ACNielsen online poll: 58-42

How seriously to take this Fairfax/Nielsen online poll? It should be made clear from the outset that this is not one of your Sky News/NineMSN type jobs – its sample of 1425 was “selected from Nielsen’s ‘Your Voice’ database” to represent “a broad cross-section of the nation”. As the dedicated YourVoice website explains, this database consists of 90,000 people who have volunteered to provide market research data in exchange for prizes and, if they can put up with it long enough, gifts. The result of the poll is a 58-42 two-party lead for Labor, who also lead 50 per cent to 37 per cent on the primary vote. As the table accompanying the report makes clear, these results are remarkably similar to those of ACNielsen’s most recent phone poll, conducted from September 6-8.

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.

96 comments on “ACNielsen online poll: 58-42”

Comments Page 1 of 2
1 2
  1. all within the usual range/moe…

    i also have the view that labor will keep on out-campaigning the libs, notwithstanding occasional glitches, so expect to see bouyancy in the labor numbers. i for one am not buying into the ‘soft’ labor vote story until i see some real evidence of it.

    i also think pro-labor sentiment is more visible in this election: maybe this will reflect in higher measured numbers…but its just a maybe…no empirical support for this surmise.

  2. William, what are you still doing up? Get some sleep.

    Re: the poll. I just noticed this item up on the SMH site, before flicking back to Pollbludger to find you already had a new stream going.

    You would have to treat this with a bit of scepticism, I would’ve thought. The methodology is novel, with nothing to compare it to, and there could be a bias built into the 90,000 population from which the sample is taken. There is no evidence to say that this 90,000 represents the voting population. Nevertheless, a nice headline for Labor fans.

  3. Yes Blindoptimist it does look that Labor has a concrete 54-55% tpp vote and I also think that the Morgan soft vote theory is not reflective of those voters who think that the country is heading in the right direction since Rudd became leader and are going to vote Labor.

    This is an interesting way to poll and it gets those voters who may not have a land line but are active online.

    I’m just about polled out, smeared out, and feared out. I hope he calls the election soon.

  4. all samples have some bias. the pollster’s skill is in weighting the samples to correct for bias. 1425 respondents is a goodly number. if the demographic and geographic weightings are done properly, it can be as valid a result as any other…or am i wrong?

  5. Willy

    There are two claims of accuracy.

    1) The 1425 is selected from the 90,000 to represent the full population (which is believable as they take quite a few details of each registrant, so they can match groups to known demographics)


    2) it has been reported that such a poll on election eve 2004 got within 1% of the actual election result.

  6. aj,

    the polls have been so consistent, and match up with reports of enormous swings measured in some unlikely places – like grey and parts of sydney. like you, i think lots of people wish the election was already over. every day howard postpones frustrates people and reminds them he is running scared. he should go without delay, but of course, he is hesitant…

  7. Blindoptimist: but the poll is not sampling the whole voting poulation, it is sampling a population of 90,000 peopole who have signed on to this online marketing tool. I think it is risky to project any finding from this population on to the population of voters as being representative, no matter what weighting or massaging is done.

  8. It’s the primary vote, yet again, that remains so decisive.

    I think I’ve only said this about six million times, but looking at a nation-wide 2PP vote doesn’t tell us anything, not least because it comes up with this “coalition” vote thing, which is wholly artificial. Most electorates around the country do not have a three-horse race involving the Nats, the Libs and Labor, and it means even less in the cities where there is no National Party.

    If this poll tells us anything it is that Labor is light years ahead on primaries–an effect that is only going to be magnified in the major capital cities.

    Coalition demolition in suburbia.

  9. willy,

    i can see your point, but since we don’t really know the composition of the 90,000 or the respondents to the survey, we can’t say how represntative the result is. but we can say the pollsters know about this stuff. they must allow for it in some way….or is it just a crude tally of the numbers?

    nielson mostly survey all kinds of other things, so you’d expect their 90,000 would be a nicely tuned cross-section. i don’t think the brewers, carmakers, banks etc would pay nielson for lousy data…not for long anyway.

  10. If it’s done accurately it’s as valid as any other polling method and it’s seems to have worked well overseas. Of coarse government supporters will say Nationals and Liberals don’t go online as much as Labor people but I think that’s a load of bull.

  11. Pant pant..just run over to Antony Green’s election thingy and put these numbers in. The scale only went to 57.3%, but this would give Labor 106 seats to Liberal 42. Not that I think that Labor would get this sort of outcome, but I do think that (like you said Blindoptimist) some unexpected seats will go Labor’s way.

