A third possibility is that we find positive and significant interaction terms, indicating that declines in news
coverage would have their strongest effects on people high in attention. Combined with significant main effects of the news content variables, this would suggest that all citizens suffer from declines in news coverage, but that those declines will be the greatest for the most politically attentive.
20 When we run the placebo tests with the interaction terms included in the models, neither news coverage nor the interactions is statistically significant.
19
most attentive to politics. A closer inspection of the magnitude of these effects, however, indicates
that the differences in the effects on high and low media use citizens are modest at best.
Table 3 about here
Figure 2 presents predicted probabilities for each of the four dependent variables.21 We plot
lines for high and low media use respondents separately. In each case, the line for high media
citizens is above the line for low media citizens. This reflects the fact that people who consume
more news are more likely to be engaged. In three of the four models, we find no substantive
difference whatsoever in the effect of news coverage on high and low media users. In addition to the
non-significant interaction terms on the coefficients for rating the House incumbent and rating the
Republican candidate’s ideology, it is evident from the slopes of the lines that the effect of news
coverage is virtually identical for citizens at different levels of media use. A comparison of the
minimum-maximum effects underscores how similar the patterns are. In the House incumbent
model, a shift from the most news coverage (80 stories) to the least (no stories) results in a decline in
the likelihood of rating the incumbent by 3.2 points for high media users and 3.9 points for low
media users. For placing the Republican candidate on the ideology scale, the minimum-maximum
shifts in probability are 7.5 and 8.1 points for high and low media users, respectively. Neither of the
differences between high and low media use is statistically significant.
Figure 2 about here
Although the interaction term in the vote intention model is statistically significant, the
fourth panel in Figure 2 reveals that the substantive difference between high and low media users is
not large. The slopes of the lines are virtually indistinguishable, and the minimum-maximum shift in
probability, while statistically significant, is 5.4 points for high media users and 4.3 points for low
21 These simulations are run with continuous variables set at their mean values and dichotomous variables at
their modes. Note that this means these simulations reflect a scenario where an incumbent is in the race (and, in the ideological placement models, when the Democratic or Republican candidate is an incumbent).
20
media users. Strictly speaking, that means that declines in coverage have a stronger effect on high
media users, but that difference is not very large. More to the point, it provides no support for the
Poor Get Poorer Hypothesis and suggests support for the Everyone Gets Poorer Hypothesis. High media use
citizens see their engagement levels drop, but only slightly more than those who consume less
media.
The one significant difference between high and low media users emerges in the Democratic
ideology model. Whereas the minimum-maximum shift in probability for high media users is 9.4
points, it is only 2.7 points for low media users. This finding suggests that declines in news coverage
will shrink the gap between the most and least attentive. After all, the gap in political engagement
between the most and least attentive citizens shrinks from 14 points (at maximum levels of
coverage) to less than 7 points (when there is no coverage of a race). But reducing inequality in
engagement happens only by reducing knowledge among the most highly aware, not by boosting
knowledge among the least attentive. This is hardly a democratically healthy method to close the
gap.
Conclusion
Together, our results form a chain that links electoral competitiveness and media
consolidation to coverage of U.S. House elections and, subsequently, to citizen engagement. When
elections are less competitive, when districts are served by large newspapers, and, albeit to a lesser
extent, when papers are corporately owned, media coverage of U.S. House campaigns is
impoverished. Each factor contributes to less, and less substantive, coverage. This diminished news
environment, then, depresses civic engagement. Citizens in districts with less campaign coverage are
less able to evaluate their incumbent and not as capable of making ideological judgments about the
candidates vying for office. They are also less likely to vote in the House election. These effects
occur for people regardless of their level of political attentiveness.
21
The fact that we find effects for everyone – not just the least attentive – illustrates a critical
theoretical point about the relationship between the changing media environment and citizen
engagement in contemporary American politics. That is, our findings suggest that the consequences
of a “post-broadcast” media environment are contingent on the level of politics we analyze. At the
national level, the proliferation of online news outlets, political blogs, and non-stop cable television
punditry gives the most politically interested citizens virtually unlimited access to political news. This
explosion of news sources, in conjunction with a simultaneous expansion of entertainment options
available to those who are not particularly politically interested, has generated a significant gap in
knowledge and participation among the public. People who care about politics can become more
knowledgeable and engaged than ever before. Those who don’t can opt out almost entirely.
At the local level, however, mainstream news organizations continue to constitute the main
– and sometimes only – source of information about House races and other lower-level contests
(Hindman 2010). When local news outlets like daily newspapers devote less coverage, and less
substantive coverage, to politics, there are few alternative sources to which citizens can turn. There
is no local Politico, no local Talking Points Memo, no local Hot Air. As a result, the quality of
information to which all citizens have access – not just the least attentive – is reduced. That, in turn,
lowers the prospects for political knowledge and participation for everyone. The consequences that
changes to the media environment carry for citizen engagement, therefore, depend strongly on the
availability of alternative sources of information. When such outlets proliferate, we are likely to see a
growing gap between the most and least attentive. But where those sources of information are
sparse, engagement is likely to decline across the board.
