‘If an election were held today...’ is probably the most oft-repeated phrase in political reporting after ‘stop the boats’ and ‘people smuggler’s business model’. Weekly polls and the slightest irrelevant movement therein are reported breathlessly by the media and picked over by both sides of politics. They’ve ceased to be a reflection of political views and become instead a driver of them. Even worse, irrelevant polls, taken years out from an election, have become so all consuming that they’re used as political weapons, dangerous enough to threaten even a popularly elected Prime Minister.

It’s understandable that polling is journalistic crack for the Press Gallery. It creates concrete figures out of the unwieldy mass of varying opinions and allows nuanced responses to be reduced to a simple yes/no, for/against, capitalist/communist result. Much easier to deal with than policy analysis or objective assessment of government performance, but it’s lazy journalism.

While we hear so much about polls in the media, we hear very little about the mechanisms that give us the results. So here are the basics on the major polling organisations and some common tactics.

A short history

In 1916 The American Literary Digest was the first modern organisation to poll its several million strong readership in an attempt to predict election results, but it was George Gallup’s American Institute of Public Opinion that first used the modern methods of weighting responses from a sample of the population to get the kind of polling we see today.

Gallup not only predicted Alf Landon’s 1936 election win, he also correctly predicted that the Literary Digest’s results would forecast a landslide win the opposite way to Franklin D. Roosevelt, based on a smaller sample size he chose to mirror the Digest’s readership.

The first Gallup style polling in Australia was done by Roy Morgan in 1941, they were also famous for creating the ‘Reactor’, more commonly referred to as ‘the worm’. Morgan Polls still run today, but, probably because they are much less frequent, have less impact than the other major polls.

Newspoll is the grandaddy of the major polling organisations in Australia. Newspoll’s standard federal polls are run fortnightly and published on a Tuesday exclusively in The Australian, although the results generally inform stories in the other papers in the News Limited stable and often bleed into articles in Fairfax, which then inevitably drive questions and stories on both television and radio news shows. The average sample size is 1,200, and all polling is done by phone — interestingly, Newspoll does conduct specialised online polling, but this is used only for marketing research, not political polls.

In 2004 Galaxy Research started polling on alternative fortnights to Newpoll as a kind of sister poll. It runs in News Ltd’s huge tabloid stable; the Herald Sun, Courier Mail, Daily Telegraph and Sunday Telegraph. Its average sample size is 1,100 and polling is done by phone on Friday, Saturday and Sunday before being printed the following week. Galaxy has made a name for itself as the most accurate of the major polls when it comes to predicting election results. Peter Brent of The Australian (and previously Mumble.com.au) argues that this is mostly because of their allocation of preference votes for the 2 party preferred figure and wider sampling of voter preferences for third parties.

Essential Report is the new kid on the block. It is run weekly by Essential Media and most often reported in Crikey. What sets it apart from the other polls is that it relies on online polling. It is also run by an openly progressive company that spruiks its ability to help win political campaigns – happily referring on their website to a number of ‘wins’, including the ACTU campaign against Howard’s Work Choices and the Wilderness Society’s campaign against the Gunns timber mill in Tasmania.

Essential’s polling is conducted by Your Source, a social and marketing research company that has access to approximately 100,000 respondents through a self-service online portal. An invitation is sent out to 7,000 – 8,000 people on this list, and this generally results in a sample size of 1,500 – 2,000, slightly larger than the Galaxy and Newspoll. Some sections of the media have expressed trepidation about Essential’s results because of its preference for online polling and avowed progressive agenda. However, given that Essential generally echo the other major polls, this doesn’t really carry any weight.

Sampling and weighting

Sample groups are taken from a random selection of phone numbers across the nation — or, in the case of Essential, a random selection of Your Source’s existing table. Of course, if the sample is genuinely random, there’s no way to guarantee that you are getting a true reflection of the total population, it could lean too heavily towards particular sections of the electorate (say, a majority of respondents over 60). Over time, with randomised samples, this should even out, but it can severely distort a single round of polling.

