Making sense of the polls

Making sense of the polls

Introduction and Analysis of the Latest Polls for the 2013 Federal Election


I am a PhD student at the University of Melbourne’s Department of Mathematics and Statistics. Since 1998, I have had great interest in electoral politics, and I keenly follow both Australian and US elections. In the lead-up to the Australian Federal election, I will be writing a post every week about the latest polls and what they mean for the number of seats that will be won by each party in the 150-member Australian House of Representatives.

In the US, Nate Silver was able to amalgamate polls for each key state and nationally for the 2012 US Presidential election; the wealth of US polling data enabled Silver to accurately predict the winner in each state. That wealth of polling data does not exist in Australia. Here, most of our polls are national, and we do not get many public polls of key seats. Even then, several seats will usually be included, and sample sizes for individual seats within the overall survey will usually be too small for meaningful analysis. As a result, I will not attempt to determine the outcome of an individual seat. However, if there is say a 4% swing to party A in a particular state, although the swings will not be uniform across seats, seats held by less than 4% that do not swing enough for party A to gain the seat will generally be made up by seats held by more than 4% that swing enough for party A to gain.

In Australia, if either Labor or the Coalition parties obtains over 51% of the two party preferred vote, that side will probably win the election. However, if the two party vote is around 50-50, the result will depend on swings among the various states. We can deduce what is going to happen in particular states from state breakdowns of national polls that are provided by the pollsters, particularly close to an election. At the moment, the 50-50 polls are showing a considerable swing to Labor in Queensland, where they could potentially gain a number of seats, thus offsetting potential losses in Victoria and Tasmania; this could allow Labor to win the next election.

This week’s polls

The table below shows the two-party result, shift in the two-party from the last result by the same pollster, dates when the poll was conducted (fieldwork dates) and the poll’s approximate sample size.

Poll Table.

Newspoll was the only poll that showed a change in the Coalition’s favour from the last result by the same pollster, with the other three polls giving Labor a small improvement. As a result, I am sceptical of the Newspoll result, and still think that the current situation is roughly 50-50.

The Poll Bludger’s Bludgertrack measure, which aggregates the latest polling data, now has the Coalition leading 50.5-49.5. State breakdowns of ReachTEL and Morgan polls show that Labor’s position appears to have worsened in Queensland, where they need to gain multiple seats to make up for expected losses elsewhere. As a result, the Coalition would be expected to win an election held now.

Some notes on these polls

  • Newspoll is published by The Australian every two weeks, and is the most well-reported poll. Their methodology is telephone interviews. This week’s poll had Rudd’s approval down to 42% from 43% and his disapproval up from 36% to 41%. On best party to handle asylum seekers, the Coalition’s lead fell from 46-20 under Gillard to 33-26.

  • Morgan is published online weekly, and its multi-mode poll uses a combination of face-to-face, SMS and internet interviewing. While Julia Gillard was Prime Minister, Morgan’s results were in line with other polls, but since Rudd became PM, his results have been about 2% better for Labor than other polls. It’s possible that this could be a methodological flaw, or that his methodology is better geared towards mobile only households.

  • ReachTEL is a robopollster (interviews conducted by a robot). US robopollsters have generally been biased towards the conservative parties, and indeed ReachTEL appears to lean about 1% to the Coalition as compared with other polls, though they haven’t taken many national polls this year.

  • Essential Report is published online weekly. Voting intention results are based on the last two weeks of data, but Essential also asks many questions on various political topics every week. Essential’s methodology is to select its participants from a panel, and I think this means its results do not change enough. When most polls had Labor 57-43 down under Gillard, Essential had Labor only down 55-45, and its 3% shift back to Labor on Rudd’s restoration was less than any other poll recorded. This week’s political topics have Labor gaining ground on most issues, but the Coalition still holds big leads on economic issues. Voters approved of the change to a floating carbon price by a 45-29 margin.


The widely-reported Newspoll will very probably mean that Tony Abbott’s Liberal leadership is safe. This is good news for Labor because if Turnbull replaced Abbott the Liberals would definitely win. The Papua New Guinea resettlement plan is clearly aimed at stopping or at least slowing the flow of boats at the election; if Labor can succeed in this it will remove a major negative. However, economic management is still considered the most important issue according to Essential, and Labor still trails the Coalition here 44-29. Labor needs to improve its economic management credentials to win the election.