Because our results show that engagement depends on the media environment, the question
arises: How can we enrich the news? Numerous scholars and observers have weighed in and
suggested various remedies, such as overhauling journalistic practices (Patterson 2013), boosting
22
reporting resources by establishing partnerships with foundations and non-profits (Downie and
Schudson 2009), and finding innovative benefactors with deep pockets (Huffington 2013). Although
these efforts could certainly improve political journalism, the biggest driver of the decline in the
quality of news is the competitiveness of elections. When elections are uncompetitive, the media
ignore them because they simply aren’t newsworthy. But when contests are closer, they generate
more coverage. Thus, the most effective route to reinvigorating campaign coverage – and thus
improving civic engagement – is likely a renaissance in the competitiveness of House elections.
Because the decline in competitiveness is principally a product of party polarization (e.g.,
Abramowitz, Alexander, and Gunning 2006), however, such a renaissance strikes us as unlikely any
time soon. Given the large body of research that has been devoted to the effects of polarization at
both the elite and mass levels, it is somewhat surprising that the relationship between polarization
and news coverage in congressional elections has been largely overlooked. But our analysis now
makes clear that polarization does more than hinder Congress’s ability to pass legislation (Binder
2003) or generate ill will between partisans in the public (Iyengar, Sood, and Lelkes 2012). By
producing uncompetitive elections that impoverish the political information environment, it also
contributes to a decline in political engagement. In this way, polarization imperils the foundation of
democracy by making it more difficult for citizens to gain the information that would help them
hold their elected officials accountable.
23
Works Cited
Abramowitz, Alan, Brad Alexander, and Matthew Gunning. 2006. “Incumbency, Redistricting, and the Decline of Competition in U.S. House Elections.” Journal of Politics 68(1):75-88.
Althaus, Scott L. 2007. “Free Falls, High Dives, and the Future of Democratic Accountability.” In
Doris A Graber, Denis McQuail, and Pippa Norris (eds.), The Politics of News, the News of Politics. Washington, DC: CQ Press.
Althaus, Scott L. and Todd Trautman. 2008. “The Impact of Television Market Size on Voter
Turnout in American Elections.” American Politics Research 36(6):824‐56. Arnold, R. Douglas. 2004. Congress, the Press, and Political Accountability. New York: Russell Sage
Foundation and Princeton, NJ: Princeton University Press. Barabas, Jason and Jennifer Jerit. 2009. “Estimating the Causal Effects of Media Coverage on
Policy-Specific Knowledge.” American Journal of Political Science 53(1):73-89. Bennett, W. Lance. 2003. “The Burglar Alarm that Just Keeps Ringing: A Response to Zaller.”
Political Communication 20(3):131-8. . 2011. News: The Politics of Illusion, 9th edition. New York: Longman. Binder, Sarah A. 2003. Stalemate: Causes and Consequences of Legislative Gridlock. Washington, DC:
Brookings Institution Press. Brocas, Isabelle, Juan D. Carrillo, and Simon Wilkie. 2011. “A Theoretical Analysis of the Impact of
Local Market Structure on the Range of Viewpoints Supplied.” Washington, DC: Federal Communications Commission. Accessed at: http://www.fcc.gov/encyclopedia/2010-media-ownership-studies (December 3, 2013).
Bruni, Frank. 2002. Ambling into History: The Unlikely Odyssey of George W. Bush. New York: Harper
Collins. Clarke, Peter and Susan H. Evans. 1983. Covering Campaigns: Journalism in Congressional Elections.
Stanford: Stanford University Press. Cohen, Marty, Hans Noel, and John Zaller. 2004. “Local News and Political Accountability in U.S.
Legislative Elections.” Paper presented at the annual meeting of the American Political Science Association. Chicago: September 2 – 5.
Cox, Gary W. and Jonathan N. Katz. 1996. “Why Did the Incumbency Advantage in U.S. House
Elections Grow?” American Journal of Political Science 40(2):478-97. Dalton, Russell J., Paul Beck, and Robert Huckfeldt. 1992. “Partisan Cues and the Media:
Information Flows in the 1992 Presidential Election.” American Political Science Review 92(1):111-26.
24
Delli Carpini, Michael X. and Scott Keeter. 1996. What Americans Know About Politics and Why It
Matters. New Haven: Yale University Press. de Vreese, Claes H. and Hajo Boomgaarden. 2006. “How Content Moderates the Effects of
Television News on Political Knowledge and Engagement.” Acta Politica 41:317-41. Downie, Leonard, Jr. and Michael Schudson. 2009. “The Reconstruction of American Journalism.”
Columbia Journalism Review. Accessed at: http://www.cjr.org/reconstruction/the_reconstruction_of_american.php?page=all (January 26, 2014).
Dunaway, Johanna. 2008. “Markets, Ownership, and the Quality of Campaign News Coverage.”
Journal of Politics 70(4):1193-1202. Edmonds, Rick, Emily Guskin, and Tom Rosenstiel. 2011. “Newspapers: Building Digital Revenues
Proves Painfully Slow.” Washington, DC: Pew Center Project for Excellence in Journalism. Accessed at: http://stateofthemedia.org/2012/newspapers-building-digital-revenues-proves-painfully-slow/ (December 21, 2013).
Erb, Jack. 2010. “Local Information Programming and the Structure of Television Markets.”
Washington, DC: Federal Communications Commission. Accessed at: http://www.fcc.gov/encyclopedia/2010-media-ownership-studies (December 3, 2013).