This is where weighting comes in. Weighting a raw sample in polling is meant to even out the bias of the sample to give a more accurate representation of the population as a whole. Weighting in Australia relies upon demographic information from the census and the polling company’s own data on age, gender and socioeconomic status. At a basic level, if your sample includes 400 women and 500 men, then you would even this out by factoring that Australia actually has a balance of 99.2 men to every 100 women – so a woman’s response in the poll would be weighted more heavily than a man’s to give a better idea of the actual gender balance*.

While the columnists love to harp on about slight movements up and down, there’s generally a margin of error of about 2-3% in polling, so most of the time changing poll result are actually meaningless. Much greater insight is gained by tracking movements over a long period of time. Of course, that doesn’t make for interesting column inches, so most of the media tend not to dwell on it.

2 years out...

2 years out from the 2010 election, the leaders of the two main political parties were Kevin Rudd and Brendan Nelson. The ALP was on a 2 party preferred score of 66% according to Newspoll. When the election was held on August 21 2010, Julia Gillard was the Prime Minister, Tony Abbott the opposition leader and the ALP won the election by a narrow 50.12% and had to gain the support of a few independents to actually win government.

2 years out from the 2007 election, the two leaders were John Howard and Kim Beazley (phoned in for another round after Latham pulled the pin in a huff), and the ALP was narrowly in the lead on 50.5%, again, according to Newspoll. When the votes were finally counted after the 2007 election, Kevin Rudd won with a 2 party preferred vote of 52.7%.

It’s clear that even from these two examples, the polls two years out don’t have any real connection to the actual election result. In both cases, the leaders were different and the political climate had changed altogether by the election. While companies like Galaxy claim they predicted the 2007 federal election to within 1% accuracy, this accuracy is based on polls immediately preceding the election.

Push Polling

‘Push polling’ is actually not really polling at all. The point of push polling is to push your agenda onto the person being surveyed, or to bring something to their attention that they otherwise may not have known about. Traditionally, this would be done in the form of standard phone polling where the questions are framed as legitimate polling to get a desired result. These days it’s more likely to be the case that push polling pops up in the form of the friendly ‘online poll’. A recent news.com.au poll asked ‘Is the Federal government too soft on asylum seekers cheating their way to Australia? Yes/No.’

So it’s not actually asking a question, it just snookers the reader into a particular position — asylum seekers are cheating their way into this country, you just get to decide whether the government is being too soft on them (or not, I guess). That’s one for the intellectuals.

What does it all mean?

George Megalogenis from The Australian has argued that Newspoll transitioning from monthly to fortnightly polling quickened the pace of the leadership revolving door in Canberra and both he and Annabel Crabb have said that the ‘year of the poll’, 2010, was one of the most demoralising years in modern politics. ‘Polled Turkey’ as Crabb described it.

Acknowledging this, both Crabb and Megalogenis have taken a pledge this year to not report on the polls, but instead report the policy and politics of parliament — hoping to educate and inform, rather than regurgitate the views of the electorate back at us. Both have written some of the best pieces of recent years in their abstention.

It can be intriguing to feel the pulse every now and then, but opinion polls have become so much more than that. Poor polling was certainly not the only reason for toppling Rudd’s leadership, but it certainly helped to push nervous back-benchers down that path only two months out from an election.

It’s widely reported in the media that the Gillard government is polling badly, but actually, according to Newspoll, they’ve only dropped 5% since the beginning of 2011, which, when you take into account a 2-3% margin of error, is almost meaningless.

Given that we are probably about 2 years away from an election, Abbott’s 14% lead over Gillard in the last Newspoll is no indication at all of the actual election results. It’s also worth remembering that she’s only 3rd in the top three worst approval ratings in the history of Newspoll — beaten out of a tied 2nd place by Bob Hawke and John Howard (who would continue to be Prime Minister for another 9 years after scraping the bottom).

And who holds the record for the highest approval rating and longest sustained approval? Kevin Rudd.

But that’s a story for another day.

*Thanks to Possum Comitatus of Pollytics.com for this simple example.

—

Wes Mountain regularly writes and draws for his blog Dickhead Frenzy and plays guitar, sings and writes for the band Little Bones. When he is not doing those other things, he also works in local government for actual money. He is on twitter: @therevmountain — and apologises for everything he has ever said there unreservedly.

+ 1

+ 1