November 6, 2017

Media Bias (Real and Perceived) and the Rise of Partisan Media



In the age of Fox News Channel, talk radio, and MSNBC, it is tempting to think we have entered an unprecedented age of biased media. But partisan bias is actually the norm in media history. As Jonathan Ladd argues, “The existence of an independent, powerful, widely respected news media establishment is an historical anomaly. Prior to the twentieth century, such an institution had never existed in American history.”

In the nineteenth century, overtly partisan newspapers were the norm. Independent journalism was a successful countervailing movement that brought with it standards of unbiased political coverage. Joseph Pulitzer, the Democratic politician and newspaperman, helped establish the Columbia School of Journalism in 1912 and the Pulitzer Prizes in 1917. The American Society of Newspaper Editors codified an impartiality principle in its 1923 code of ethics: “News reports should be free from opinion or bias of any kind.”

But Republicans have long been skeptical of the ability of an independent press to equally represent their views. Donald Trump’s declaration of any contrary coverage as “fake news” is a more extreme version of a long-standing complaint. As Ladd finds, “criticism of the institutional news media was a defining characteristic” of the conservative movement and Barry Goldwater’s nomination campaign. His supporters later founded the Committee to Combat Bias in Broadcasting and Accuracy in Media to monitor the news for bias. Complaints about media bias were common in subsequent Republican administrations and eventually came to dominate Republican campaigns. One study found that 92% of claims of media bias in the 1988, 1992, and 1996 elections came from Republicans alleging liberal bias.

These complaints were directly tied to the rise of explicitly conservative media as alternatives to mainstream news. Kathleen Hall Jamieson and Joseph Capella, who studied the Rush Limbaugh show for a full year, found that he mentioned media bias every single day. Fox News Channel was created to counter perceived liberal bias, calling itself “fair and balanced” with the motto “we report, you decide.”

Conservative skepticism of the ability of an independent press to be unbiased was a reaction to the basic demographics of journalism. Reporters have been and remain disproportionately Democrats and liberals (compared to the general public). Lars Willnat and David Weaver found that the proportion of Republican reporters dropped from an already-low 26% in 1971 to 7% in 2013. Republicans and conservatives are more common in local television news, but are rarely seen at national mainstream news outlets.

Are the Mainstream Media Biased in Favor of Democrats and Liberalism?

Even if reporters are disproportionately Democrats and liberals, they may be able to produce unbiased news by pursuing objective evidence or balanced coverage. Do they succeed?

Overall, studies examining the content of news coverage have not found consistent evidence of bias favoring Democrats or Republicans across many different elections. There are studies finding Republican bias and studies finding Democratic bias. Others find only media bias toward the frontrunner, regardless of party. Dave D’Alessio examines 99 prior studies of presidential election coverage bias and finds no consistent partisan bias.

In an analysis of 95 Senate elections, Adam Schiffer finds that factors like incumbency, finances, poll standing, competitiveness, the state of the economy, and scandals explain the bulk of variation. But he does find a remaining slight advantage for Democrats in newspaper tone, even accounting for these factors.

Even where careful researchers do find bias, they tend not to find that it favors the same party all the time. Kim Fridkin Kahn and Patrick Kenney examined the positive or negative tone of each article and its coverage of issues and candidate traits and counted the number of criticisms of each candidate. They found that newspaper coverage favors incumbent candidates that the newspaper endorsed on its editorial page. But systematic studies of this kind depend on research choices that affect the results. If one candidate is quoted attacking another candidate, that shows up as a criticism, but it could reflect either an accurate portrayal of the opposing candidate’s message or a choice by the reporter to emphasize the criticism. If a reporter mentions that one candidate has raised more money than another, it reflects both a reality about the race and a choice to mention fundraising over another indicator of support.

One might think that cataloging only the clearest instances of media bias is a solution to this problem, but media bias could be subtle. One study found, for example, that candidates endorsed by a newspaper get better-looking photographs in that newspaper.

The most famous study of media bias, by Tim Groseclose and Jeffrey Milyo, compared citations of interest groups and think tanks by media outlets with citations by members of Congress. They reasoned that an unbiased media outlet would cite material equally from sources favored by legislators in each party and from each ideological perspective. They found that most media outlets disproportionately cited sources favored by Democratic and more liberal members. Their measure of media bias allowed them to compare media outlets with members of Congress on the same ideological dimension.