    Anyone know of the tops of the heads which of the government front bench would maybe, kinda, loose their seats if this result was performed on the election day??

  12. Re 15

    You only have to look at the vastly disproportionate representation of ALP supporters on online forums and threads such as these to work out that there are less Liberals online. It’s common sense.

  13. i reckon the numbers show how voters respond to the campaigns of the contenders. labor/rudd are just miles ahead of the libs/howard. it is howard that looks out of his depth these days. try as they might to slur, undermine or club rudd, he is still regarded as smart, capable, determined. if anything, the attacks on him only tend to make him look resolute, human and impassioned: not bad qualities in a leader.

    these nielson numbers – or something similar – are what you would expect to see recorded by the front-runner, not least because howard’s negatives are more or less engraved in the public psyche.

  14. I’m not trying to say that the methodology of this poll is necessarily flawed. But it would be nice to see a few more polls in a series to compare with the other polling methods. I don’t think you can meaningfully compare this result with the last telephone poll the way the SMH graphic does.

  15. [ The only poll that counts is the election poll. ]

    So why doesn’t Howard call the damn thing and put us all out of our misery? Could it be because Howard knows the polls are correct?

  16. a better way to look at these numbers is to ask: what is the best kind of number howard could realistically expect to achieve. considering that about 45% of the public can’t stand howard any more and 15% are indifferent, the best primary number he can hope for is about 40%. you can add 5% to estimate the 2PP when the undecideds are shared around. this is the best he can get because nearly half the electorate have completely had their fill of him.

    it follows, in a 2-horse race, that if horse A is only going to get at most 45%, horse B has to get 55% or more.

    hence the results. the longer people ask the same questions, the more they will get the same answers.

    the real question about the polls is: will people vote the way they say they will.

    the pollsters should be drilling down to test the depth of people’s intentions.

  17. Hmm. It will be fascinating post election to learn if this poll was any closer to the final outcome than others. Especially given certain past and present pollsters publishing in msm seem to question their own results.

  18. [You only have to look at the vastly disproportionate representation of ALP supporters on online forums and threads such as these to work out that there are less Liberals online. It’s common sense.]

    Not much common sense in this comment. The reason ALP supporters are better represented is because their views are more commonsensical. The Tory line is indefensible. I work in market research – there are a HIGHER proportion of coalition voters online than ALP voters.

  19. blindoptimist 21 makes good points. Why are the obvious questions not asked.

    Back to this poll, despite the selection as described, smacks of ‘guess the right answer’ given it is incentive driven. Built in mutant factor.

    Disappointees may find themselves voting other than.

  20. On the subject of new polling techniques. are the phone polls conducted only on land lines? or are mobiles included (I know VOIP’s arnt as they are seldom listed). Surely this would skew the findings by limiting the sample communities available the poor and young are increasingly only using the mobile due to line rentals.

  21. Im loving the way the polls are quickly crushing the desperate attempts by Lib staffers (and the few dodgy media apologists still onside) to spin some momentum.

    Id put Textor’s internal poll leaks in the same class as Howard’s recent call on how close Eden-Monaro was.

    File under p for ‘pffft’.

  22. I’m sorry, but I have a problem with a poll that relies on a certain demographic actually participating, rather than one selected at random. Sure, there are going to be people that refuse to participate in a phone-up poll, but it’s gotta be more accurate than relying on people to phone-in.

    Sure, the numbers are the same, but that could just as easily be coincidence.

  23. 16
    aj Says:
    September 25th, 2007 at 2:29 am
    Pant pant..just run over to Antony Green’s election thingy and put these numbers in. The scale only went to 57.3%, but this would give Labor 106 seats to Liberal 42. Not that I think that Labor would get this sort of outcome, but I do think that (like you said Blindoptimist) some unexpected seats will go Labor’s way.

    Anyone know of the tops of the heads which of the government front bench would maybe, kinda, loose their seats if this result was performed on the election day??

    Using the Oz Politics calculator ( ) it comes up with the following (to answer your question): [this calculator assumes uniform swing and you can put in any number]

    Bennelong – Howard [6.61 to Labor]
    Wentworth – Turnbull [8.23 to Labor]
    Gippsland – McGauran [3.04 to Labor]
    Longman – Brough [3.99 to Labor]
    Higgins – Costello [1.98 to Labor]
    North Sydney – Hockey [0.7 to Labor]
    Menzies – Andrews [0.07 to Labor]

    Everyone else in the Cabinet keeps their seats under this scenario. New margins:

    Warringah – Abbot @ .55
    Mayo – Downer @ 2.85
    Berowra – Ruddock @ 3.08
    Wide Bay – Truss @ 1.47
    Lyne – Vaile @ 2.67
    Curtin – Bishop @ 3.88
    Bradfield – Nelson @ 6.81
    Groom – MacFarlane @ 8.07

    Largest remaining margin is Mallee in VIC @ 14.01

  24. #17 rcandelori I suggest the online community represents the current feeling of the population, thats why the Liberals have to pay people to do leafleting. Are you suggesting in you post that Liberals are living in the past ?