Eveland, William P. and Dietram A. Scheufele. 2000. “Connecting News Media Use with Gaps in
Knowledge and Participation.” Political Communication 17:215-37. Farnsworth, Stephen J. and S. Robert Lichter. 2011. The Nightly News Nightmare: Media Coverage of U.S.
Presidential Elections, 1988-2008, 3rd edition. Lanham, MD: Rowman & Littlefield. Fleisher, Richard and Jon R. Bond. 2004. “The Shrinking Middle in the US Congress.” British Journal
of Political Science 34(Jul):429-51. Fogarty, Brian J. 2013. “Scandals, News Coverage, and the 2006 Congressional Elections.” Political
Communication 30(3):419-33. Gentzkow, Matthew. 2006. “Television and Voter Turnout.” Quarterly Journal of Economics 121(3):931-
72. Gentzkow, Matthew and Jesse M. Shapiro. 2010. ‘‘What Drives Media Slant? Evidence from U.S.
Daily Newspapers.’’ Econometrica 78(1):35-71. Gentzkow, Matthew, Jesse M. Shapiro, and Michael Sinkinson. 2011. “The Effect of Newspaper
Entry and Exit on Electoral Politics.” American Economic Review 101(7): 2980-3018. George, Lisa and Felix Oberholzer-Gee. 2011. “Diversity in Local Television News.” Washington,
DC: Federal Communications Commission. Accessed at: http://www.fcc.gov/encyclopedia/2010-media-ownership-studies (December 3, 2013).
25
Gershon, Sarah. 2012a. “Press Secretaries, Journalists, and Editors: Shaping Local Congressional
News Coverage.” Political Communication 29(2):160-83. . 2012b. “When Race, Gender, and the Media Intersect: Campaign News Coverage of
Minority Congresswomen.” Journal of Women, Politics, & Policy 33(2):105-25. . 2013. “Voter Reaction to Media Coverage of Anglo, Latino and African American
Congresswomen: An Experimental Study.” Political Research Quarterly 66(3):702-14. Goldenberg, Edie and Michael Traugott. 1984. Campaigning for Congress. Washington, DC:
Congressional Quarterly Press. Graber, Doris A. 2010. Mass Media and American Politics, 8th ed. Washington, DC: Congressional
Quarterly Press. . ed. 2010. Media Power in Politics, 6th edition. Washington, DC: Congressional Quarterly Press. Hayes, Danny. 2010. “The Dynamics of Agenda Convergence and the Paradox of Competitiveness
in Presidential Campaigns.” Political Research Quarterly 63(3):594-611. Hindman, Matthew. 2010. “Less of the Same: The Lack of Local News on the Internet.”
Washington, DC: Federal Communications Commission. Accessed at: http://www.fcc.gov/encyclopedia/2010-media-ownership-studies (December 3, 2013).
Huffington, Arianna. 2013. “Bezos, Heraclitus, and the Hybrid Future of Journalism.” Huffington
Post. August 14. Accessed at: http://www.huffingtonpost.com/arianna-huffington/future-of-journalism_b_3756207.html (January 26, 2014).
Iyengar, Shanto, Guarav Sood, and Yphtach Lelkes. 2012. “Affect, Not Ideology: A Social Identity
Perspective on Polarization.” Public Opinion Quarterly 76(3):405-31. Jacobson, Gary C. 2012. The Politics of Congressional Elections, 8th edition. New York: Pearson. . 2000. “Party Polarization in National Politics: The Electoral Connection.” In Polarized
Politics: Congress and the President in a Partisan Era, Jon R. Bond and Richard Fleisher, eds. Washington, DC: Congressional Quarterly Press.
Jerit, Jennifer, Jason Barabas, and Toby Bolsen. 2006. “Citizens, Knowledge, and the Information
Environment.” American Journal of Political Science 50(2):266-82. Kahn, Kim Fridkin and Patrick J. Kenney. 1999. The Spectacle of U.S. Senate Campaigns. Princeton, NJ:
Princeton University Press. Kirchhoff, Suzanne M. 2009. “The U.S. Newspaper Industry in Transition.” Washington, DC:
Congressional Research Service, July 8. Accessed at: http://digitalcommons.ilr.cornell.edu/cgi/viewcontent.cgi?article=1639&context=key_workplace (December 21, 2013).
26
Larcinese, Valentino. 2007. “Does Political Knowledge Increase Turnout? Evidence from the 1997
British General Election.” Public Choice 131(3-4):387-411. Larson, Stephanie Greco. 1992. Creating Consent of the Governed: A Member of Congress and the Local
Media. Carbondale: Southern Illinois University Press. Lawrence, Eric, John Sides, and Henry Farrell. 2010. “Self-Segregation or Deliberation? Blog
Readership, Participation, and Polarization in American Politics.” Perspectives on Politics 8(1):141-57.
Leighley, Jan. 1991. “Participation as a Stimulus of Political Conceptualization.” Journal of Politics
53(1):198-211. Manheim, Jarol B. 1974. “Urbanization and Differential Press Coverage of the Congressional
Campaign.” Journalism Quarterly 51(4):649-53. Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale University Press. Mondak, Jeffrey J. 1995. “Newspapers and Political Awareness.” American Journal of Political Science
39(2):513-27. Napoli, P. M., & Yan, M. Z. 2007. “Media ownership regulations and local news programming on
broadcast television: An empirical analysis.” Journal of Broadcasting and Electronic Media 51(1):39-57.
