News

It's called the "Bradley effect," named for former Los Angeles mayor Tom Bradley.

In 1982, polls showed the popular black Democrat nine to 22 points ahead of Republican George Deukmejian in the race for California governor. But Mr. Bradley wound up losing.

In that and several other races since, pollsters have suggested a racially charged phenomenon may be at work: To avoid the appearance of bias, a statistically significant swath of white voters hide their true intentions until they step into the voting booth, effectively tainting the polls.

Some pundits say that may explain why the polls were so wrong in the New Hampshire Democratic primary, where Hillary Clinton won a stunning come-from-behind victory over Barack Obama.

"Despite all the talk of how little race matters in this campaign, it is clear that race is still a big deal in biracial campaigns," argued David Kuo, a former policy adviser to U.S. President George W. Bush. "And it has showed up for the first time in a measurable way in the 2008 presidential race."

He isn't alone in that view. The political blogosphere was abuzz yesterday with speculation about the discomforting return of the Bradley effect, which political scientists thought had vanished after a spate of polling anomalies in the 1980s and 1990s.

But anecdotal evidence seems to support the hypothesis that it has returned. A string of at least eight polls in the lead-up to the vote showed Mr. Obama with a commanding lead of anywhere from seven to 13 percentage points, well beyond the margin of error.

The polls, meanwhile, accurately predicted the Republican contest between seven white male candidates. The polls consistently showed Arizona Senator John McCain ahead of former Massachusetts governor Mitt Romney, and that's exactly the way it ended up.

Canadian pollster Tim Woolstencroft of the Strategic Counsel said the notion that voters would deceive pollsters about Mr. Obama because he's black isn't so far-fetched. "We see this pattern in Quebec, where polls overestimate the support for the separatists," he said.

But don't cast New Hampshire voters as closet racists just yet. There are a host of other plausible explanations for Ms. Clinton's late surge. Among them: Pollsters may have overestimated Mr. Obama's bounce from Iowa; Ms. Clinton could have had a better get-out-the-vote machine; independents might have opted late in the game to vote in the Republican primary because they thought the media had already anointed Mr. Obama president; New Hampshire put candidates' names in alphabetical order this year so Ms. Clinton was first on the ballot; women voters overwhelmingly flocked to Ms. Clinton.

Ultimately, experts may conclude that all these factors, like a perfect storm, played a part in lifting Ms. Clinton, the new comeback kid, to victory in New Hampshire.

New Hampshire-based pollster Dick Bennett of American Research Group, who also predicted an Obama win, said the answer to the poll mystery is much simpler. The pundits misinterpreted the impact of Ms. Clinton's unexpected display of emotion in a New Hampshire café, seeing it as a sign of imminent failure, instead of a galvanizing moment for voters, especially women.

Unlike most of his rivals, his polls showed Ms. Clinton with a very late surge.

"We did not have a polling problem; we just ran out of time," Mr. Bennett said.

*****

False expectations

Some reports have been calling Hillary Clinton's New Hampshire primary win one of the biggest upsets in modern U.S. political history, but that's only because the opinion polling gave everyone a false expectation that the results would be different.

It's not the first time this has happened.

FRANCE

French voters were shocked to find that Nationalist leader Jean-Marie Le Pen came second in the first round of the 2002 French presidential vote, putting him on the ballot with winner Jacques Chirac on the second-round ballot. The 12 polls published during the week before the vote had given Lionel Jospin an average of 18 per cent support to Mr. Le Pen's 12.7 per cent. But on election day, Mr. Pen earned 16.9 per cent and the second-round spot, knocking Mr. Jospin out. It was thought that the inaccurate polls actually affected the results, as it gave voters the false impression that the usual second ballot between the top conservative and liberal parties was assured, allowing them to cast their first vote for minor parties.

BRITAIN

Final opinion polls in the 1992 British general election gave the Conservatives between 38 per cent and 39 per cent of the vote, about 1 per cent behind the Labour Party, leading to predictions of a minority parliament. On election day, however, John Major's Tory Party won its fourth consecutive election by more than 7 percentage points. The Market Research Society launched an inquiry into the results, and laid part of the blame on what it called the Shy Tory Factor. It attributed more than 2 percentage points of the error to Conservative Party voters reluctant to reveal their true intentions.

UNITED STATES

The infamous "Dewey Defeats Truman" headline in the Chicago Daily Tribune has its origins in polling. Before the 1948 U.S. presidential election, pollsters were all predicting that Thomas Dewey would beat Harry S Truman. Polling was a relatively new field, and because Gallup had so accurately predicted the 1936 outcome, election watchers had an over-confidence in the young science. The pollsters, however, had stopped surveying a week before the vote, and missed the shift of support away from third-party candidates and toward Mr. Truman.

Staff