Here Lies the ‘Bradley Effect’

When Hillary Clinton defied an ocean of pre-election polling that placed her anywhere from five to 13 points behind Barack

When Hillary Clinton defied an ocean of pre-election polling that placed her anywhere from five to 13 points behind Barack Obama" class="company-link">Barack Obama and won last January's New Hampshire primary, it was described as one of public opinion polling's darkest hours.

Sign Up For Our Daily Newsletter

By clicking submit, you agree to our <a rel="noreferrer" href="">terms of service</a> and acknowledge we may use your information to send you emails, product samples, and promotions on this website and other properties. You can opt out anytime.

See all of our newsletters

So dark that a group of academics and polling professionals spent the last months piecing together a dense study of last year's pre-primary polling, in large part to explain what went wrong in New Hampshire. Released this week, their findings point to an assortment of ho-hum factors—"variations in weighting procedures," "patterns of non-response," and "variations in likely voter models" are among them—that might have produced the botched call. 

Maybe they're right. It's an interesting report. But it mostly misses the point. Rather than devoting exhaustive study to the various technical factors that might have misleadingly inflated Obama's pre-New Hampshire standing, we ought to take a step back and realize that the New Hampshire discrepancies never amounted to the polling/media crisis that they were widely assumed to represent.

To be sure, Clinton's thoroughly unexpected four-point victory was an embarrassment to the pollsters—and to the chorus of reporters and political analysts (present company very much included) who unflinchingly accepted the findings and spent the days before New Hampshire discussing Clinton's imminent demise as a certainty.

But far worse was the hysteria, in the form of pervasive conspiracy theories, intense media introspection and self-flagellation, and baseless campaign narratives that emerged from the shock of Clinton's unlikely triumph.

This began moments after Clinton delivered her victory speech, when Eugene Robinson, appearing on MSNBC, posited that Obama might have been the victim of the so-called "Bradley Effect," by which—supposedly—white voters lie to pollsters when black candidates are running for office, claiming on the phone to be supporters (so as not to appear racist) only to vote the other way in the privacy of the ballot box.

There was some superficial appeal to this thinking—New Hampshire is overwhelmingly white, after all, and hey, something had to be responsible for Obama's shocking loss.

But this ignored the fact that just five nights earlier, tens of thousands of voters in Iowa—a state just as white as New Hampshire—had, in record-shattering numbers, given up their evenings to attend meetings at which they publicly declared their allegiance to Obama, who performed slightly better in the final caucus results than previous polling had suggested. Where was the Bradley Effect, keeping the white Iowan public in their living rooms that night?

More importantly, it ignored the fact that the Bradley Effect really hasn't been documented that well. Supposedly, the disappearance of Los Angeles Mayor Tom Bradley's polling lead on Election Day in 1982 was attributable to such white treachery. But there were other factors at work that day, too. Nor was the Bradley Effect a prominent factor in the 26 years between Bradley's California loss and Obama's New Hampshire defeat. Harold Ford, for instance, a black Democrat running for the Senate in Tennessee in 2006, found no drop-off between his pre-election standing and the final results. None of this means that racism doesn't affect black candidates; it only suggests that citing the Bradley Effect as a vehicle for such racism is dubious.

For the rest of the campaign, clear through the primary season (when Obama comfortably won the most overwhelmingly white states in the union) and all the way until the eve of the general election, the New Hampshire "example" was routinely cited in the press—all because the pre-primary polls had been wrong.

In the end, of course, there was no Bradley Effect, during the rest of the Democratic primaries or in November. And, notably, this week's exhaustive report found no evidence that it had been at work during the New Hampshire primary. (Again, this is not to say that racism didn't cost Obama votes in 2008—it clearly did—but these were not voters who were hiding their opposition to him, which is how the Bradley Effect supposedly works.)

Through all of this, a simple, logical and non-technical, albeit unsexy, explanation for the New Hampshire discrepancies was there for anyone willing to consider history: It was just New Hampshire being New Hampshire.

Polling before the New Hampshire primary is always volatile—extremely so. The main reason for this is that the state's contest comes after the Iowa caucuses. And those caucus results, filtered through obsessive media coverage, invariably alter on a wide scale the perceptions of New Hampshire voters. It's called the Iowa bounce, or as George H. W. Bush put it in 1980 (when he went from trailing in New Hampshire by 19 points to leading by six in the days after scoring an upset win in Iowa): "the Big Mo."

Bush ended up losing the '80 New Hampshire primary to Ronald Reagan in a landslide. The Big Mo evaporated when Reagan staged his "I paid for this microphone!" moment at a Nashua debate. Back then, there were five weeks between Iowa and New Hampshire. After the debate, Bush's numbers did nothing but bleed, until his ultimate loss—by nearly 30 points—was no surprise. Overall, though, his example is indicative of New Hampshire polling volatility—a journey from 19 points down, to six up, to 27 behind, all in the course of a few weeks.

Since '80, the Iowa-New Hampshire window has narrowed considerably. By '88, when Bush again sought the G.O.P. nod, it was down to just eight days. That was enough time for Bob Dole, who scored a decisive victory in Iowa (with Bush, the sitting vice president finishing a humiliating third, behind Pat Robertson) to turn a pre-Iowa double-digit deficit in New Hampshire into a slight lead—33 to 30 percent over Bush, according to an ABC News poll taken in the final three days before the '88 New Hampshire vote. Other outlets found the same result: a slight Dole lead or an outright tie.

The trend seemed obvious, and the press began playing up Bush's demise. "The senator from Kansas spent most of the day stumping the state with the lighthearted assurance of a frontrunner," The New York Times wrote of Dole's campaign activity on the day before the primary. Members of Bush's own campaign (much like Clinton's last year) happily played along, too. As Dole erased Bush's lead within two days of the Iowa result, a Bush aide told The Washington Post: "Six days is not enough time to turn this colossus around if we are sliding down that slippery a slope."

Final result: Bush 38, Dole 28.

But that's the New Hampshire primary: extreme volatility triggered by the Iowa result mixed with a short campaign window (down to just five days in '08). No conspiracy theories necessary.

Here Lies the ‘Bradley Effect’