Same conclusion, different spin

Backlash: Bracks risks losing 16 seats, Poll finds ALP is out of touch
– headline in print edition of the Melbourne Herald Sun, 24 October 2006

Libs face crushing loss at poll: One in eight to vote for Greens
– headline in print edition of the Melbourne Age, 24 October 2006

Two rather different interpretations of poll results showing that we are on track to the expected Victorian election outcome – the return of the Bracks government with a slightly lower share of the vote. The actual two-party preferred estimates are from ACNielsen in The Age, Labor 56, Liberal-National (though there is not actually a Coalition agreement) 44; and from Galaxy in the Herald Sun Labor 52, Liberal-National 48.

The major difference between the two polls is their assessment of the minor party and independent vote, 12% according to Galaxy and 18% according to ACNielsen. Galaxy puts the Greens at 7% and ACNielsen at nearly double that, 13%. Because Greens preference to Labor, the ALP ends up with a larger 2-party preferred vote in the ACNielsen poll.

Galaxy puts its margin of error at +/-3.5%, and ACNielsen at +/-3%. They are both going to need all their margins of error, Galaxy up, ACNielsen down, to reconcile the different estimates of the Green vote.

Update: Newspoll’s Victorian election poll is in The Australian today (a small amount of information is online). It’s midway between Galaxy and ACNeilsen on the 2-party preferred, 54-46. It puts the Greens at 7%, the same as Galaxy, and the same as the Morgan Poll in September. Yesterday in comments, Pollwatcher thought that Galaxy’s low result might have been because they did not read out the Greens as an option, but Newspoll seems to have done so by asking ‘which one of the following would you vote for?’ with the Greens appearing in their table with all other minor parties in ‘other’. However the actual Green vote in 2002 was about mid-way between yesterday’s polls, on 9.7%. I’d guess that actual Green support is closer to 7%, but if they campaign more effectively than other minor parties and independents they will pick some of the stray uncommitted and protest voters.

14 thoughts on “Same conclusion, different spin

  1. The 3.5% margin of error is calculated for a party with 50% of the vote. It’s much higher for smaller parties, which means that most Australian opinion polls are useless for estimating minor party voteshares. Wikipedia has a nice discussion.


  2. Andrew – I thought there might be something like that, thanks for pointing it out. I will read the Wikipedia article carefully later to avoid future errors. And hope that Galaxy is closer to the true figure…


  3. While Andrew Leigh is right, I don’t think that the discrepancy is primarily due to margin of error issues.

    AC Neilson’s previous polls had the Greens on 10,11, 10 and 11. The jump to 13 is hardly surprising in this context given the recent focus on climate change.

    One the other hand, the last few Newspolls for Victoria have run something like 5, 6,6 (going from memory), while Morgan has mostly been around 7-8 with the last one at 5.5. I haven’t seen a Galaxy poll in Vic this year.

    Clearly there is some systematic difference between the different polling companies. If you aggragate all the AC Neilsons and all the Newspolls you get a fairly small margin for error for each, yet they average 11 and 6 respectively.


  4. I would wager that the difference in the Green numbers comes from the fact that ACNielsen reads them out and Galaxy doesn’t.

    When the Greens are read out then the disaffected Labor vote goes Green on primary, then returns to Labor on 2PP. When Greens aren’t read out disaffected Labor voters just go to the Coalition.

    The Greens received 9.8% at the last election – its implausible to think their vote will drop (although 13% is probably stretching it a bit as well).


  5. [blockquote]Andrew Leigh Says:

    October 24th, 2006 at 10:33 am
    The 3.5% margin of error is calculated for a party with 50% of the vote. It


  6. Harry – I’m not sure that pokies are a big vote switcher. After all, people are already against them and Labor has a comfortable lead. Governments can get away with many unpopular policies provided they are still better than the Opposition in a few major areas.


  7. David, you are correct that the standard error is lower for minor parties. That said, most polls are still not very good for estimating minor party vote shares as they have a much higher standard error relative to their size, i.e. it is much more likely for the Greens vote to be twice as large in reality than it is for the Liberal vote.

    Out of interest we can calculate the 95% CIs for each party (if all those pesky assumptions are met):

    Labour 42%


  8. Andrew, I think there is widespread popular opposition to pokies at all levels – 86% of Labor voters oppose them. Surely not a bad policy to at least emphasise. The Greens are doing well with this policy.

    The water policy is a dud and the free fares policy not that wonderful.


  9. David/Matt, you’re quite right – poll estimates for minor parties have smaller standard errors than poll estimates for major parties. Apologies for causing confusion, and thanks for correcting my mistake.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s