One of the big problems with polls is that people have no idea what they mean. A few months ago, there was a poll showing that Howard was the most popular PM in the past twenty five years. He got a whopping 35%. Impressive? No, not when you consider that he was the only Liberal PM in the survey. It’d be like doing a survey and asking whether you’d prefer to see Gillard, Rudd, Dreyfus, Shorten or Abbott as PM. One would expect Abbott to win that one. Howard actually received a lower percentage than the Liberal Party did in 2007 election. So, in other words, many Liberals preferred one of the Labor leaders. Yet this was written up as Howard being the most popular PM in the past twenty five years.
The other problem with polls is that they don’t distinguish between the possible, the probable and the impossible.
Ok, let’s imagine I’m preparing a dinner for a hundred people and to give myself a guide for the menu, I use the following poll:
The vegetarian option is liable to be very accurate. Most people who identify themselves as vegetarians aren’t liable to change in between now and the dinner. If it’s particularly appealing, on the night, a few extra may order it, but I’m not likely to need less than the poll estimate. But when it comes to the other three options, there are all sorts of things which may make a difference. For example, news of an outbreak of food poisoning associated with chicken may mean that very few people order the chicken. Similarly, rumours of “mad cow” disease may reduce the number of beef eaters considerably.
These things may change what a person initially indicated. And that’s even before, when presented with the final menu, we discover that the beef comes with brussel sprouts because they’re good for you. (“Yuk, brussel sprouts – think I’ll get the seafood after all!”) Or before the anaphylactics discover that the chicken is covered with a peanut sauce.
When it comes to political polls, the media conveniently overlooks that most elections get decided by the people who only start to pay attention in the final few weeks. So when we start hearing about polls meaning this or that, we should be given more information about the circumstances of the poll. Was there an undecided element, and how large was it? How random was the poll? Did the sample include a range of ages and occupations? What state(s) was it taken in? Were there differences between the states? In determining the result of the Federal Election, these things will be more significant than the actual 2-3% fluctuations that the media obsesses over – even as they acknowledge that a 2% swing is in the potential margin of error, so it may not even exist. I’ve even heard, after an improvement of 1%, commentators hypothesise about the reason for the “improvement”.
Polls can create their own momentum. If everyone else thinks this Government is bad, who am I to disagree? But they can also create a backlash. “I’m not sure that I want to re-elect this Government, but I’m concerned that the Opposition will get too big a majority like in Queensland.” The question for the coming election is which of those is likely to happen.
In short, the poll doesn’t tell us who are the committed vegetarians and who will only be deciding after seeing the menu. We’re encouraged to believe that it’s all done and dusted, that the election campaign counts for nothing and that Julia Gillard should hand Tony the keys to the Lodge. (Barry Cassidy doesn’t believe that she’ll lead Labor to the election. Or was that last week?) But I’d still like to see a poll which asked whether the person was sure of their voting intentions, or whether they were still to make up their mind. Now, THAT might give us something interesting to discuss.