Making sense of the polls

Making sense of the polls

FINAL Federal Poll Analysis

Projected Winner: the Coalition

The table below has the usual poll, two party result, change from last issue, fieldwork dates and sample size.

final poll table.

Newspoll has not made a major election error with its final poll for a very long time, and its final poll here has the Coalition’s lead at 54-46 assuming 2010 preference flows. Even if this year’s preference flows are better for Labor than in 2010, the Coalition would still be well ahead in this final Newspoll. It is on this basis, and on the basis that none of the final polls have Labor even tied, that I am calling the election for the Coalition, which is likely to win a 1996 style landslide.

There are possibly two rays of hope for Labor. One is the Lonergan robopoll, conducted using only mobiles, which puts the Coalition’s lead at only 51-49. However, this poll’s sample size is too small to take it seriously. Another ray of hope is that Morgan released interim results for his poll on Friday which had Labor in a worse position than his final results; this implies that Friday night’s sample was better for Labor than Wednesday and Thursday nights. Morgan has the Coalition’s lead at 53.5-46.5 on respondent-allocated preferences compared with 54.5-45.5 on previous election preferences, but the Palmer United Party’s how to vote cards put the Coalition ahead of Labor. As a result, Labor will probably not get as many preferences from PUP as would be predicted using respondent-allocated preferences.

A notable feature of the poll table is that robopollsters ReachTEL and Lonergan are better for Labor than live interview pollsters Nielsen and Newspoll. ReachTEL conducted three separate one-night surveys this week that showed the surge for PUP. These surveys are tabulated in the Notes.

The Poll Bludger’s final BludgerTrack poll aggregate has the Coalition leading by 53.5-46.5, on primaries of 44.3% for the Coalition, 33.9% for Labor, 9.4% for the Greens and 12.4% for Others. The Coalition is predicted to win 90 seats to Labor’s 57.

Notes on These Polls

  • Essential still has 15% who say it is “quite possible” that they could change their vote. Since the campaign began, 16% say their opinion of Kevin Rudd and the Labor party has gone up, while 40% say it has gone down. A similar question on Tony Abbott and the Liberal party finds 25% who say their opinion has gone up, and 35% that it has gone down.

  • Galaxy has the Coalition’s lead on the economy up from 43-35 last week to 47-34; this probably shows the impact of Labor’s costings blunder. Abbott is thought to have run the better campaign by a 43-36 margin over Rudd. 44% say that if the Coalition wins, they will win because Labor deserved to lose, while 28% say the Coalition deserves to win.

  • Lonergan’s robopoll was conducted using only mobiles. I am not sure why such a small sample size was used.

  • Morgan’s final poll used a combination of SMS, phone and online interviewing. His normal polls use face-to-face instead of phone interviewing. Morgan first started using this multi mode methodology early this year, and before that his face-to-face polls were notoriously biased towards Labor. Since Morgan started using his new methodology, that bias has virtually disappeared. However, Morgan’s multi mode polls without the face to face component over the last week have favoured the Coalition, suggesting that perhaps Morgan’s online panel is pro-Coalition. It appears that normally there is cancellation of a pro-Labor face-to-face effect, and a pro-Coalition something else, but without the face-to-face component Morgan’s polls have become pro-Coalition.

  • Nielsen gives Rudd a net approval of -9, down 1% from two weeks ago.

  • Newspoll gives Rudd a net approval of -25, up 1% from last weekend.

  • Robopollster ReachTEL’s final four poll results are presented here; they conducted one-night surveys from Tuesday to Thursday this week, each with over 3,500 respondents. ReachTEL gives primaries to one decimal place, but rounds two party votes to the nearest integer. However, two party votes can be calculated to a decimal place using the previous election’s preferences. ReachTEL is assuming that PUP preferences split 55-45 to the Coalition, rather than the 60-40 that would be suggested on previous election preferences.

ReachTEL.

The table shows how PUP has surged from 2.4% on the 26 August to 7% on 5 September; this surge seems to have come more at the expense of the Coalition than Labor.