Just a few weeks ago Hillary Clinton was the frontrunner for the Democratic nomination to contest the presidential election but that was before the Illinois senator stormed to victory in the Iowa caucus.His campaign seemed to have built up an unstoppable momentum, which was reflected in opinion polls that had him comfortably ahead. So how did the pollsters get it so wrong?
One possible, if unsavoury, explanation is the so-called Bradley effect.
The phenomenon was named after Tom Bradley, the long time mayor of Los Angeles, and describes the difference between what members of the public will say in relation to a black candidate when asked by pollsters and the change in their behaviour when they actually vote.
Bradley, who is black, ran as the Democratic candidate for governor in 1982, but, after polls showed he was consistently in the lead, he was a surprise loser.
It was suggested that voters may have told pollsters they supported the black candidate, because they were embarrassed to admit they were racist, but that when it came to voting in private they supported his white opponent, precisely because he was not black.
The Bradley effect was also cited in 1989 when Douglas Wilder won the contest for governor of Virginia by a razor-thin margin after leading comfortably in all the pre-election opinion polls.
There is even a precedent in a Democratic primary. In 1988 a huge number of Democrats told pollsters they supported the black candidate Jesse Jackson but actually voted for Michael Dukakis.
The same pollsters who predicted Obama's victory yesterday correctly anticipated a victory for John McCain - but there was no black candidate to distort opinions in the Republican primary.
Obama may have triumphed in Iowa but that was a public caucus where there was no opportunity for voters to surreptitiously change their stated intentions in the comfort of a polling booth, as they could in New Hampshire, one of the whitest states in the US.
But before jumping to the conclusion that racist voters cost Obama the primary, there is evidence that contradicts the impact of the Bradley effect in New Hampshire.
In 2006, Democrat Deval Patrick was elected as the first black governor of Massachusetts - which neighbours New Hampshire to the south - and the final result accurately reflected his lead in the opinion polls.
Perhaps more significantly, in seven polls preceding yesterday's vote Obama's share of the vote came in at an average of 38.7%, with Clinton averaging approximately 30%.
Although the final results have yet to be announced it looks like the Illinois senator will poll approximately 36% and his rival 39%.
Those figures would seem to suggest that rather than voters abandoning Obama because of the colour of his skin, people who had never stated their support for him were drawn towards Clinton's campaign.
With another caucus, in Nevada, on January 19 and a primary in South Carolina - where 50% of eligible Democratic voters are black - on January 26, Obama's chances of overall victory, and whether it was the Bradley effect that defeated him in New Hampshire, could soon become a lot clearer.
© Guardian News and Media Limited 2008