Making sense of the polls

Making sense of the polls

Final Election Results and a Polling Critique

Final results have now been released for the September Federal election. The Coalition won 45.6% of the primary vote, up 1.9% on 2010. Labor won 33.4%, down 4.6%, the Greens received 8.7%, down 3.1%, and all Others, including Palmer United Party (PUP) had 12.4%, up 5.8%. PUP itself debuted with 5.5% of the vote. The Coalition won the national two party preferred (2PP) vote by 53.5-46.5, a 3.6% swing to the Coalition. In the House of Representatives, the party standings are: Coalition 90 seats, up 17, Labor 55, down 17, 2 Independents, 1 Green, 1 Katter’s Australian Party and 1 PUP.

The table below shows summary results for each state. The columns are the state, then that state’s total seats. The other columns refer to the Coalition’s seats in that state, % of the seats, net seats gained, state Coalition 2PP and state 2PP swing to the Coalition.

final House.

State Notes

  • In NSW, Labor was lucky to win 26 of the 48 seats in 2010 on only 48.8% of the 2PP vote. Although the 3.2% swing in NSW was not as bad as in some other states, it was going to be costly for Labor due to their 2010 fortune. The Coalition gained two seats from retiring Independents and eight seats from Labor. Contrary to pre-election expectations, swings were higher outside Sydney than within Sydney, and four of Labor’s eight losses were regional seats.

  • In Victoria, the Coalition easily won the three most marginal Labor seats of La Trobe, Corangamite and Deakin, but were unable to gain more, with Labor just holding on in McEwen. Sophie Mirabella was the only sitting Coalition member to lose, suffering a shock loss to an Independent in Indi.

  • In Queensland, the swing against Labor was the least of any state, and Labor only lost two seats. The Coalition lost Fairfax to Clive Palmer by 53 votes.

  • WA gave the Coalition its highest 2PP vote at 58.3%, a 2.3% swing, but no seats changed hands.

  • Despite a 5.5% swing in SA, Labor only lost one seat.

  • While Tasmania had the strongest swing to the Coalition (9.4%), it still gave Labor its highest 2PP vote of any state (51.2%). But although Labor won the 2PP vote, it now only holds one of Tasmania’s five seats, as it lost the three northern seats to the Coalition, while Andrew Wilkie easily retained Denison.

  • In the territories, Labor easily retained both ACT seats. In the Darwin-based seat of Solomon, there was a small swing to Labor, and Solomon is now the Coalition-held seat that came closest to being won by Labor.

Poll Critique

The primary vote figures and Coalition 2PP votes for the final polls are compared with the election result in the table below. Bold numbers in the polls are where the poll was within 1% of the election result. Since not all polls gave breakdowns of PUP and Other votes, Other in the table is any vote not for a major party or the Greens.

poll critique.

Newspoll was clearly the best final poll, getting all five numbers within a point of the actual result. Morgan’s primary vote numbers were all out, but he fluked the two party result, exactly hitting the mark. Galaxy, Nielsen and ReachTEL were all fairly good, but Essential’s final poll overestimated Labor’s position, and the final mobile-only Lonergan poll was a disaster.

The 2PP estimates of the pollsters were generally accurate because two errors cancelled. Most polls overestimated the combined Labor/Green vote, but this was offset by the change in preference flows from 2010 to 2013. Had 2010 flows been used for the 2013 election, Labor’s 2PP would have been 1% worse than it was. Of the polls in the table, only Morgan used respondent-allocated preferences, and all others used 2010 preferences. Newspoll is somewhat fortunate that its accurate primary vote numbers rounded to a Coalition 2PP of 54%, and not 55%.

Another issue with polling is assessing how closely the polls followed the general trend. When Rudd returned as Prime Minister, Labor immediately bounced back to near-parity, and this situation was held throughout July 2013. Labor started to lose ground in early August, and kept losing votes until the election.

In tracking this rise and then fall, Neilsen and Galaxy were the best polls. ReachTEL was somewhat skewed towards the Coalition in July, and Newspoll had a couple of pro-Coalition outliers. Morgan badly overshot in July, at one stage giving Labor a clear lead, though this poll was better in the election campaign. Essential was the worst behaved of the regular polls; it badly lagged the other polls by still showing a Coalition lead in July, then it gave Labor a 50-50 tie in August when other polls were showing Labor declines.

Individual seat polling, whether by live interview or robopoll, was skewed towards the Coalition. There were two polls that showed Rudd losing his seat of Griffith, but he retained it comfortably. This polling also had Labor heading for a massacre in Western Sydney, with Werriwa, McMahon, Greenway and even Blaxland in danger, but Labor held these four seats easily. At this election, the question of whether individual seat polls or national polls were correct was decisively settled in favour of the national polls.

At this election, Greens preferences favoured Labor by 83-17, PUP preferences favoured the Coalition by 54-46, and all Other preferences favoured the Coalition by 53-47. These preference flow figures can now be used by pollsters.

Kevin Bonham has more on the polls.

WA Senate

I have previously posted on the final Senate results, and on the WA Senate count, where almost 1400 votes went missing. The Australian Electoral Commission (AEC) has now released details of these votes, and Antony Green says that if the missing votes were included in the count, the critical margin between the Shooters & Fishers and the Australian Christians would be just one vote. The AEC has lodged a petition with the Court of Disputed Returns to have the WA Senate result declared void, and have a re-run election. It is likely that the Court will agree with the AEC’s petition.