Bias Beyond a Reasonable Doubt

THE ONGOING UPROAR over Dan Rather and CBS News has intensified concern about whether the mainstream media have a liberal bias. Some analyses, such as those by the Pew Research Center, document the strong tendency of journalists to describe themselves as liberal. This propensity--also prevalent, alas, among professors--is interesting but does not prove bias in coverage. Reporters might maintain objectivity despite their personal viewpoints, or the conservative leanings of most company owners might offset the liberal inclinations of the journalists.

In this spirit, in February 2003, the former New York Times executive editor Howell Raines said at a meeting of journalists: "Our greatest accomplishment as a profession is the development since World War II of a news-reporting craft that is truly nonpartisan." Paul Krugman went further in his New York Times column of November 8, 2002, when he asserted that the media actually had a conservative slant: "Some of the major broadcast media are simply biased in favor of the Republicans, while the rest tend to blur differences between the parties."

The question is, Who is right? Is there a left- or right-wing bias, or have the media actually managed to be objective? A serious assessment requires quantification of the output put forth by the media. The best analysis I know along these lines is the ongoing study "A Measure of Media Bias," by professors Tim Groseclose of UCLA and Jeffrey Milyo of the University of Missouri.

These researchers use a clever statistical technique to construct an objective measure of conservative or liberal bias in the news coverage of major U.S. television and radio stations, newspapers, magazines, and the Internet. Their main finding is that the liberal inclination of the mainstream media is clear. Among 20 major outlets, Fox News and the Washington Times emerge as conservative, but the other 18 range from slightly to substantially left of center.

Groseclose and Milyo's analytical method begins not with the media but with the voting records of members of Congress. They use the well-known ratings of members' voting records issued by Americans for Democratic Action (ADA), a self-described liberal organization. First, they adjust the members' ADA ratings for the 1990s to ensure comparability over time and between the House and Senate. The ADA score has a 0-100 scale, with 0 meaning that a legislator voted with the ADA 0 percent of the time and 100 signifying 100 percent agreement. The researchers use scores scaled to correspond to the House ratings in 1999. On this scale, the average ADA score for 1995-99 in the House and Senate was 50.1, when senators were weighted by state population, and the District of Columbia was assigned phantom liberal legislators. If members of Congress reflect the views of their constituents, we can view "50" as close to the position of the average voter.

Among well-known liberal senators, John Kerry had an adjusted ADA rating of 88, close to Ted Kennedy's 89. On the conservative side, Bill Frist had 10, whereas John McCain had 13. Results closer to the center were Joe Lieberman's 74, John Breaux's 60, Arlen Specter's 51, and Olympia Snowe's 43.

The next step in the research is to measure the tendency of Senate and House members in their speeches to cite 200 prominent think tanks. Citations considered were limited to those that referred favorably to a view or fact reported by a think tank. For example, the Heritage Foundation was cited favorably by legislators whose average ADA rating was 20, substantially conservative. Also highly conservative were Americans for Tax Reform (19), the Family Research Council (20), the National Right to Life Committee (22), and the Christian Coalition (23). Liberal think tanks included the Center on Budget and Policy Priorities (88), Citizens for Tax Justice (88), the Consumer Federation of America (82), the Economic Policy Institute (80), the National Organization for Women (79), and the NAACP (75).

The last step is to measure the tendency of 20 prominent media outlets to cite favorably the same 200 think tanks. An important point is that the researchers considered only programs or stories labeled "news." They excluded editorials, talk shows, and the like. The idea was to assess political bias in programs billed as news, not the more transparent slant contained in self-identified opinion pieces. The periods assessed ranged from 1990 to 2003. The researchers used these data to calculate, effectively, an ADA rating for each media outlet. The idea is that outlets that refer favorably to conservative think tanks are reasonably viewed as conservative, whereas those that refer favorably to liberal think tanks are plausibly labeled liberal.