Nicholson, Stephen P. 2003. “The Political Environment and Ballot Proposition Awareness.”
American Journal of Political Science 47(3):403-10. Oberholzer-Gee, Felix and Joel Waldfogel. 2009. “Media Markets and Localism: Does Local News
en Español Boost Hispanic Voter Turnout?” American Economic Review 99(5)2120-8. Orman, John. 1985. “Media Coverage of the Congressional Underdog.” PS: Political Science and Politics
18(4):754-9. Patterson, Thomas E. 1980. The Mass Media Election: How Americans Choose Their President. New York:
Praeger. . 1994. Out of Order. New York: Alfred A. Knopf, Inc. Patterson, Thomas E. 2013. Informing the News: The Need for Knowledge-Based Journalism. New York:
Vintage. Price, Vincent and John Zaller. 1993. “Who Gets the News? Alternative Measures of News
Reception and Their Implications for Research.” Public Opinion Quarterly 57(2):133-64.
27
Prior, Markus. 2005. “News v. Entertainment: How Increasing Media Choice Widens Gaps in Political Knowledge and Turnout.” American Journal of Political Science 49(3):577-92.
. 2007. Post-Broadcast Democracy: How Media Choice Increases Inequality in Political Involvement and
Polarizes Elections. New York: Cambridge University Press. Rainie, Lee and Aaron Smith. 2012. “Politics on Social Networking Sites.” Washington, DC: Pew
Research Center. Accessed at: http://pewinternet.org/Reports/2012/Politics-on-SNS.aspx (August 18, 2013).
Rennhoff, Adam D. and Kenneth C. Wilbur. 2012. “Local Media Ownership and Media Quality.”
Information Economics and Policy 24(3-4):231-42. Rosenstiel, Tom and Amy Mitchell. 2011. “The State of the News Media: An Annual Report on
American Journalism.” Washington, DC: Pew Center Project for Excellence in Journalism. Accessed at: http://stateofthemedia.org/2011/overview-2/ (December 21, 2013).
Savage, Scott J. and Donald Waldman. 2013. “Who Values the Media?” In Telecommunications Demand
and Investment: The Road Ahead, James Alleman and Paul Rappoport, eds. New York: Springer Publishing Company.
Schaffner, Brian F. 2006. “Local News Coverage and the Incumbency Advantage in the U.S. House”
Legislative Studies Quarterly 31(4): 491-511. Schaffner, Brian F. and Patrick J. Sellers. 2003. ‘‘The Structural Determinants of Local
Congressional News Coverage.’’ Political Communication 20(1):41–57. Smith, Aaron and Joanna Brenner. 2012. “Twitter Use 2012.” Washington, DC: Pew Research
Center. Accessed at: http://pewinternet.org/Reports/2012/Twitter-Use-2012/Findings.aspx (August 18, 2013).
Snyder, James M. and David Strömberg. 2010. “Press Coverage and Political Accountability.” Journal
of Political Economy 118(2):355-408. Theriault, Sean M. 2008. Party Polarization in Congress. New York: Cambridge University Press. Theriault, Sean M. and David Rohde. 2011. “The Gingrich Senators and Party Polarization in the
U.S. Senate.” Journal of Politics 73(4):1011-24. Tidmarch, Charles M. and Brad S. Karp. 1983. “The Missing Beat: Press Coverage of Congressional
Elections in Eight Metropolitan Areas.” Congress & the Presidency 10(1):47-61. Vavreck, Lynn, Simon Jackman, and Jeff Lewis. 2010. “How the Ownership Structure of Media
Markets Affects Civic Engagement and Political Knowledge, 2006 – 2008.” Washington, DC: Federal Communications Commission. Accessed at: http://www.fcc.gov/encyclopedia/2010-media-ownership-studies (December 3, 2013).
28
Vermeer, Jan Pons. ed. 1987. Campaigns in the News: Mass Media and Congressional Elections. Westport, CT: Greenwood Press.
Vinson, Danielle C. 2003. Local Media Coverage of Congress and Its Members: Through Local Eyes. Cresskill,
N.J.: Hampton Press. Westlye, Mark C. 1991. Senate Elections and Campaign Intensity. Baltimore: Johns Hopkins University
Press. Zaller, John. 2003. “A New Standard of News Quality: Burglar Alarms for the Monitorial Citizen.”
Political Communication 20:109-30. . 1999. ‘‘A Theory of Media Politics: How the Interests of Politicians, Journalists, and
Citizens Shapes the News.’’ University of California, Los Angeles. Typescript.
29
Figure 1: The Relationships among Electoral Competitiveness, Media Consolidation, and House Race Election Coverage, 2010
Note: News data come from a content analysis of local newspaper campaign coverage (6,004 stories overall) in all 436 House districts from October 2 – November 2, 2010. Competitiveness data are from the Cook Political Report, and circulation and ownership data were collected by the authors.
