By Mark Davis
And the gong for the most accurate election-eve opinion poll goes to . . . cue drum roll.
Now that the Australian Electoral Commission has finished its count of the two-party preferred vote, we can make a final assessment of how well the pollsters performed in predicting election 2010.
The simplest way of measuring pollster accuracy is to look at the ''absolute error'' – the difference between the pollster's prediction of the share of the two-party vote and the actual outcome.
On election night, the AEC had Labor on 50.7 per cent of the two-party preferred vote and the Coalition on 49.3 per cent. The verdict then was that the Morgan Phone poll and Essential Media's online poll - which had both predicted a 51-49 split in the the two-party vote - shared the bragging rights as the most accurate polls.
But now we have the AEC's final count of the two-party vote - which puts Labor on 50.12 per cent of the two-party vote - it turns out that the most accurate pollster was Newspoll.
Newspoll's election-eve survey of 2500 voters predicted Labor would secure 50.2 per cent of the two-party vote. That gave Newspoll a minuscule error of just 0.08 percentage points. This pushes the Morgan Phone and Essential Media polls into equal second place - they were 0.9 percentage points out on the final two-party result.
The last JWS Research ''robopoll'' before election day came in third place with an error of 1.5 points. Nielsen, Galaxy and the Morgan Face to Face poll came in equal fourth place with errors of 1.9 percentage points.
Newspoll's tiny 0.1 point error was its best performance in a federal election since 1996.
But Essential Media deserves a special mention.
Essential was regarded with scepticism by some pollwatchers because it was a new entrant to the pollster stakes, used online surveys and is a pro-Labor public relations firm.
Yet Essential had not only the second best prediction of the two-party preferred vote, but was the most accurate pollster in predicting first preference votes.
Essential's election-eve prediction of the Labor, Coalition and Greens primary votes had an average absolute error of just 0.3 points, compared to 1.1 points for Morgan Phone, 1.5 points for Newspoll, Nielsen and Morgan Face to Face and 2 points for JWS Research.
Political activists and ordinary voters are often suspicious about polls, doubting the veracity of their findings and the motivations of media organisations who report their results.
Yet election 2010 turned out to be a good year for the pollsters.
The average election-eve error of all seven polls on the two-party preferred vote was 1.3 percentage points. This was the best collective result for the pollsters since 1998.
And each of the organisations accurately predicted the result within the statistical margins of error of their polls - which is all that any pollster purports to do. The table shows that has been the case, with a few exceptions, over the last several federal elections.
Mark Davis is national editor of the Fairfax National Bureau .