Use of Political Labels in Major Newspapers
data revised and updated, 3/22/02; see notes below

[When I came to CBS in New York in 1981], I noticed that we pointedly identified conservatives as conservatives, for example, but for some crazy reason we didn't bother to identify liberals as liberals. ... This blindness, this failure to see liberals as anything but middle-of-the-road moderates, happens all the time on network television. ...

In the world of the Jennings and Brokaws and Rathers, conservatives are out of the mainstream and have to be identified. Liberals, on the other hand, are the mainstream and don't have to be identified . Bernard Goldberg, Bias, p. 57-59.

The following table shows the results of a search done on the words "liberal" and "conservative" within seven words of the names of prominent politicians, public figures, and organizations -- a method that picks out the labeling of political views with better than 85% accuracy. Except where indicated, all searches were done on the Dialog "papersmj" database, which includes the texts of about twenty newspapers, including The New York Times , the Los Angeles Times , the Washington Post, The Boston Globe, the Miami Herald , the San Francisco Chronicle , the Chicago Tribune , and numerous others. The method and results are discussed further below.

NOTES:

1. Cf. Bernard Goldberg, Bias, p. 57: "Robert Bork is the "conservative" judge. But Laurence Tribe, who must have been on the CBS Evening News ten million times in the 1980's. . . is identified simply as a "Harvard law professor." But Tribe is not simply a Harvard law professor. He's easily as liberal as Bork is conservative."

2. Cf Goldberg, Bias, p. 57: "If we [the press] do a Hollywood story, it's not unusual to identify certain actors, like Tom Selleck or Bruce Willis, as conservatives. But Barbra Streisand or Rob Reiner, no matter how active they are in liberal Democratic politics, are just Barbra Streisand and Rob Reiner."

NOTES ON THE METHOD

3/22/02: This table is updated from an earlier version, with a number of ratios recomputed using different search strings. It also adds a few more names and corrects a few errors that appeared in the earlier version .

The database used in the study included all articles from The Denver Post, the New York Times, the Arizona Republic, the St Louis Post-Dispatch, the Detroit Free Press, the Los Angeles Times, the Boston Globe, the Chicago Tribune, the Philadelphia Inquirer, Newsday, the San Francisco Chronicle, the Rocky Mountain News, the Miami Herald, USA Today, The (Portland) Oregonian, the Atlanta Journal-Constitution, the Baltimore Sun, the Christian Science Monitor, the Cleveland Plain Dealer, and the St. Petersburg Times. Some of these files went back as far as 1982, others began as recently as 1992.

The papers used in the survey of "liberal" papers were the Washington Post, the New York Times, and the Los Angeles Times.

Dialog searches were generally done in such a way as to pick up as many combinations as possible of names & titles; e.g., they would have found "Senator Kennedy," "Ted Kennedy," "Teddy Kennedy," "Senator Edward Kennedy" "Edward M. Kennedy" and so forth. Instances of last names alone were not counted, not just because of the potential for ambiguity with common surnames like Lear and Frank, but because partisan labels are usually restricted to the first occurrence of a name in an article, where a full identification would be used (E.g., "Conservative North Carolina Senator Jesse Helms said yesterday,,,."). Care was also taken to use additional distinguishers with common names like John Warner. Rob Reiner was searched on only in records that did not contain either "Archie Bunker" or "All in the Family" so as to eliminate references to "Archie Bunker's liberal son-in-law," and Anthony Kennedy was searched on only in documents that contained the word "court."

Some people have suggested to me that the figures for leaders like Trent Lott are kept low by the fact that their leadership titles are often used. This may be so, but a look at the percentages for other conservatives shows that the generalization is not affected by this factor. As for the other legislators, I avoided searching on "liberals" whose ADA ratings are lower than the Democratic average, like Dick Gebhart, or on "conservatives" whose ADA ratings are higher than the Republican average, like Chuck Hagel. That may not be a reliable guide of someone's politics, but every other standard is even more subjective.

It has also been suggested that the survey would be more accurate if it were restricted only to news stories and omitted editorials and letters to the editor. To check whether this was a factor I ran the counts for six legislators over while exclu ding editorials and letters. Not surprisingly, the frequency of labeling was reduced overall, but the ratio of conservative to liberal ascriptions was almost exactly the same.

I did not tabulate the frequencies ofÝ strings like "right wing," "extreme right," "progressive," "leftist," etc., since there is no easy way to accommodate the obvious differences in connotation among these terms. On the basis of preliminary searches, though, I can safely say that the inclusion of all such labels would not have affected the overall proportions. Thus while the label "right wing" is used far more often than the label "left wing," it is counterbalanced by other labels like "on the left," "progressive," "leftist," and so on. When I compared Helms and Wellstone with regard to all possible labels (no easy task, since it requires hand-sorting to eliminate the extraneous uses of 'right' in phrases like "right to life," "right now," "right away," "he is right," and so on, not to mention the -- much rarer -- instances of phrases like "left the meeting early"), I found that the overall proportion of labeling between the two was almost exactly as it was if only the labels "liberal" and "conservative" were considered.

Searches were done on the strings "conservative?" and "liberal?" so as to pick up plural forms. Occasionally, of course, the relevant citation involved "noise," as when someone writes: "Robert L. Barr (R-Ga.) has spoken to Council of Conservative Citizens. Sen. Trent Lott (R-Miss.) has tried to distance himself from the group." To gauge the effect of such examples, I examined 100 citations by hand, by taking the first 25 hits for each of Lott, Boxer, Wellstone, and DeLay. I found 12 examples in this set that could not be described as ascriptions of political labels. These were evenly balanced between conservatives and liberals. Or to put it another way, 88 percent of the hits did in fact involve ascriptions of political labels, which ensures that the results can be assigned a high level of confidence. Typical hits looked like the following:

Forty-five senators, ranging from conservative Republican Jesse Helms to liberal Democrat Paul Wellstone, backed a stronger measure. . .

Sen. Paul Wellstone (D-Minn.), a liberal who supports some of the same goals. . .

. . . said Sen. Paul Wellstone, D-Minn., an outspoken liberal who often has broken with fellow Democrats.

If the Senate has room for Paul Wellstone "-the liberal Democrat from Minnesota-"it has room for Oliver North. . .

With a sample of more than 100,000 instances of the names of legislators alone, examples of the irrelevant juxtaposition of labels and names do not materially affect the significance of the results.

Note that this method should only be used to make comparisons between figures who are discussed in roughly the same range of contexts in the press. It makes no sense to compare the results for Barbara Streisand with those for Clarence Thomas, for example, since the vast majority of articles that mention Streisand talk about her movies or her records, in which connection her political views are irrelevant. For that matter, it makes no sense to compare Streisand to Charlton Heston, who in recent years is far less often discussed as an actor than as a political figure. But a comparison of Heston with Norman Lear is fair enough. The results are most reliable when the figures or organizations compared are likely to be mentioned in the same range of contexts. We can draw strong conclusions from differences in the frequency of labeling of John Paul Stevens and Clarence Thomas, Barney Frank and Dick Armey, or Michael Kinsley and George Will.

return to Geoff Nunberg's "Fresh Air" pieces
return to "Fresh Air" piece on bias