0
30
Sto
ries
Toss-Up Lean Likely Safe
COMPETITIVENESS
0
30
<100K 100-200K >200K
CIRCULATION
0
30
Ind./Family Corporate
CORPORATE
0
100
% M
entio
ning
Bot
h
Toss-Up Lean Likely Safe0
100
<100K 100-200K >200K0
100
Ind./Family Corporate
0
110
Issu
es
Toss-Up Lean Likely Safe0
110
<100K 100-200K >200K0
110
Ind./Family Corporate
30
Table 1: Predicting the Volume and Substance of Newspaper Coverage in House Elections, 2010
Number of Stories
Both
Candidates Mentioned
Number of
Issue Mentions
Number of
Different Issues Mentioned
Competitiveness and Media Consolidation
Cook Rating 0.148* 0.156* 0.158* 0.046* (0.044) (0.035) (0.042) (0.021)
Circulation -0.017* -0.011* -0.017* -0.004^ (0.004) (0.004) (0.004) (0.002)
Corporate Ownership 0.033 -0.128* -0.148^ -0.035 (0.099) (0.072) (0.095) (0.058)
Electoral Context
Open Seat -0.145 0.114 0.076 0.064 (0.138) (0.099) (0.137) (0.071)
Uncontested -0.252 --- -0.804* -0.283* (0.167) (0.230) (0.129)
Quality Candidate 0.249* 0.147* -0.085 -0.167* (0.104) (0.087) (0.116) (0.057)
Candidate Spending 0.008* 0.003* 0.002 0.001 (0.002) (0.002) (0.002) (0.001)
District Features
Percent White -0.002 0.001 0.003 0.004* (0.003) (0.002) (0.003) (0.002)
Median Income 0.007 0.011 0.041 0.000 (0.042) (0.034) (0.050) (0.032)
Percent College Educated 0.016* 0.003 0.012 0.003 (0.006) (0.006) (0.008) (0.005)
Market Convergence 0.126 0.851* 0.779* 0.518* (0.431) (0.205) (0.310) (0.189)
N of Stories --- 0.037* 0.034* 0.016* (0.002) (0.003) (0.002)
Constant 2.297* 0.891* 2.531* 0.839* (0.261) (0.228) (0.288) (0.172)
Log Likelihood -2237.484 -1025.353 -5315.758 -840.221
N 405 380 405 405
Notes: Cell entries are Poisson coefficients. Robust standard errors clustered on newspaper are in parentheses. Levels of significance: * p < .05, one-tailed; ^ p < .10, one-tailed.
31
Table 2: Predicting Political Knowledge and Participation in House Elections, 2010
PLACEBO TESTS
Know House
Majority
Rate Congress
Rate House
Incumbent
Rate Democrat’s
Ideology
Rate Republican’s
Ideology
House Vote
Intention
Information Environment
N of Stories 0.001 -0.002 0.009* 0.006* 0.007* 0.004* (0.002) (0.004) (0.003) (0.003) (0.003) (0.002)
High Media Use 0.967* 1.258* 0.809* 0.723* 0.765* 0.793* (0.054) (0.113) (0.059) (0.042) (0.038) (0.042)
Electoral Context
Competitiveness -0.043 0.072 0.079 -0.007 0.180* 0.012 (0.026) (0.049) (0.051) (0.039) (0.044) (0.030)
Open Seat -0.112* -0.011 0.286* 0.627* 0.311* -0.238* (0.065) (0.113) (0.137) (0.114) (0.107) (0.060)
Uncontested 0.099 -0.006 -0.019 0.343 0.352* -0.560* (0.078) (0.155) (0.151) (0.237) (0.138) (0.091)
Quality Candidate 0.015 -0.088 0.005 0.121 0.221* 0.017 (0.054) (0.103) (0.119) (0.088) (0.081) (0.065)
Democratic Spending 0.004 -0.000 0.009* 0.014* 0.003 0.009* (0.003) (0.005) (0.005) (0.005) (0.004) (0.003)
Republican Spending 0.005* 0.000 0.001 0.005* 0.013* 0.001 (0.002) (0.002) (0.004) (0.002) (0.005) (0.002)
Democratic Incumbent --- --- --- 1.857* --- --- (0.100)
Republican Incumbent --- --- --- --- 1.337* --- (0.088)
Demographics
Strength of Partisanship 0.263* 0.288* 0.162* 0.195* 0.194* 0.559* (0.019) (0.033) (0.024) (0.019) (0.017) (0.020)
Education 0.337* 0.185* 0.120* 0.141* 0.106* 0.119* (0.016) (0.028) (0.021) (0.014) (0.013) (0.013)
Age 0.031* 0.046* 0.041* 0.025* 0.025* 0.027* (0.002) (0.003) (0.002) (0.002) (0.001) (0.001)
Income 0.112* 0.078* 0.056* 0.048* 0.047* 0.079* (0.007) (0.011) (0.007) (0.006) (0.006) (0.006)
White 0.355* 0.245* 0.386* 0.028 0.189* 0.301* (0.053) (0.084) (0.068) (0.057) (0.051) (0.049)
Constant -3.225* -1.348* -2.015* -3.824* -3.432* -2.877* (0.119) (0.172) (0.140) (0.153) (0.151) (0.101)
Pseudo R2 0.172 0.147 0.127 0.196 0.139 0.155 Log Likelihood -25600.793 -11777.579 -19575.655 -27904.346 -29426.527 -27109.436 Chi-square 1791.229 797.462 1221.785 1742.168 1307.279 2133.496
N 44,315 44,224 43,872 41,295 43,281 44,379
Notes: Cell entries are logistic regression coefficients. Robust standard errors clustered on congressional district are in parentheses. Levels of significance: * p < .05, one-tailed.
