A review of Left Turn: How Liberal Media Bias Distorts the American Mind, by Tim Groseclose

The vast majority of conservatives undoubtedly believe the mass media are hopelessly liberal. What we need, however, are not more denunciations of tendentious journalism but more, and better, evidence. Tim Groseclose, a professor of political science at UCLA, has made an earnest and impressive effort to address the central problem in deciding whether there is systematic media bias: finding an objective way of measuring bias, both an individual's and news outlet's. The first task is done by assigning to any politician, news outlet, or oneself a Political Quotient (P.Q.) based on whether you agree, disagree, or have no opinion on a policy question. The questions are those formulated by Americans for Democratic Action (ADA) to rank members of Congress. The book lists ten of these questions and Groseclose's website lists all forty. Here's an example:

On August 26, 2009, the Senate voted on the confirmation of Sonia Sotomayor to be a justice on the Supreme Court. Democrats favored her confirmation 58-0; Republicans opposed it 9-31.

I would have favored her confirmation.
I can't decide.
I would have opposed her confirmation.

For every "yes" answer, you get one point, for every "no" response, zero. The higher your P.Q., the more liberal you are; the lower the score, the more conservative. Among the most conservative politicians, according to this method, are Jim DeMint and Michele Bachmann, and among the most liberal are Nancy Pelosi and Barney Frank. No surprises here.

To gauge the bias of the mass media, Groseclose develops another measure, the Slant Quotient (S.Q.). This is a number that shows how often a news outlet cites one or more of some 200 think tanks. (Groseclose did not count a citation if a reporter either criticized a think tank or gave it an ideological label, but these were rare events.) The S.Q. of a news outlet corresponds roughly to the P.Q. of legislators who cite it.

The bigger the S.Q., the more liberal the news outlet. The most liberal newspaper is the Detroit Free Press (S.Q. = 81.5), the New York Times is very liberal (67.3), and the Wall Street Journal is left of center (55.1). The Journal? Yes, because the news section and the editorial sections of that newspaper are opposed to one another. As one observer put it, the editorial and news sections are "as politically polarized as North and South Korea."

* * *

Groseclose's findings are consistent with other studies of media bias that rely on non-political measures. John Lott and Kevin Hassett made a list of technical economic news reported by the United States Department of Commerce, such as stories about employment, GDP growth, and retail sales. They then looked at the headlines newspapers ran about these stories, discovering that papers are 20 to 40% more likely to print a negative headline if a Republican is in the White House than if a Democrat is there.

Two economists, Matthew Gentzkow and Jesse Shapiro, programmed a computer to construct a list of "politically loaded phrases"—ones not merely descriptive but also connotative—uttered by members of Congress. On the basis of these phrases, Gentzkow and Shapiro counted the use of these phrases by 400 daily newspapers. They assigned the equivalent of a S.Q. to each paper. Among the phrases most used by Democratic members of Congress are "tax cut for the wealthiest," "arctic national wildlife," "oil companies," and "civil rights," while those most used by Republican members included "global war on terror," "death tax," "partial birth abortion," and "illegal aliens." A media outlet received a high S.Q. if it used mostly liberal phrases and a low S.Q. if it used primarily conservative ones. The most liberal newspapers were the New York Times, Los Angeles Times, and Washington Post. The most conservative papers were the Washington Times and the Wall Street Journal.

The slant quotient of major newspapers was almost the same whether one used the Groseclose or the Gentzkow-Shapiro method, with one exception—the Wall Street Journal. But Gentzkow-Shapiro's computer program assembled and counted all of the politically loaded phrases, a large fraction of which came from opinion pages; as a result, the Journal is more conservative than Groseclose, who counted only news stories, reported. Gentzkow and Shapiro were also able to show that this political orientation did not have a lot to do with the ideology of their owners, at least when one person owned several papers. When this occurred, the political outlook of a paper's readers explained about 20% of the political slant that Gentzkow and Shapiro found. By contrast, the political contributions of the owners (a rough measure of their ideology) had no effect on the slant. Groseclose says that whether one uses his method, that of Lott and Hassett, or that of Gentzkow and Shapiro, the conclusion is that the mass media carry stories that are liberal.

