Opinion pollsters love elections. For a start they get lots of publicity virtually free as they are plastered all over the media. More importantly they get a taxpayer-funded audit on the soundness of their methods and work which they can flog around the commercial world.
People in commerce want to know what people are thinking. Car manufacturers want to know whether people prefer fuel economy over safety, environment over power, comfort over performance and so on. Cosmetic manufacturers want to know whether people care about animal experiments and so on. So they want to know whether the polling company, or its commercial arm, can give them sensible answers for a reasonable cost.
Pollsters can answer this question. In the aftermath of an election, a pollster can point to a poll of voters immediately before an election in which they predicted that X per cent said they would vote Labor and Y per cent would vote Coalition. And also point to the actual result in which X per cent, plus or minus a tad, actually voted Labor and Y per cent, plus or minus a tad, voted Coalition.
For commercial purposes, however, the pollster has to be only roughly right. Selling corporations are only interested in significant trends. If things are fairly line ball they are not interested in changing a whole marketing or manufacturing process. From a commercial point of view, if a pollster can reveal a significant shift in opinion, that will be enough. We are talking 60-40 or 70-30 here.
On the other hand, every election since the two-party-preferred system settled down in the late 1920s has been something less than a 57-43 split. Indeed, any split bigger than 54-46 would always result in a massive landslide to the party that got 54. In 1996, for example, the Coalition’s 53.6 per cent of the two-party preferred vote got it the whopping landslide of 94 seats to Labor’s 49, and 64 per cent of the seats.
The short point is that elections are decided on a few percentage points. A pollster only needs to be a couple of percentage points out to have egg all over his face in voterland. But in commercial land being just a couple of percentage points out is pretty good.
So what is one to make of the hullabaloo in the Murdoch press when the Newspoll shifted a couple of points to the Coalition? Most writers in the Murdoch press and elsewhere frothed like generals re-fighting the past couple of wars. Howard and Coalition always come from behind to win, so here we go again.
Well, a couple of points is within the margin of error of the poll. In a poll like Newspoll you can say that if you repeat the poll 100 times, in 95 cases the results would be within 3 percentage points of the poll you have got.
From a commercial point of view, the polls get it right all the time: about half the electorate votes Labor and the other half Coalition. If you sampled a new product and found only about half of the market preferred it, you would abandon the project. You would not much care if the pollster was out by 5 per cent or so. But in an election, that middle 5 per cent is absolutely critical.
A two-percentage-point shift in one poll means nothing, particularly in these circumstances. In this election, a higher than usual percentage of people have made up their mind. Newspoll is tracking only 5 per cent uncommitted and 1 or 2 per cent refused to answer. A Burson-Marsteller survey of 1156 voters as far back as August showed 77 per cent of voters had firmly decided for whom they would vote. Of them, 56 per cent said they would vote Labor and 34 per cent would vote Coalition.
In that environment, the 2 per cent shift is more likely to be from poll error than from people changing their mind.
What does matter in 2007 is the long and persistent results in favour of Labor for the past 11 months of Kevin Rudd’s leadership. It is like no other election in recent history. Labor is consistently polling more than it needs to win.
But that said, Labor needs 52.1 per cent of the two-party preferred vote to win this election. The Coalition could win it with just 48 per cent on the vote. This is based on the 2004 result and the seats needed to win in 2007.
In 2004 Labor got 47.3 of the two-party preferred vote. It needs 16 seats to win a majority – 76 seats in the House of Representatives. The 15th of those seats on a even swing is John Howard’s seat of Bennelong which seems he will almost inevitably lose. It needs an even swing of 4.8 per cent to win 16. You add the 4.8 to 47.3 and get 52.1 per cent. To get a hung Parliament it needs 15 seats or 3.3 per cent even swing which is 50.6 per cent of the vote.
In 20 of the 23 elections since World War II Labor has needed more than 50 per cent to win and the Coalition less than 50 per cent. The Coalition won in 1954, 1961, 1969 and 1998 with less than 50 per cent of the two-party preferred vote. Labor won 1990 on that basis.
The trouble for Labor in this highly urbanised country, it holds a lot of urban seats with large majorities – wasted vote.
Another reason is incumbency. Governments can be more effective in holding on to marginals than Oppositions. Sitting members have media and postal advantages and the pork barrel. Under Bob Hawke, Labor got its lowest percentage required to win – 47.4 in 1987.
So the system makes it hard to gain Government in Australia. Even so, these polls show that to be almost certain, but perhaps not by such as large margin as the raw figures suggest, given the pro-government bias in the system.