The formal vote for president takes place, not today, but on December 15 when the electors, formally known as the Electoral College, meet in their respective state capitals and cast their ballots, which are then dispatched to Washington and formally counted before a joint session of Congress on January 6. Only then is the outcome of the election officially proclaimed.

The winner-take-all arrangement at state level translates into an exaggerated margin of victory in the Electoral College, even when the contest for the popular vote is close. This wouldn't matter too much because few Americans take any notice of the electoral vote, but in the past some presidents, Bill Clinton most recently, have used the electoral vote to claim a mandate that they couldn't derive from the popular vote.

Clinton was first elected in 1992 with just 43 per cent of the popular vote and nearly 70 per cent of the electoral vote. Of course, the same distortion can result in a ''runner-up presidency'' where the Electoral College produces a winner (in the electoral vote) who has fewer popular votes than his or her opponent, as it did in 1876, 1888 and 2000. This is a serious distortion of the democratic process if you happen to believe that, in a democracy, the wishes of the majority, or at least the plurality, ought to prevail even when the election is very close in terms of popular votes. The prospect of the winner losing and the loser winning this year is remote because it is more than likely that Obama's lead in the opinion polls will translate into a margin of popular-vote victory sufficient to overcome the risk of a distorted electoral vote. When the Electoral College chose Bush over Gore in 2000 and Benjamin Harrison over Grover Cleveland in 1888, the margin of victory in the popular vote was in both cases less than 1 per cent, and it would probably take a race as close as that this year to produce another runner-up president.

Yet it is still theoretically possible. What counts in an American presidential election is not so much how many popular votes the candidate has, but rather how those popular votes are distributed among the states. If somehow John Kerry could have transferred to Ohio 120,000 of his more than one million surplus votes in California in the 2004 presidential election, then he would have become president notwithstanding Bush's three-million-plus lead in the popular vote.

If this year does produce another runner-up president, it would more likely be McCain than Obama. The polls in the big, safe Democratic states that no one is taking any notice of right now because they are so predictable, like California, New York, Illinois and New Jersey, all show a huge margin of popular vote victory for Obama. In New York, for example, Obama is nearly 30 per cent ahead of McCain and performing significantly better than Kerry did in 2004 or Al Gore in 2000. But, in order to win, the surplus votes that Obama will amass in those big states are not as important as making sure of a popular vote victory, no matter how close, in Ohio or Florida, or in a combination of Colorado, Nevada and New Mexico.

That also dictates the nature of the campaign itself. Candidates inevitably concentrate on winning the popular vote in a handful of close states. Neither bothers with those states that either candidate has in the bag. Obama hasn't personally campaigned in California since August 16 and McCain hasn't campaigned in Texas since he received the Republican nomination at the beginning of September. Most Americans see nothing of the election campaign in their states.

The Electoral College deviates from basic democratic principles in other ways. Because all states are guaranteed three electoral votes irrespective of their population size, votes are not equal. Small states benefit unfairly. One person, one vote, one value doesn't apply in America. A voter in Alaska, for example, has four times as much influence over the electoral vote as a voter in New York. The winner-take-all arrangement means that the choice of a voter who has supported the losing candidate in a state counts for nothing in the Electoral College. In fact, the system effectively transfers the votes of the losing candidate at state level to the national Electoral College total of the winning candidate from that state. The system also allows electors to be free agents when they formally cast their votes on December 15. They can vote for whomever they choose and it doesn't even have to be for one of the recognised candidates. Most electors stick to the candidate on whose ticket they were elected but now and again an ''unfaithful elector'' in search of cheap publicity will show their utter contempt for democratic principles and be a maverick to get their 15 minutes of fame.

Apart from the possibility of a runner-up president being elected, the worst distortion of the democratic principle occurs if no candidate reaches a majority of 270 or more votes in the Electoral College. Then the election would be decided by the House of Representatives with each state having just one vote. Thankfully, that hasn't happened since 1824 and is unlikely to happen with only two serious candidates in the contest.

For those who believe in majoritarian democracy, there isn't much hope of changing this state of affairs. The Electoral College has proved to be remarkably resistant to reform notwithstanding strong public support for a change. Neither are the prospects for reform helped by the attitude of the two major parties. Both seem to think that they have more to gain from the existing system than they would have from the uncertainty of a direct popular vote. Even the events of 2000 didn't move the Democrats to support the abolition of the Electoral College. And if that didn't, an Obama victory this year is hardly likely to be a catalyst for reform.

John Hart is reader in political science at the Australian National University.