But another study found that Groseclose and Milyo’s methodology did not produce stable estimates over time. All outlets they studied appeared to be more moderate or conservative in later years. Brendan Nyhan leveled a different critique of Groseclose and Milyo: Democratic office holders and media organizations both cited more neutral experts, rather than favoring liberal perspectives: “Technocratic centrist to liberal organizations like Brookings and the Center on Budget and Policy Priorities tend to have more credentialed experts with peer-reviewed publications than their conservative counterparts.” But conservatives also tend to doubt the ability of academic experts to maintain neutrality for the same reason that they are skeptical of the media: academics are also disproportionately Democrats and liberals. So the media bias debate reflects a broader conflict over whether technocratic expertise can be seen as neutral.

Despite these difficulties, studies of the relative treatment of Democratic and Republican candidates and their affiliated interest groups are relatively easy to conduct because there is a straightforward comparative baseline: balanced coverage of each candidate or each group. But that means the focus has been mostly on partisan biases, especially in election coverage, rather than broader ideological biases in the selection or description of political issues and events.

It may be easier, however, for reporters to treat two general election candidates or two parties in a congressional debate equally than it is to avoid ideological influences on coverage that stem from the reporters’ underlying values. There is no evidence of a reporters’ conspiracy to help the Democrats, but reporters with liberal viewpoints might (even inadvertently) choose different news to cover or frame it differently than does the much smaller number of conservative reporters.  

The Promise of New Measures of Media Bias

State-of-the-art measures of media bias build on Groseclose and Milyo’s strategy of comparing media outlets with members of Congress, but use “big data” approaches geared toward analyzing large sets of texts. This approach tries to capture more subtle biases in the use of language or the discussion of issues, which it might be more difficult for reporters to control.

In 2010, Matthew Gentzkow and Jesse Shapiro used the full 2005 Congressional Record to generate phrases disproportionately used by more liberal or conservative members of Congress and then compare the usage of those phrases across newspapers. This approach sidesteps some of the objections to Groseclose and Milyo’s methodology by going beyond group citations, but it is still dependent on the idea that liberal and conservative legislators both use equivalently biased phrases and that, to be unbiased, reporters should tend to use them in roughly equal measure.

The approach still raises perennial difficulties in interpretation. The phrases disproportionately used by Democrats, according to Grentzkow and Shapiro, include some partisan messaging tropes—such as “workers’ rights,” “nuclear option,” “sniper rifles,” and “privatize Social Security”—but also nonpartisan, generic terms such as “trade agreement,” “American people,” “budget deficit” and “war in Iraq.” Other Democratic phrases, such as “veterans’ health care,” “Congressional Black Caucus,” and “minimum wage” reflect real partisan differences, but these are differences in issue agenda rather than message. Republican phrases similarly include poll-tested partisan language such as “death tax,” “illegal aliens,” “oil for food scandal,” and “personal retirement accounts,” but also neutral language referring to the names of courts and justices. Again, there are phrases that reflect issue priorities but not biased analysis, such as “stem cell” and “government spending.” When reporters use these phrases, it may reflect anything from the neutral language in a paper’s style guide to biased choices of what issues to cover to unfairly repeating one side’s message more often. It is also difficult to distinguish between a reporter echoing liberal or conservative talking points and one side just being more “on message” in their communications with reporters.

Gregory Martin and Ali Yurukoglu recently used the same method to look for biased phrase usage on CNN, Fox News, and MSNBC in 2000, 2004, and 2008. They turned up significant but unsurprising Republican bias on Fox News and Democratic bias on MSNBC. Yet once again, the word lists found to be biased (based on congressional communications) alternatively instill confidence and raise questions. The most “Democratic” phrases included neutral language like “African American,” “Republican leadership,” and “Bush administration,” along with more partisan phrases like “social justice” and “working families.” The most “Republican” phrases included neutral phrases like “federal government” and “new refinery,” alongside such ideologically tinged language as “death tax,” “illegal aliens,” and “limited government.”

In a thoughtful appendix, the authors use several alternative measures to find similar trends. They analyze each year independently and all years together, they use different estimators of ideological position based on the same texts, and they produce an alternative measure: the time given to Democratic and Republican guests on each network. All measures produce the same basic contrast: Fox News is to the right of CNN and MSNBC and Fox News and MSNBC are getting more polarized over time. There is no reason to doubt these basic findings.

The improvement in measurement is quite welcome. But scholars should not lose sight of qualitative differences among media outlets. Regardless of word usage patterns, we should still categorically distinguish between CNN, which goes out of its way to find balanced panels of commentators even if it does not always succeed, and Fox and MSNBC. Similarly, we should be able to separate the journalism displayed on such Fox programs as Special Report with Bret Baier and Fox News Sunday from clearly biased programming such as the Sean Hannity Show. Even if the front page of a nonpartisan newspaper is subtly biased, this is qualitatively different from the bias openly displayed on the editorial page. Losing sight of that difference downplays the change wrought by the rise of openly partisan and ideological media.

