More value in polls with geographic breakdowns

31 July 2015

Boredom might be the feeling that many people get when they hear or see news stories about opinion polls.  Such is the saturation of polls in today’s media cycle that countless people just tune out at the mere mention of them.  Except when parliaments are hung, meaning that no side of politics can govern in its own right without the need for support from crossbenchers and one single resignation or death could potentially change governments, why should it matter so much what the polls say, especially when no election is clearly around the corner?

On my part, I only pay close attention to opinion polls every three months or so.  The reason is that one of the main polling entities, Newspoll, publishes its findings with geographic breakdowns every three months, or on a quarterly basis in the context of a year.  Throughout the rest of the year, Newspoll will publish findings every fortnight, albeit with less detail than when it does geographic breakdowns, but reading fortnightly polls won’t matter too much.

There are three lots of findings that I like looking at, two of which are published in major newspapers.  You can read the findings from Newspoll in the AUSTRALIAN.  There are findings from another polling entity, Ipsos, in several newspapers published by Fairfax Media, and they are published as Fairfax/Ipsos polls – of the Fairfax newspapers, I go to the AUSTRALIAN FINANCIAL REVIEW because it includes geographic breakdowns.  And another long-running polling entity, Roy Morgan Research, used to publish its findings in print but now only publishes on electronic pages, with some geographic breakdowns in its findings.

As such, the most recent opportunity to compare findings with geographic breakdowns was in June this year.  And the Newspoll and Fairfax/Ipsos and Morgan findings made for interesting reading.  My focus for the moment is on the two-party-preferred vote, meaning whether voters ultimately prefer Labor or the Coalition, especially if their first choice for voting isn’t either Labor or the Coalition.  The findings of all pollsters in June showed a swing of 6-7 per cent to Labor, thus implying an easy Labor at the next Federal election.

In the House of Representatives, where governments are formed, there are 150 seats.  The Coalition governs with a 90-55 lead over Labor, with a quintet of crossbenchers also there.  In that quintet are Adam Bandt of the Greens, Bob Katter and Clive Palmer of political parties respectively bearing their surnames, and Independents Cathy McGowan and Andrew Wilkie.

Based on the results of the last election, a swing of about 3.1 per cent to Labor would cost the Coalition its majority, but Labor needs a swing of about 4.3 per cent its way to win with a majority – I refer to having a majority because this means being able to govern without the need for support from the Greens or Independents or anyone else, and the memory of the minority Labor Government from 2010 to 2013 is still fresh in many people’s minds, sometimes uncomfortably so.

Within or under the predicted swing 6-7 per cent to Labor, as per the polls’ findings, there are thirty-one Coalition seats across the country.  New South Wales and Queensland have ten seats apiece, Victoria has four seats, Tasmania has three seats, Western Australia has two seats, and South Australia and the Northern Territory have one seat apiece.  Going back to the needed swing of 4.3 per cent to Labor, NSW and Victoria would both shed one seat less, but Queensland would shed six seats less, while WA would see no Coalition seats lost.

However, at election time, swings aren’t always uniform, and often seats within range of the predicted swing actually don’t fall.  And sometimes there are varying factors in each state or region.

For example, the Morgan findings actually showed a swing to the Coalition in Tasmania, with no seats lost there.  The Newspoll and Fairfax/Ipsos findings showed a swing of 6-7 per cent to Labor in Victoria, but the Morgan findings showed a swing of only 2-3 per cent to Labor and hence no seats lost there.  The Newspoll findings showed a swing of 8-9 per cent to Labor in WA, but the other polls’ findings showed a much larger swing and more seats lost there.  And while the Newspoll and Morgan findings both showed figures implying a loss of thirty-one seats for the Coalition, as would occur in a uniform swing, the Fairfax/Ipsos findings seemed to show fewer seats lost.

These findings show why more value appears in polls with geographic breakdowns, although they don’t happen very often.  But publishing them more often mightn’t be ideal, given how bored people seem with them nowadays.