The art of prophecy, Mark Twain once said, is very difficult, especially with respect to the future. Yet after months of pre-poll speculation, it’s worth spending a sober moment asking the question: who got it right, and who got it wrong?
First, how did the election-eve polls perform? With the Electoral Commission now estimating the Coalition's two-party preferred vote at 52.5 per cent, the best-performing poll appears to have been the Galaxy poll (52), followed by ACNielsen (54), Newspoll (50) and Roy Morgan (49). The wooden spoon goes to the ANU Online Poll (46), suggesting that we have some time to go before Internet polls are as accurate as phone polling.
One curious fact is that three of the pollsters - Newspoll, Morgan and the ANU Online Poll - were further away from the eventual election result than their reported margin of error. One possible explanation for this is that voters simply changed their minds at the last moment. But recent work by Simon Jackman, an Australian political scientist at Stanford University, indicates that the problem may be that polling margins of error are bigger than we think not the 2-3 per cent that pollsters typically report, but potentially as large as 5-6 per cent. Jackman argues that the best way to deal with this is to poll the polls, combining multiple opinion polls together to determine the true mood of the electorate.
What if we assess the pollsters not on their predicted two-party preferred vote, but simply on whether they picked the winner? Thumbs up to Galaxy and ACNielsen. Thumbs down to Morgan and the ANU Online Poll. Thumbs sideways to Newspoll. Comparing the pollsters' track records over recent elections, newcomer Galaxy now has a perfect record (1/1), as does ACNielsen (2/2). Based on elections since 1987, Morgan is now 3/7, while Newspoll's fence-sitting makes them 4/6 with one no-ball.
What of the betting markets? Since election betting opened in mid-2003, the Coalition has always been firm favourite. Even throughout the first half of 2004, with Latham ahead in many of the polls, the betting markets remained confident that Howard would be returned. By election eve, the betting market was paying $1.20 for a Coalition win, and $4 for a Labor win, suggesting that they thought the Coalition had a 77 per cent chance of victory.
How about the betting odds in the marginals? Writing in the Australian Journal of Political Science after the 2001 election, Justin Wolfers and I noted that the favourites had won in 43 of the 47 seats where Centrebet had allowed punters to place a bet. This time, Centrebet offered betting in 33 marginal seats, and on current returns, the favourite appears to have won in 27/33 instances - somewhat down from the stellar predictive power of the betting market in the 2001 election. Yet among the 33 seats, the betting markets did correctly predict a net gain to the Coalition. Overall, the betting markets seem to have again performed at least as well as the polls.
Because betting odds change daily, we can also use them to track the performance of the parties over the course of the campaign. While Labor was rated a 40 per cent chance to win on the day Howard called the election, it steadily slipped behind during the 6-week campaign. Although Latham's stock rose on the day of the ALP campaign launch, it fell sharply after the Jakarta bombing on September 9, and again in the wake of Howard's speech to Tasmanian timber workers last Wednesday.
How did the pundits go? While most forecast a Coalition victory, few foresaw that Howard would increase his majority.
A Sunday Telegraph survey of ten experts on October 3 found that three thought Latham would win, while seven thought Howard would win, but with a smaller majority than in 2001.
Writing in the Herald on polling day, Alan Ramsey stated firmly, "At the very least Labor will eat into the Government's majority".
Mike Carlton wrote: "This time the Coalition will get back, but it will be trudge and plod." The strong swing to Howard surprised most pollsters and virtually all pundits, though not the betting markets.
Election tragics might turn their eye to the November 2 US presidential election. Since the US election coincides with the Melbourne Cup, it seems only appropriate to look to the bookies. According to the Iowa Electronic Markets, George Bush is favourite. But John Kerry is closing fast, and is now a 45 per cent chance.