NEW YORK - POLITICIANS fear them, voters mistrust them, but many US opinion polls were impressively accurate in predicting Mr Barack Obama's landslide defeat of Mr John McCain.

Several polling organisations were as giddy with success as Mr Obama fans.

'We were dead on centre. We don't want to be obnoxious about it, but we had a really good year,' said Mr Peter Brown, assistant director of the Quinnipiac University Polling Institute.

While few polls predicted exactly those numbers, the six point spread was matched, or almost matched in seven of 15 pre-election surveys listed on the realclearpolitics.com website.

'If you followed all the polls and didn't cherry pick one or two, then that gave a very good sense of what to expect on election night', said Mr Scott Rasmussen, of rasmussenreports.com.

'Our national tracking poll of 52-46 was exactly right ... It was a much better night than people had feared', he said.

The more polling numbers are broken down state by state, the more variety in performance.

For example, in Ohio, Obama won by a four point margin. Quinnipiac had predicted him winning by seven points, while SurveyUSA showed two points, and Fox News/Rasmussen predicted a tie.

In Florida, Mr Obama won by just over two points. Quinnipiac got this right, as did SurveyUSA, which predicted three points, while Fox News/Rasmussen saw Mr McCain winning by one point.

But this was an election posing unusually tricky problems.

One was not knowing whether Mr Obama's apparent support was skewed by the so-called Bradley effect, in which a respondent claims to back a black candidate to avoid sounding racist, then votes for the white candidate.

Another was determining whether legions of young Obama enthusiasts who voiced support to pollsters would actually vote.

Yet another involved the difficulty of reaching the rapidly growing population that uses only mobile telephones, when polling relies on landline calls.

All the same, the often maligned pollsters avoided major errors.

Even exit polls - which flopped spectacularly in 2004 when they showed Democrat John Kerry seemingly beating President George W. Bush - confidently indicated a hefty Obama win.

Having been badly burnt in 2004, some big media organisations, such as cable news giant CNN, treated the exit polls with huge caution.

But Ms Carroll Doherty, associate director at the Pew Research Center, said that 'excesses and mistakes of the past were corrected' in the exit polls.

And in the pre-election polling, 'a lot of effort' went into adjusting for potential problems, including trying to screen respondents for latent racism.

A mobile phone sample was also factored in to try to overcome that deficiency.

As for the Bradley effect, there turned out to be 'not much of an effect', Ms Doherty said.

'Polling is not a science. It's a combination of a science and an art', Mr Brown said. -- AFP