32
Table 3: Predicting Political Knowledge and Participation in House Elections, 2010,
by Media Use
Rate House Incumbent
Rate Democrat’s
Ideology
Rate Republican’s
Ideology
House Vote Intention
Information Environment
N of Stories 0.008* 0.003 0.006* 0.003^ (0.003) (0.003) (0.003) (0.002) High Media Use 0.739* 0.534* 0.716* 0.713* (0.083) (0.062) (0.056) (0.057) N of Stories x High Media Use 0.005 0.014* 0.003 0.006* (0.005) (0.004) (0.002) (0.003) Electoral Context
Competitiveness 0.080 -0.005 0.181* 0.013 (0.051) (0.039) (0.044) (0.030) Open Seat 0.284* 0.616* 0.309* -0.239* (0.137) (0.113) (0.107) (0.060) Uncontested -0.019 0.341 0.350* -0.559* (0.151) (0.234) (0.138) (0.090) Quality Candidate 0.006 0.123 0.222* 0.017 (0.119) (0.087) (0.081) (0.065) Democratic Spending 0.009* 0.014* 0.003 0.009* (0.005) (0.005) (0.004) (0.003) Republican Spending 0.001 0.005* 0.013* 0.001 (0.004) (0.002) (0.005) (0.002) Democratic Incumbent --- 1.857* --- ---- (0.100) Republican Incumbent --- --- 1.334* --- (0.088) Demographics
Strength of Partisanship 0.162* 0.195* 0.194* 0.559* (0.024) (0.019) (0.017) (0.020) Education 0.120* 0.141* 0.106* 0.119* (0.021) (0.014) (0.013) (0.013) Age 0.041* 0.025* 0.025* 0.027* (0.002) (0.002) (0.001) (0.001) Income 0.056* 0.048* 0.047* 0.079* (0.007) (0.006) (0.006) (0.006) White 0.387* 0.032 0.189* 0.302* (0.068) (0.057) (0.051) (0.049) Constant -2.002* -3.772* -3.416* -2.858* (0.140) (0.155) (0.154) (0.102) Pseudo R2 0.127 0.197 0.139 0.155 Log Likelihood -19575.655 -27904.346 -29426.527 -27109.436 Chi-square 1213.080 1760.798 1392.045 2222.935 N 43,872 41,295 43,281 44,379
Notes: Cell entries are logistic regression coefficients. Robust standard errors clustered on congressional district are in parentheses. Levels of significance: * p < .05, one-tailed; ^ p < .10, one-tailed.
33
Figure 2: The Effects of News Coverage on Political Engagement, by Media Use
Notes: Predicted probabilities are based on the regression equations presented in Table 3. All continuous variables are set at their means and dichotomous variables at their modes.
0.5
0.75
1
80 70 60 50 40 30 20 10 0
Prob
abili
ty
N of Stories
Rate House Incumbent
Low Media Use
High Media Use
0.5
0.75
1
80 70 60 50 40 30 20 10 0
Prob
abili
ty
N of Stories
Place Democratic Candidate on Ideology Scale
0.5
0.75
1
80 70 60 50 40 30 20 10 0
Prob
abili
ty
N of Stories
Place Republican Candidate on Ideology Scale
0.5
0.75
1
80 70 60 50 40 30 20 10 0
Prob
abili
ty
N of Stories
Report Vote Intention
34
Appendix A Newspaper Selection
Very little political science research has sought to analyze media coverage of House elections
from more than a handful of districts. Thus, there is no accepted method of identifying the local
news outlets that serve a particular House contest. To identify the appropriate newspaper for each
House race, we first consulted maps of each congressional district and identified the largest city in
each district. We then determined whether the city had a daily newspaper that we could access
through one of several electronic databases or through the newspaper’s online archives. In the vast
majority of cases, this was a straightforward, though time-consuming, task. In the few cases for
which we could not gain access to newspaper coverage from the district’s largest-circulation daily
paper, we relied on coverage from the largest paper in an adjoining congressional district.
We identified every news story in each congressional district from October 2 through
November 2, 2010 (Election Day) that mentioned at least one of the two major party candidates. We
included in the sample straight news reports, news analyses, editorials, and op-ed columns. We did
not code letters to the editor. We did not restrict the analysis strictly to “campaign” stories because
we assume that any information about the House candidates is potentially relevant for voters. As a
result, our coding includes a comprehensive analysis of the media coverage to which voters could
have been exposed in the lead-up to the election. Our analyses do not include independent and
minor-party candidates.
Table A1 provides an overview of our media data. The figures in the table represent
summary statistics on the circulation size of the newspapers in our sample, the number of stories
about the congressional race, and the length of those stories. The circulation of the newspapers and
the amount of attention to the House race varies quite a bit, as one would expect, given differences
in district composition and competitiveness.
35
Table A1: Summary Statistics on Newspaper Sample and News Stories
Mean
Standard Deviation
Minimum
Maximum
Daily Circulation of Newspaper 195,149 202,379 6,772 876,638 Number of Stories 14 13 1 81 Average Number of Words in a Story 696 274 34 3,275
Notes: Circulation and number of stories reflect data from all 436 congressional districts. Average number of words reflects the 406 districts with at least one story mentioning a major-party candidate. The total number of stories is 6,004. News Content Analysis
Coders read the full text of each article, recording a number of pieces of information. The
key variables for the purposes of this paper were references to issues in connection with a candidate.
