Label Whores, Take Two:

My recent article on media bias ["Label Whores: Bernard Goldberg may not be wrong about the presence of bias in the media -- he's just wrong that it's 'liberal,'" TAP, May 6, 2002] touched a number of conservative nerves, as people variously disputed or pooh-poohed my finding that the average liberal has a thirty percent greater likelihood of being given a label in the press than the average conservative does.

One response to the piece came from Bernard Goldberg himself, whose bestseller Bias has given wide circulation to the notion that the press define liberals as the mainstream by labeling conservatives far more than they do liberals. In an op-ed piece in the Miami Herald, Goldberg offers two numbers to prove his point about labeling. First, he says that a six-month search of The New York Times showed that the word "conservative" popped up in news stories 1,580 times; "liberal" only 802 times.

Well, but so what? Goldberg didn't bother to check how many of those instances of "conservative" and "liberal" were used as labels of American politicians or interest groups, much less to relativize those numbers to the occurrences of the names of each. For that matter, he didn't even try to screen out occurrences of "conservative" that referred to European political parties, business suits, or investment strategies, not to mention occurrences of "liberal" that referred to loan repayment terms and helpings of gravy. In short, these figures are utterly meaningless.

Goldberg's other number involves one of those specious comparisons that critics of liberal media bias are prone to. In this case, he points out that "the Los Angeles Times ran only 98 stories about the Concerned Women for America and identified the group as conservative 28 times. But The LA Times ran more than 1,000 stories on the National Organization for Women and labeled NOW liberal only seven times."

But that's meretricious, in every sense of the term. Concerned Women for America is a self-identified conservative Christian group (it opposes, among other things, abortion, homosexual adoption, hate-crime legislation, the AmeriCorps volunteer program, and the teaching of "ill-conceived Darwinian theory" in the schools). Whereas NOW makes a point of rejecting explicitly partisan labels -- the appropriate description of the group is "feminist." To insist on labeling it as "liberal" would be to assume that to be pro-choice makes you by definition a liberal, by which criterion Goldberg ought to be equally indignant that the press doesn't use the "liberal" label for Christine Todd Whitman or Tom Ridge.

Goldberg has made a specialty of loaded comparisons like this one -- in Bias, he complains that the media routinely label Rush Limbaugh as a conservative talk show host, but don't label Rosie O'Donnell as a liberal TV talk show host. I mean, Rosie O'Donnell? A serious critic would have chosen an example better suited to making the point -- comparing Limbaugh to someone like Michael Kinsley or James Carville, say. But it's precisely the blatant speciousness of the Limbaugh-O'Donnell comparison that commends it to people like Goldberg and his devotees, who delight in imagining how annoying it will be to people on the other side. You have the sense Goldberg's not really interested in persuading anyone -- this is for the bleachers.

Brent Bozell's column on my TAP article develops this strategy at length. Bozell claims that I ignored studies by the Media Research Center that show discrepancies in the labeling of what he takes to be conservative and liberal groups. For example, he says, newspaper stories on the Competitive Enterprise Institute included a conservative label 28 percent of the time, compared to less than one percent for the Sierra Club, and that Concerned Women for America is labeled far more often than Planned Parenthood.

But those comparisons are as transparently loaded as Goldberg's are. After all, the Sierra Club membership came close to adopting a resolution favoring immigration restriction a few years ago, and Planned Parenthood proudly points out that Peggy Goldwater was the founder of its Arizona chapter. To insist that the press describe these groups as liberal amounts to demanding that it adopt the lexicon of the right on a wholesale basis, like a baseball manager demanding that the team's own fans should determine the strike zone. Again, this one is for the bleachers.

It's notable that Bozell doesn't mention any figures for well-known groups like the Americans for Democratic Action (ADA) or the Center for Justice, who fairly deserve to be labeled as liberal or progressive. As it happens, I did counts for a number of political organizations like these, and if I wanted to play Bozell's game I could point out that ADA and the Center for Justice are labeled far more often than conservative groups like the National Association of Scholars, the Center for the Study of Popular Culture, or the Competitive Enterprise Institute. But that would be misleading -- the fact is that there's a lot of unaccountable variation in the frequency of labeling of groups, with some groups on both sides, like the Heritage Foundation and ADA, being labeled far more than others.

Other responses to my study are worthy of more serious discussion. The blogger Edward Boyd went to the trouble of replicating a part of the study on the last six months of the Nexis "Major Papers" database (probably not the best period to pick, since the coverage of American politics has been decidedly atypical in the months following September 11). Boyd used the ten names that I used in my test set, and found that conservatives on average were labeled as "conservative" about fifteen percent more often than liberals were labeled as "liberal."