Clever Ways to Study Media Bias

Assuming that scholars can measure the relative ideological position of news organizations, it is still more difficult to locate the zero point of unbiased news. As Tim Groeling has pointed out, we do not know the population of real-world events that might neutrally be covered, so it is hard to tell if news judgment is biased or reflects neutrality. Does Trump receive negative coverage because the press hates him, or because he has produced a lot of detrimental news for himself?

Several recent studies of media bias use clever methods to get around these difficulties, but do not find consistent results. One study used an experiment where similar letters to the editor favoring each presidential candidate were sent to different newspapers. The researchers, perhaps surprisingly, found that newspapers were more likely to publish the letter against the candidate that it had endorsed, possibly as a way of balancing the opinion section. But the judgment of the editors may not correspond to the choices of that newspaper’s reporters.

Another useful method is to compare coverage of similar events. One study showed that the same level of approval ratings for different presidents was covered differently by each network, reflecting partisan biases. Another study compared the coverage of similar economic statistics in different administrations, finding that the partisan editorial stance of the newspaper was associated with more positive coverage of the same unemployment rate for the endorsed presidents, but that other economic statistics had no partisan slant. A third study used instances where members of Congress switch political parties to look at changes in coverage for the same members in different parties. It found no significant party differences.

Another difficulty is that even biased coverage could reflect the media outlet’s perception of its audience’s views, rather than its own biases. One useful study found biased newspaper coverage of unemployment based on the party holding the White House, but found that this bias was substantially moderated after a competing newspaper closed. Apparently the newspapers took partisan approaches primarily to distinguish themselves from the other paper in town.

Perceptions of Bias Are Reducing Trust in News and Empowering Partisan Media

Regardless of the academic research, the public, and political professionals, are responding to their own perceptions of bias. William Eveland and Dhavan Shah found that even when Republicans and Democrats watch the same coverage, Democrats believe that it is more favorable for the Republican candidates and vice versa. Another study found that even presenting the same information but varying the label of where it came from (CNN or Fox) also changes the perceptions of bias.

Not surprisingly, then, faith in the news media has been in decline across the board. Yet mainstream media outlets get especially poor trust ratings from conservatives. Because Republicans regularly cite political bias in the mainstream media, their electorate only trusts  explicitly conservative alternatives.

Given the strength of partisanship in contemporary political life, it is understandable that Republicans would be more likely to suspect and react to perceived biases among (disproportionately liberal) reporters. It would be surprising if the political views of reporters did not in any way affect their selection of news stories or their portrayal of events.

Only two out of the hundred largest newspapers endorsed Trump in 2016, so there is no reason to believe that perceptions of bias will be diminishing any time soon. Trump’s constant attacks on the media and invocations of “fake news” mean that the Republican electorate is likely to further reduce its trust in mainstream media and rely even more on explicitly conservative media. Fox News and talk radio have large effects on their audiences, and even on the friends of viewers and listeners who do not themselves watch or listen. They have also changed the behavior of legislators, instilling fears of primary challenges and base backlashes.

Even if Republican skepticism of mainstream media impartiality is justified, the rise of explicitly ideological media as the central information sources for Republicans is still alarming. It is as if Republicans reacted to perceived biases on the front page of the New York Times by deciding to only read the editorial page of the Wall Street Journal.

The die may therefore be cast for increasingly replacing traditional journalism with partisan media. Even as Democrats are more accepting of the mainstream news media, the recent doubling of MSNBC viewership suggests that liberals may eventually succeed at copying the conservative media model. The dominance of independent, trusted, putatively impartial media is not a natural state of affairs. It is a twentieth-century phenomenon that is not guaranteed to survive.

Be careful what you call biased news; complaints about mainstream journalism are empowering outlets that are not even trying to be even-handed.

Matt Grossmann, a Senior Fellow of the Niskanen Center, is associate professor of political science and Director of the Institute for Public Policy and Social Research at Michigan State University. He is a coauthor of Asymmetric Politics: Ideological Republicans and Interest-Group Democrats (Oxford University Press, 2016) and the author of Artists of the Possible: Governing Networks and American Policy Change since 1945 (Oxford University Press, 2014) and The Not-So-Special Interests: Interest Groups, Public Representation, and American Governance (Stanford University Press, 2012).