Who will vote for Obama

It seems to be all over bar the voting but the final days of a presidential race can sometimes produ

This is the 13th presidential campaign I have followed, as a teenager and

as an adult, and the only previous campaign that generated anything like the same passion and enthusiasm was the first of those: John Kennedy’s in 1960. For many people, including myself, the excitement of this campaign is the prospect of an African-American president who could change the direction of his country, and perhaps the world, after the barren Bush years.

However, we should not allow excitement to mask reality. The Obama-McCain contest has generated a number of myths about America's electorate - and it has also generated the polling evidence to extinguish those myths. What is that evidence? And how far can we trust the polls that tell us that Obama is heading for an emphatic victory?

Here are four myths and the real conclusions that can be gathered from the poll data.

Myth 1: Less-educated white voters have deserted the Democrats because the party chose Obama as its candidate. As our chart shows, it is true that John McCain leads Obama among those Americans who never went to college - but so did George Bush four years ago. In fact there has been virtually no change in this group since 2004. The real point, therefore, is more subtle: Obama has failed to make the kind of inroads with the poorer educated that he has made among graduates. Within that group, for example, he has converted John Kerry's 11-point lead among those with postgraduate degrees into a 28-point lead today - a 17-point shift, double the national average.

The same shift can be detected among those whose top educational qualification is an undergraduate degree, where Bush led by 6 per cent in 2004 but Obama leads by 11 per cent today. Even if Obama has not done as well as he might have hoped among less-educated voters, he has more than offset this by his appeal to the college-educated, who make up 75 per cent of all US voters.

A similar story emerges when the data are sorted by income. Obama has maintained, but not added to, the Democrats' traditionally strong lead among Americans whose household income is less than $50,000 (around £30,000) a year. But he has wiped out the Republicans' previously double-digit leads on higher earners. A further shift to Obama in the final days of the campaign might see him achieve the improbable feat of winning more votes than McCain in $100,000-plus households.

Myth 2: Obama's support has suffered because many women voters, upset by the Democrats' failure to nominate Hillary Clinton, will not vote for him. Not so. Obama has gained more ground among women than among men. In Britain, women are usually slightly more conservative then men. But in US elections the gender gap is different. Men are more likely to vote Republican and women to vote Democrat.

This year, the gender gap, far from narrowing as the result of an anti-Obama female backlash, has actually widened. Obama now leads among women by 14 per cent, compared with McCain's 4 per cent lead among men.

Myth 3: Many Democrats won't vote for Obama. Well, some won't; but some Democrats always defect. In 2004, 11 per cent of Democrats voted for Bush. Almost exactly the same proportion this year are likely to vote for McCain. In contrast, slightly more Republicans, and significantly more independents, are likely to vote for Obama than they did for Kerry. Converts to the Obama banner far outweigh any deserters from the supposedly vulnerable base.

Myth 4: McCain appeals less than Bush to the religious right. In fact, McCain's lead among the four in ten Americans who attend church at least once a week is actually higher than Bush enjoyed four years ago, as is McCain's lead among those who describe themselves as "conservative". Perhaps Sarah Palin has done more than we might expect to shore up the right-wing base.

In contrast, infrequent and non-churchgoers have swung Obama's way, as have liberals and, most importantly of all, the 45 per cent of Americans who describe themselves as "moderate". Among this vital group, Kerry's 9 per cent lead four years ago has been transformed into a 26 per cent lead for Obama.

On the other hand, some parts of the conventional wisdom are true. Obama really has won over younger voters. Four years ago, the minority of under-30s who bothered to vote backed Kerry by a modest margin of 54-45 per cent. Today Obama leads with this group by two to one. His challenge is to convert that latent support into big numbers of real votes.

Luckily, however, Obama can win without a vastly higher turnout among young voters. The next age group up - the 30-44 year-olds, many of whom have mortgages and young families and who are worried by today's financial crisis - have also shifted to Obama in a big way. Four years ago, they backed Bush by twice the national margin; today they put Obama ahead by twice the national average. Here is a huge election-deciding shift among people who can generally be relied on to come out and vote.

The shift among older groups is smaller. Among 45-59 year-olds, the shift is a modest (though still useful) six points; but the over-60s have proved resistant to Obama’s charms. They backed Bush by an 8 per cent margin in 2004, and look like backing McCain by the same margin on 4 November.

So, is it really all over bar the voting? Or could Obama still be denied victory by a late swing,

or inaccurate polls, or both? History provides two examples of the polls getting things badly wrong.

In 1948 the Chicago Tribune announced “Dewey Defeats Truman” – a paper Truman held up with glee when he won

So, is it really all over bar the voting? Or could Obama still be denied victory by a late swing, or inaccurate polls, or both? History provides two examples of the polls getting things badly wrong.

In 1948, the final polls showed the Republican challenger, Thomas Dewey, leading the Democratic president, Harry Truman, by 5 per cent. Initial returns on the night from rural areas seemed to confirm this, leading the Chicago Tribune to announce on the front page of its early editions: "Dewey Defeats Truman" - a paper that Truman held up with glee the following day when it turned out he had won, not lost, by 5 per cent.

One or both of two things went wrong: the pollsters' sampling methods contained a pro-Republican bias; and/or there was a late surge in support for Truman, for the polls finished their final interviews two weeks before election day.

In 1980, the polls predicted a close race between Jimmy Carter and Ronald Reagan - when Reagan in fact won by a 10 per cent margin in the popular vote and a landslide in the electoral college. On that occasion, there probably was a last-minute swing. There was only one television debate between Carter and Reagan, and it took place a week before polling day. Reagan emerged the clear victor and assuaged the doubts of many voters that he was up to the job of president.

This was important, because the fundamentals were not on Carter's side. He was an unpopular president, the US economy was in trouble, interest rates were in double figures, and the US had been humiliated by Iran, which held 52 American diplomats hostage in Tehran. The late swing to Reagan was probably a case of the fundamentals finally asserting themselves.

This year, the fundamentals are on Obama's side, and polling will continue until the last moment. So, unless something extraordinary happens, a clear polling lead for Obama probably will translate into an Obama victory (even if a few voters are telling pollsters that they will vote for an African-American candidate when in fact they won't).

So am I predicting an Obama victory? Yes. Am I confident about this prediction? Yes. Every bit as confident as I was in April 1992 when I said that Neil Kinnock was about to become prime minister.

Peter Kellner is president of YouGov

The 2008 data for this analysis comes from YouGov-Polimetrix. It interviewed 26,000 Americans in order to provide a state-by-state projection for the television network CBS. Its overall voting figures - Obama 49 per cent, McCain 43 per cent - are close to the average of recent polls; the size of the sample has enabled us to look at different groups of voters in detail. The figures are compared with those from the 13,000 interviews conducted by the Edison/Mitofsky exit poll for the TV networks and AP news agency on the day of the 2004 election. I am grateful to colleagues at YouGov-Polimetrix for analysing the data for me and to the Hoover Institution at Stanford University for inviting me to study the presidential election