  25. If you have a database of 90,000 people. You could select from swinging seats and types of professions. Thereby getting a more accurate result. This would be how they get better results for their customers.

  26. It’s another good polling result for the ALP, but I just wish the campaign would start already so we can start getting more leaks on individual seat polls.

  27. The question seems to be not whether the 90,000 whole is proportionate to the electorate, but whether it is possible to select ~1400 from the 90,000 who are, methinks.

    Unless there is a significant demographic missing totally or massively under-represented, this poll could be more accurate, as you (in theory) know more about each person, and are able to weight the results better than you would otherwise be able to to match the total population.

  28. Don’t think this poll methodology is too novel. In my work (not political) I have used AC Neilsen and found them very accurate. AC Neilsen, like all polling companies, are looking for ways to overcome the problems of traditional survey techniques. Typically it is very hard to get phone and interview surveys to include all demographic groups evenly, because some groups are consistently less likely to be home when the surveys are done. This may then lead to bias in the results. This is especially true of young adults, who are often one of the big unknowns in voting patterns. So don’t think this is automatically less reliable than Newspoll etc. In fact, it may be more accurate. The difficulty is that they are hard to directly compare. So don’t compare this to Galaxy; compare to previous results.

    Any technique will have biases, unless every age group used all forms of technology equally, and was at home equal proportions of the time. I have thought for some time that Galaxy tends to be slightly pro-ALP, Newspoll slightly pro-Liberal, and AC Neilsen the most reliable. Their past record on picking things liike the swing to Pauline Hansen etc, is very good.

    “Random selection” is also misleading. Surveyers try to get representative samples in each age group. This often requires asking particular people in households to answer, and declining to survey others. Newspoll rang my household two weks ago, and declined to survey me but asked to survey my wife, because she was in the required age group and I wasn’t. The surveyer explicitely commented that they had trouble getting a sample in her age bracket (we fell either side of a dividing line).

    All of this is to say that all of the survey groups try to get it right, but short of holding the actual election now, it is a difficult thing to predict. This poll is as good a method as any other. If you are sitting in a seat with a lot of over 50s, it may be less reliable. If you are sitting in a seat with a lot of Generation X and Y, it is probably the most accurate.

  29. I am a little sceptical about these online polls, however I do recall a few weeks ago talk about a British company that was bringing this type of polling to Oz in October.

    Their experience in the UK claimed to have a 1% accuracy.

  30. One more thing – the use of this sort of technology also varies with income and education level. For example the North Sydney area is older than average but also wealthy and very highly educated. It has a well above average level of internet use, even amoung over 60 yr olds. There is no reason to believe this result doesn’t apply to Bennelong.

  31. The Flat Earth Society is also under represented, as well as the creationalists. The university types are over represented. (subversives). But don’t tell my two sons I said that.

  32. It would be very easy to set up a computer program, to select exactly the right type of people to be surveyed. The more accurate the results, the more business they acquire.

  33. In regard to surveys missing out on mobile phones, this aspect was continually raised in the last US election by Democrat supporters. The results showed that it had no effect. What did not show up was the religious vote. That will not have an effect in the next US election.

  34. The Age must be getting desperate to continue Rudd’s honeymoon if they can’t even publish a poll conducted using the standard method.

    I personally know numerous lifelong Age readers who have cancelled their subscriptons and switched to the Oz as they have grown tired of the one-sided reporting and blind bias of the paper’s political reporting team.

    So it would not surprise me that remaining Fairfax readers are more likely to vote Labor than the general population.

    I suspect that Rudd’s latest blunders and the arrogance of the Labor team have dented his support, although maybe not enough for Howard to close the gap by the end of the year.

    The Business Coalitions TV Blitz showing the village idiot mouthing off about Workchoices, and being corrected by the sensible guy, are almost as believable as the ads on Latham’s council record – and will have an impact in the usual places (outer suburbs of cities plus country towns).

  35. With regards to the sample pool, couldn’t those signing up for ‘prizes’ and ‘gifts’ be labeled ‘aspirational’? This is not good John…

Comments are closed.

Comments Page 1 of 2
1 2