(Before undertaking the content analysis, two coders participated in several hours of practice coding,
using news stories from House elections in previous years. This allowed us to refine the coding
scheme and to minimize confusion and maximize consistency between the coders.)
We tracked every time an issue was mentioned, beginning with a list of issues commonly
included in previous studies and then recording references to additional issues as they emerged in
the coverage. We then classified each issue into eight broad categories following previous scholars’
coding schemes: (1) Civil and Social Order, (2) Defense, Security, and Military, (3) Social Welfare, (4)
Taxes and Spending, (5) Foreign Affairs, (6) Race and Social Groups, (7) Government Functioning,
and (8) Economy. Table A2 presents a list of the 173 issues we identified.
36
Table A2: Specific Issue References Coded from News Coverage
“Women’s Issues”
“Men’s Issues”
Defense, Security, Military
Taxes and Spending
Race and Social Groups
abortion, advocacy for women, birth control/contraception, children’s issues (child care), domestic violence, ERA/pay equity, family planning, pornography, women’s health care, women’s issues (not abortion, contraception)
Afghanistan, crime, criminal justice system, death penalty, defense, defense spending, Guantanamo Bay, guns, hunting rights, intelligence, Iran, Iraq, military, national security, nuclear weapons, Pakistan, Patriot Act, public safety, securing the border, security, terrorism, war
Afghanistan, defense, defense spending, GI bill, Guantanamo Bay, intelligence, Iran, Iraq, military issues (bases, benefits, health care, pay), NASA/space, national security, nuclear weapons, Pakistan, Patriot Act, security, veterans’ affairs, war
arts programs, balanced budget, budget/spending, Bush tax cuts, business, debt ceiling, debt or deficit, earmarks/pork, funding for local projects, government size/power, oil subsidies, other program funding, research and development, spending, taxes/tax breaks
advocacy for women, affirmative action, civil rights, Don’t Ask, Don’t Tell, ERA/pay equity, gay rights, marriage equality, Native American issues, race advocacy, racial equality, seniors, workplace discrimination, workplace diversity
Civil and Social Order Social Welfare Economy Foreign Affairs Government Functioning
abortion, alcohol, assisted suicide, bullying, civil liberties, crime, criminal justice system, death penalty, domestic violence, English as the national language, gambling and casinos, guns, hate crimes, hunting rights, illegal drugs, immigration, police or fire funding, pornography, privacy, public safety, religion/religious issues/creationism taught in schools, school prayer, securing the border, separation of church and state, social issues, stem cell research
9/11 workers health plan, birth control/contraception, BP oil spill, cap and trade, children’s issues/child care, climate change, education, energy/electricity/coal/nuclear power, entitlements, environment, family planning, health care/health insurance/Obamacare, homelessness, Medicaid, medical research, Medicare, mining, natural gas, oil drilling, oil pipelines, prescription drugs, school vouchers, social security, social services, student loans, teacher salaries, utilities, water, welfare, wildlife/forests, women’s health, women’s issues (not abortion, contraception), work safety, workers’ compensation
agriculture, auto industry, bailout, banks, business, Cash for Clunkers, consumer protection, credit card reform, economy, ethanol subsidies, farms, federal employee wages, Freddie Mac/Fannie Mae, free enterprise, gas prices, global currency, housing/foreclosures, inequality (economic), infrastructure, jobs, labor, manufacturing, minimum wage, mortgage rates, net neutrality, outsourcing, personal finances, poverty, redistribution of wealth, regulations, retirement, stimulus, TARP, technology, tourism, transportation, unemployment, unions, Wall Street reform
Africa, China, diplomacy, foreign policy, human rights, international issues in health, Israel, Mexico, Middle East, other specific country, spending on foreign aid, trade
campaign finance reform, constitutional amendments, decreasing partisanship in Congress, disaster relief/ FEMA, ethics, FDA, government reform/transparency, insurance reform (not health care), lobbying, PACs, personal scandal, reforms to congressional campaigns, term limits, tort reform, wages for members of Congress and other elected officials
37
Appendix B: Variable Description, Newspaper Content Analysis
Variable
Range
Mean
Standard Deviation
Coding
DEPENDENT VARIABLES Number of Stories
1 – 81
14.41
13.20
Indicates, among districts that received any coverage of the House race, the total number of stories that mentioned at least one of the major-party candidates.
Both Candidates Mentioned
0 – 60
7.65
9.55
Indicates, among contested districts that received any coverage of the House race, the total number of stories that mentioned both major-party candidates.
Number of Issue Mentions
0 – 409
48.61
6.46
Indicates, among districts that received any coverage of the House race, the total number of mentions that substantive issues received in the overall amount of coverage of the House race (see Table A2 for a full list of issues that were mentioned).
Number of Different Issues Mentioned
0 – 8
4.30
2.27
Indicates, among districts that received any coverage of the House race, the total number of issue categories addressed at least once in the overall amount of coverage the House race received. The categories are: (1) Civil and Social Order, (2) Defense, Security, and Military, (3) Social Welfare, (4) Taxes and Spending, (5) Foreign Affairs, (6) Race and Social Groups, (7) Government Functioning, and (8) Economy.