* * *

Arguments have been made against the Groseclose approach. One holds that it's a mistake to measure journalistic reliance on the views of think tanks or advocacy groups, because liberal ones have better experts and more influential publicists who sell their projects more successfully than do conservative think tanks. But Groseclose finds no evidence (such as the number of Ph.D.s or publicity experts on a staff) that these things differ among think tanks or give liberal ones more stature than conservative ones.

One could also argue that reporters, despite their personal opinions, voluntarily eschew bias out of a sense of civic and professional duty. Groseclose, however, finds little evidence of this self-discipline in two case studies about media coverage of partial birth abortion and the 2001 Bush tax cuts. In the case of the tax cuts, two statements were factually true: first, the rich would get a bigger reduction in their taxes than the poor but, second, that after the tax cut the rich would pay a larger share of federal income taxes than before the cut. Liberal media (those with a high S.Q.) would overwhelmingly report the first fact while conservative outlets (those with a low S.Q.) would report the second fact. Both facts are true; media bias comes from choosing which to report.

* * *

Does capitalism prevent liberal bias? As Eric Alterman, author of What Liberal Media? (2003), puts it, "You're only as liberal as the man who owns you." If Alterman's claims are true, then all the political views expressed by journalists working for a given outlet would be very similar. But watch MSNBC in the morning (when Joe Scarborough is a featured speaker) and in the evening (when Rachel Maddow talks). No one could possibly say that they have identical views: Scarborough is conservative and Maddow liberal. And the Gentzkow-Shapiro paper shows that reporters do not follow ownership bias; on the contrary, most are hired to reinforce reader bias.

But there is no reason to rest Groseclose's argument on a sample of two TV pundits. In 2000 the Pew Research Center published the results of a survey it did of journalists. Each was asked, "How often, if ever, do you think the journalists you know avoid a newsworthy story because the story would be embarrassing or damaging to the financial interests of a news organization's owners or parent company?" Three percent of national journalists said it was commonplace while 72% said it never or rarely happened. Essentially the same results emerged if you substituted advertisers or other journalists for owners. One of my closest friends, who ran a large media company for many years, complained steadily that changing the politics of what reporters covered required a huge investment of time, resources, and good will that made it almost a waste of time to try. Katharine Graham, the late publisher of the Washington Post, made the same observation. Alterman, to put it bluntly, does not know what he is talking about.

Of course, very little of Groseclose's careful analysis would be important if the media and media bias had no effect on readers and viewers. For many years, scholars of the media claimed precisely that, due to the public's "selective attention." That is, people simply screened out media reports that disagreed with the viewpoints they held prior to turning on the television or opening the newspaper. But good studies have been done of late clearly supporting the view that the media does matter.

One of them, by Stefano DellaVigna and Ethan Kaplan, found that as Fox News became available in areas it had not served before the Republican vote for president increased more in those areas than in ones where Fox was long-established. The vote shift was not large—about 0.43%—but big and reliable enough not to be dismissed as statistical noise. Some Yale scholars (Alan Gerber, Donald Green, and Daniel Bergan) conducted a study that offered a free subscription to the Washington Post (a liberal paper) to some residents of Washington's Virginia suburbs, while to others they offered a free subscription to the Washington Times (a conservative paper). The randomly selected Post subscribers gave 3.8% more of their votes for the Democratic candidate for governor than did the randomly chosen Times readers. Here, as in the DellaVigna and Kaplan study, the media have an effect.

* * *

Many questions have been and will be raised by these findings; social science is a competitive enterprise that relies, for the truth, on finding many studies that arrive at the same conclusion. For example, will Groseclose's way of calculating Political Quotient and Slant Quotient hold true for many different time periods and a different variety of interest-group views? Will the Yale findings in northern Virginia hold true for a larger sample of voters? Until such questions are answered, we can regard the view that the media matters and has a liberal bias as unproven but strongly supported.