Nielsen JWS Research marginals mega-poll (and Galaxy and Morgan)

The super-sample poll published by Fairfax, covering 22,000 voters in 54 seats (about 400 each), turns out to have been conducted not by Nielsen, but an outfit called JWS Research whose automated phone polling on the weekend were widely noted at the time. The Sydney Morning Herald sells its managing director John Scales as “a renowned pollster” who was director of Morgan from 1992 to 1995, and research director at Crosby Textor from 2002 until earlier this year. However, its methodology is untried, at least in the Australian context: presumably pollsters go to the expense of hiring interviewers for a very good reason (UPDATE: Lukas in comments points to the assessment of legendary US polling analyst Nate Silver, that “there’s not really convincing evidence about whether IVR polls [robopolls] are inferior to regular ones”). Nonetheless, the general tenor of the results is in keeping with polling overall, give or take a few surprises. Taken together, the results point to 79 seats for Labor and 68 for the Coalition with three independents – curiously, the poll did not cover Melbourne.

Queensland. Labor to lose Brisbane, Bonner, Petrie, Leichhardt, Forde, Dawson, Flynn and Dickson, but hold Moreton, Longman and Herbert.

New South Wales. Labor to lose Lindsay, Bennelong, Macarthur and Robertson while gaining Paterson and Cowper; Labor to hold Greenway, Dobell, Page, Eden-Monaro, Gilmore and Macquarie.

Victoria. Labor to gain McEwen, La Trobe and Dunkley, but lose Corangamite.

Elsewhere. Labor to gain Boothby, lose Hasluck and Swan, and hold Solomon.

We also have Galaxy polling another 800 voters in four Queensland marginals – Bonner, Forde, Herbert and Longman – and found happier results for Labor than the Queensland sample from the weekend’s super-survey, with a swing of 3.5 per cent rather than 5.4 per cent. Morgan has also published small-sample polls from Bennelong (300) and La Trobe (200) conducted just this evening, which respectively have the Liberals leading 50.5-49.5 and Labor leading 53-47.

UPDATE: And now state-by-state breakdowns from the past two weeks of Newspoll, both of which had the national result at 52-48. The state two-party are results are 52-48 to Labor in New South Wales (a 1.7 per cent swing), 52-48 to the Coalition in Queensland (2.4 per cent swing), 55-45 to Labor in Victoria (0.7 per cent swing to Labor), 56-44 to Labor in South Australia (3.6 per cent swing to Labor) and 57-43 to the Coalition in Western Australia (3.7 per cent swing to the Coalition).

UPDATE 2: Recent state-level polling results in terms of Labor swing:

TOTAL NSW VIC Qld WA SA
JWS Research -1.1 -2.1 2.4 -5.3 -0.5 -1.6
Galaxy marginals -1.7 -3.1 1.6 -4.3 -2.1 0.0
Newspoll (2 week) -0.7 -1.7 0.7 -2.4 -3.7 3.6
Nielsen (2 week) -1.7 -2.7 3.0 -3.4 -2.7 0.6
Essential (2 week) -1.7 -5.7 0.7 -3.4 0.3 1.6

Author: William Bowe

William Bowe is a Perth-based election analyst and occasional teacher of political science. His blog, The Poll Bludger, has existed in one form or another since 2004, and is one of the most heavily trafficked websites on Australian politics.

3,759 comments on “Nielsen JWS Research marginals mega-poll (and Galaxy and Morgan)”

Comments Page 1 of 76
1 2 … 76

  10. Nick of McEwan or Melbourne or wherever. I live near McEwan. I work in Melbourne. And I think you are onto something. Keep up the good work.

  18. Obviously we’d all like to see which seats were polled. Still even if that is, say 30 seats, it is almost 1000 responses per seat, which is getting pretty solid.

  20. Ghostwhovotes may nudge Bluey for the “three votes” in my Brownlow/Dally M Medal for greatest impact during PB coverage of election!

    – hope Bluey is asleep and doesn’t see this post!

  23. El Nino, I can beat that I live in McEwan and work in Melbourne!.

    Good to hear from somone green on the ground in Melb, that is an interesting contest.

  31. 28,000 respondents? Is the is robosurvey by any chance? If not it would have to be cumulative, 28,000 X 5 mins =….alot of time to be doing fieldwork

  36. [If this poll is accurate and it pans out that way on Saturday Australia will have its own Mason-Dixon line.]

    Dibs on being the Union!

  37. El Nino,
    Cheers. I visit my parents and friends in McEwen regularly. I live and study in Melbourne, and work in Maribrynong (although I had to look that up just now). I think we’re both onto something. I’m going to bed now. Night all.

  39. Andrew wrote:
    [Why then, GP, did the libs establish the process of costings for campaigns?]

    http://www.budget.gov.au/1996-97/honest.pdf

    Basically, it was to remove the “privileged access” to the public service that previous caretaker governments enjoyed.

    There was no inference or implication that this was to submit election costings to the “independent” Treasury because the Treasury cannot be independent by virtue of its complete submission to the executive government of the day.

  43. 8,000 respondents? Is the is robosurvey by any chance? If not it would have to be cumulative, 28,000 X 5 mins =….alot of time to be doing fieldwork

    Comes to 292 8-hour days.

Comments are closed.

Comments Page 1 of 76
1 2 … 76