Federal Poll Analysis - D-day Minus Four

This Week's Polls

As usual, the table below shows the poll, two party result, change from that poll's last issue, fieldwork dates and sample size.

The two party figures in the above table obscure the increase in the "Others" vote over the last week. Newspoll has Labor down 4% on primary to 33%, but the Coalition is also down 1% to 46%, with Others climbing 4% to 11%. In Morgan, Others took 2% from the Coalition and 0.5% from Labor, and the Coalition now leads Labor 43-34 on primaries with Others on 12%. Morgan's commentary indicates that Clive Palmer's Palmer United Party (PUP) is up to 4% nationally and 7.5% in Queensland. Essential also has the Others vote up 1% to 11%, and has told the Poll Bludger that PUP constituted 4% of the vote.

These three polls all have the combined major party vote below 80%. The Pauline Hanson influenced election of 1998 is the only recent election where the major parties combined did not receive at least 80% of the vote. I believe that the movement to Others in this week's polls is caused by voters who dislike Labor's costings debacle, but are reluctant to vote for the Coalition. Because Others are loosely committed, they could still be won by either party. If Labor has a strong last week, they could regain much lost ground, and make the result close. However, another blunder like last Thursday's costings fiasco, and it could be a Coalition landslide victory.

Update 4 September: A ReachTEL robopoll taken Tuesday night 3 September from a sample of over 3,800 has the Coalition's lead at 52-48, a 1% shift to Labor from last week's ReachTEL. In this poll, the PUP's gains have come at the Coalition's expense; PUP's primary is up 2% to 4.5%, while the Coalition primary falls from 46.5% to 44%, with Labor steady on 35.5%. ReachTEL has had a house effect that favours the Coalition by 1%, so it is possible that this poll could effectively be 51-49. Last week's ReachTEL did not show a house effect, but it may be that the Coalition's lead spiked after the costings debacle, and that this effect is now washing out of the system.

Electorate-level polling from Galaxy's robopolling unit is much more in line with national polling than any other electorate-level polling. In the US, Public Policy Polling and Survey USA are two robopollsters that have reasonable track records, so the method of robopolling is not necessarily a bad method, and it seems that Galaxy is able to do it well in Australia.

In Queensland we have had two polls. A Nielsen Queensland poll taken last weekend from a sample of 1000 had primaries of 45% for the Coalition, 31% Labor, 8% Greens, 8% PUP and 4% Katter Party. When respondents were asked for their preferences, this produced a two party result of 53-47 to the Coalition, a swing to Labor of 2% in Queensland. On the other hand, a Newspoll survey of the seven most marginal Labor-held Queensland seats taken between last Tuesday and Sunday from a sample of 800 recorded a 5% swing to the Coalition, suggesting that Labor would only hold three Queensland seats post-election. This has continued the pattern in this election campaign where the national polls have been better for Labor than the individual seat polls.

With Essential moving in line with the other polls, the current national poll situation is about 53-47 to the Coalition, unchanged from last week. Newspoll has the Coalition lead over Labor on primary votes at 13%, while Morgan and Essential show a Coalition primary vote lead of 9%. Labor can perhaps be thankful that the costings blunder did not cost it more support. A small positive for Labor is that respondent-allocated preferences continue to show stronger results for Labor than if the last election's preferences are used. If 1% is added to Labor's two party estimate, they only trail by 52-48, though they would still need to make up much ground in the last week to win from there.

I will have a final post on Saturday morning regarding the final pre-election polls.

Notes on These Polls

• Galaxy surprisingly has Labor gaining ground on economic management. On this question, Labor is up 2% to 35% and the Coalition down 2% to 43%. However, this poll was half conducted before the costings debacle on Thursday 29 August.

• Newspoll has Rudd's approval down 4% to 32% and his disapproval up 6% to 58% for a net approval of -26; clearly some of this fall reflects the costings debacle. Newspoll's approvals have been much harsher on Rudd than other pollsters; Nielsen's Queensland poll gives Rudd a -5 net approval, and Essential also has Rudd at -5. While Abbott is now the preferred PM, his personal ratings also fell somewhat from a net approval of -7 the week before to -10 now.

• Morgan's respondent-allocated preferences were 52-48 to the Coalition, compared with 52.5-47.5 using the previous election. Morgan also conducted a poll using phone and Internet surveying on Wednesday and Thursday nights, which found the Coalition ahead by 53-47 using respondent-allocated preferences and 54-46 using previous election preferences. This poll is not comparable with Morgan's usual polls, which use face-to-face, SMS and Internet surveying.

• Essential is now giving weekly voting intention results, rather than averaging over two weeks. The movement in the table is from last week's 51-49 poll; the one before that had a pro-Labor sample. Rudd's net approval is -5, compared with -4 last week. Compared with early July, Rudd has lost ground on positive attributes and gained on negative ones, but he still leads or ties Abbott on most positive attributes, and trails him on most negatives. The exceptions are that Rudd is perceived as being more arrogant and erratic than Abbott. The Coalition has a 47-32 lead over Labor on economic management; that margin is unchanged from late July. However, 23% say the most important economic issue is unemployment; only 11% say that about the budget surplus. Labor holds a slender 38-36 lead on better party to handle jobs. 15% say it is still "quite possible" that they could change their mind on who to vote for, including 39% of Other voters.

Adrian Beaumont, PhD Student, Department of Mathematics and Statistics, University of Melbourne

This article is republished from The Conversation under a Creative Commons license. Read the original article.