Not surprisingly, a few conservative bloggers trumpeted Boyd's results as having "refuted" my claims. But even if Boyd's results were valid, that conclusion wouldn't hold. What Goldberg argued, after all, was that there was a massive disproportion in the labeling of conservatives, which is not the same as a fifteen percent difference. Still, Boyd's result surprised me, since the American papers in the Nexis database are largely the same ones I looked at.

But there turns out to be a very big fly in Boyd's ointment. He himself points to the problem when he notes that the database he used contained some English-language foreign papers that might have skewed the results. In fact, fully 32 of the 80 papers in the database are foreign, ranging from the Sydney Telegraph to the Scotsman, the Tokyo Daily Yomuri, and The Jerusalem Post. And when I ran these searches in the Nexis "non-US news" database, which includes all of the foreign papers in the database that Boyd looked at, it turned out that foreign papers label American conservatives more than four times as often as they label liberals -- possibly because of their point of view, but more likely because "liberal" often has another meaning in foreign contexts and because American conservatives like Jesse Helms, John Ashcroft, and Trent Lott are much better known abroad than liberals like Barbara Boxer, Barney Frank, Tom Harkin, or Paul Wellstone.

That disparity introduces a strong tilt in favor of labeling conservatives into the overall data. In fact, when you correct Boyd's results for the relative disproportion of labels in the foreign papers in the database -- a matter of fairly simple math -- you find that the likely rate of labeling in the American papers in the database favors the labeling of liberals by an 18 percent margin. In short, Boyd's data confirm my own, or at least as best as one can make sense of such a small and noisy sample.

One other point worth mentioning is that Boyd did another search that included not just the labels "conservative" and "liberal," but also the labels "right wing" and "left wing," which increased the disparity in the labeling of conservatives to around 30 percent. Conservative media critics have often claimed that the press uses "right wing" a good deal more often than "left wing," and in this they're absolutely right. In my own data, for example, I found that Jesse Helms was described as "right wing" about thirty times as often as Paul Wellstone was described as "left wing." But if you are going to look at "left wing," you're obliged to look at the other labels the press uses for liberal politicians, as well -- terms like "progressive," "on the left," "leftist," and so on. In my own data, it turned out that these labels were applied to Wellstone slightly more frequently than the analogous labels with "right" were applied to Helms. And when I did some searches in the same database that Boyd used, I found that the inclusion of terms like "progressive," "leftist," and "on the left" would have increased Wellstone's rate of labeling by about fifty percent, and doubled Barney Frank's. In for a penny, in for a pound.

I'm inclined to lay the deficiencies in Boyd's study to a lack of sophistication, rather than to the out-and-out disingenuousness of a Goldberg or Bozell. In any event, all these numbers make it clear that labeling simply doesn't come out the way that Goldberg and others say it does. So it isn't surprising that some conservatives have reacted to my survey by depreciating the labeling differences entirely. As Bozell puts it: "Nunberg found liberal politicians were tagged in 4.8 percent of stories to conservatives' 3.6 percent. There's your 30 percent. Big, big deal." And one blogger characterized the results by saying "A staggering 1.2% difference!. . . Come on guys, one percent?"

That's a patently misleading way to interpret these results, of course. (If the murder rate in your city went from 3.6 to 4.8 percent, would you be reassured by a police chief who explained that the increase was only 1.2 percent?) But the very fact that conservatives are pooh-poohing the labeling disparities suggests that they've decided to back off from this argument and move on to looking at other, more subjective forms of close reading.

That's fair enough, but in this connection I was struck by the fact that none of the critics took on the single most extraordinary result in the data I looked at -- this one involving, not labeling, but the way the press talks about the bias story itself. In the newspapers I looked at, the word "media" appears within seven words of "liberal bias" 469 times and within seven words of "conservative bias" just 17 times -- a twenty-seven-fold discrepancy. (As it happens, the disproportion is about the same in the database that Boyd looked at -- 72 to 3).

Now there's a difference that truly deserves to be called staggering. But how should we explain it? Certainly critics on the left haven't been silent about what they take to be conservative bias in the media, whether in the pages of political reviews or in dozens of recent books. But the press has given their charges virtually no attention, while giving huge play to complaints from the right about liberal bias. That's hardly what you'd expect from a press that really did have a decided liberal bias, and in fact the discrepancy is far greater than anything you could explain by supposing that reporters were merely bending over backwards to be fair -- in that case, after all, you'd expect them to give at least a polite nod to the other side, as well.

The media may not have invented the "liberal bias" story, but people like Goldberg and Bozell couldn't have put it over without their active help.