Any Coverage
0, 1
0.93
0.25
Indicates whether the House race received any newspaper coverage (1) or not (0).
INDEPENDENT VARIABLES – Competitiveness and Media Consolidation Cook Rating
0 – 3
0.58
1.03
Indicates the Cook Political Report’s rating of the competitiveness of the race as of October 5, 2010. It varies from safe seat (0) to toss-up (3).
Circulation
0.68 – 87.66
19.51
20.24
Indicates the circulation size of the newspaper (in 10,000s).
Corporate Ownership
0, 1
0.80
0.40
Indicates whether the newspaper serving the district is corporately-owned (1) or independent or family-owned (0).
38
Variable
Range
Mean
Standard Deviation
Coding
INDEPENDENT VARIABLES – Electoral Context Open Seat
0, 1
0.10
0.30
Indicates whether the race is for an open seat (1) or not (0).
Uncontested
0, 1
0.07
0.25
Indicates whether the race is uncontested (1) or not (0).
Quality Candidate
0, 1
0.25
0.44
Indicates whether the race includes a challenger or open seat candidate with previous electoral experience (1) or not (0).
Candidate Spending
0 – 163.43
21.26
18.24
Indicates the total amount of money (in $100,000 units) both major-party candidates, combined, spent on the race.
INDEPENDENT VARIABLES – District Features Percent White
7.90 – 99.20
77.36
20.09
Indicates the percentage of the district that is White (based on 2000 census data).
Median Income
20,924 – 91,571
51,186
12,765
Indicates the district’s median income (based on 2000 census data).
Percent College Educated
6 – 51
21.67
7.31
Indicates the percentage of residents in the district who have earned at least a Bachelor’s degree (based on 2010 Kids Count data).
Market Convergence
0.03 – 1.00
0.16
0.10
A ratio that represents the number of media markets covering any part of the district divided by the total number of districts covered by those markets.
39
Appendix C: Variable Description, Public Opinion Data
Variable
Range
Mean
Standard Deviation
Coding
DEPENDENT VARIABLES Know House Majority
0, 1
0.70
0.37
Indicates whether respondent knows that the Republicans hold a majority of seats in the U.S. House of Representatives (1) or not (0).
Rate Congress
0, 1
0.90
0.19
Indicates whether respondent offered an answer to a question asking him/her to evaluate the way the U.S. Congress is doing its job (1) or not (0).
Rate House Incumbent
0, 1
0.81
0.30
Indicates whether respondent offered an answer to a question asking him/her to evaluate the member of Congress serving the district (1) or not (0).
Rate Democrat’s Ideology
0, 1
0.64
0.48
Indicates whether respondent offered a rating of the Democratic House candidate’s ideology (on the liberal – conservative spectrum) (1) or not (0).
Rate Republican’s Ideology
0, 1
0.64
0.48
Indicates whether respondent offered a rating of the Republican House candidate’s ideology (on the liberal – conservative spectrum) (1) or not (0).
House Vote Intention
0, 1
0.79
0.41
Indicates whether respondent indicated a preference for a candidate in the general election to the U.S. House of Representatives (1) or not (0).
INDEPENDENT VARIABLES –Information Environment N of Stories
0 – 81
14.89
13.47
Indicates the total number of stories that mentioned at least one of the major-party candidates running for the U.S. House in the district.
High Media Use
0, 1
0.43
0.49
Indicates whether respondent reported getting news from three or more sources (1) or fewer (0).
INDEPENDENT VARIABLES – Electoral Context Competitiveness
0 – 3
0.61
1.05
Indicates the Cook Political Report’s rating of the competitiveness of the race as of October 5, 2010. It varies from safe seat (0) to toss-up (3).
Open Seat
0, 1
0.11
0.31
Indicates whether the district features an open seat (1) or not (0).
40
Variable
Range
Mean
Standard Deviation
Coding
INDEPENDENT VARIABLES – Electoral Context (continued) Uncontested
0, 1
0.69
0.25
Indicates whether the race is for an open seat (1) or not (0).
Quality Candidate
0, 1
0.27
0.44
Indicates whether the race included a challenger or open seat candidate with previous electoral experience (1) or not (0).
Democratic Spending
0 – 55.73
11.38
11.70
Indicates the total amount of money (in $100,000 units) the Democratic candidate spent on the race.
Republican Spending
0 – 116.53
10.45
11.48
Indicates the total amount of money (in $100,000 units) the Republican candidate spent on the race.
Democratic Incumbent
0, 1
0.55
0.50
Indicates whether the incumbent in respondent’s district is a Democrat (1) or not (0).
Republican Incumbent
0, 1
0.36
0.48
Indicates whether the incumbent in respondent’s district is a Republican (1) or not (0).
INDEPENDENT VARIABLES – Demographics Strength of Partisanship
0 – 3
2.00
1.06
Indicates how strongly respondent identifies with a political party. Ranges from pure independent (0) to strong party identifier (3).
Education
1 – 6
3.80
1.41
Indicates the highest level of education respondent completed. Ranges from less than high school (1) to post-graduate degree (6).
Age
18 – 91
53.00
14.30
Indicates respondent’s age.
Income
1 – 14
8.32
3.48
Indicates respondent’s annual household income. Ranges from less than $10,000 (1) to more than $150,000 (14).
White
0, 1
0.78
0.42
Indicates whether respondent is White (1) or not (0).