Do Opinion Polls Tell Us Anything?

By September 18, 2012Politics, Research, Society

Recent polls suggest that Labor is on the up and the Coalition down. Can the Gillard Government actually win another term? The carbon tax is a thing of the past; Labor is now on its positive agenda. And so on.

Maybe so. New polls become news in themselves, but no one ever asks, or is told, exactly what these numbers mean. How accurate are they, anyway? I spent twenty years in the world of survey research, and loved it. At the beginning of that time (1965 to 1985) the methodology was new in Australia, and much of what I was doing was being done for the first time.

The core of the sample survey is the fact that a representative and randomly gathered set of people can for some purposes stand for the whole population. There will be error associated with it, and that is called ‘sampling error’. For a properly collected sample of 1000 the sampling error will be about 2.5 per cent: if 50 per cent of them say that they are going to vote Labor, then the true range of that proportion will be 47.5 to 52.5 per cent — 95 per cent of the time (in the other five per cent of the time, the range will be even greater). My sample was 2054, which didn’t make it twice as accurate, but allowed me to break down the sample into smaller groups and compare the responses from different groups: for example, women in cities compared with women in the bush.

Because I was interested in a lot of aspects of political attitudes and behaviour the interviews took some time, and were done on a face-to-face basis, using professional interviewers. I had selected names from the electoral rolls, with electorates themselves being being grouped and then sampled so that I had the right proportions of rural and urban electorates from each state. I had written to each person explaining what I was doing, and that the work was being done from within a university, and in the public interest. The response rate was greater than 80 per cent.

Those were the days! The first surveys were done in 1967 and 1969, and the third in 1979, when the response rate was not so high. By then many others were in the same field, especially for commercial work about products. People were now much less ready to answer the door, or to take part. Polling shifted to the telephone, but once you do that  immediacy and trust are lost. Recently some polling has moved to the online technique, and that comes with its own new problems.

Sampling error is hardly ever reported, and there is non-sampling error as well. The way questions are phrased, and the order  in which they come, can affect the response. Some people are ‘nay-sayers’, while others are anxious to please, and offer the answer they feel will be most acceptable. Some don’t know, and some don’t care, but will give you an answer anyway, because they think they ought to, or don’t want to appear ignorant. Experience allows those conducting the survey to winnow the responses to some degree, but what you’ve got is what you’ve got.

And there is no federal election going on at the moment, so being asked which party you would vote for is a hypothetical. A poll conducted the day before the election is in the right context, and is much more likely to tap real feelings and real intention. Even then there are still voters who are unsure, or haven’t really thought about it, and will make their mind up once they have a ballot paper in their hand (and some will then leave it blank).

OK, the tutorial is over. What are we learning, if anything? In my opinion there have been enough little events to have moved the proportions in Labor’s favour. The Prime Minister has been on a bit of a promises binge, perhaps to forestall a Rudd challenge, perhaps to move away from the carbon tax. The State Coalition Governments are on the nose because of their cuts to services, and some of that bad odour has drifted over to the federal arena. Ms Gillard is more popular, and Mr Abbott is less popular (not that he has ever been especially popular).

Is it the beginning of a trend that will see the Gillard Government triumphantly returned to power? No one, including me, has any idea. For what it’s worth, I doubt it.

Are we going to see a challenge from within the Liberal Party, since 63 per cent seem to prefer Mr Turnbull and only 30 per cent like Mr Abbott? Well, no. The great Turnbull-likers live within the Labor Party, where 73 per cent prefer Turnbull to Abbott. Among Coalition supporters, though Turnbull is marginally preferred, opinion is much more even.

It is hard for the media to get past the image of a horse race when politics is the subject. Everything is about who will win. Alas, we have perhaps another year of it to endure, unless Ms Gillard decides to go early, while she appears to have a decent chance.

Leave a Reply