Now that the election is done and dusted, it’s possible to look back and compare the actual results with the voting intentions reported by opinion polls during the course of the year, and with the predictions made by commentators on the basis of those opinion polls. In summary, the verdict is that the level of Labor support on polling day was lower than reported by virtually every opinion poll during 2007, and that consequently the great majority of predictions made by commentators of the election outcome, based on those polls, were wrong.
Read the full article here.
December 9, 2007 at 6:32 pm
Yes and no. There are 2 types of higher-income voters: 1) those in the public sector/caring professions who have drifted steadily to the left under Howard, 2004 actually showed this; 2) private-sector professionals who moved steadily towards Howard especially in 2004, a lot in this group thought Howard was yesterday’s man, big business is concerned about climate change for example, they liked Rudd’s Keating-lite style, but although this group probably swung strongly to Labor compared to 2004 they will remain a Liberal heartland. They are a group which has done very well materially in a globalised economy unlike public sector professionals. Turnbull would be their idol, they liked Costello but they wanted Howard to go. The different swings in Bennelong, Higgins and Wentworth match to this.
December 9, 2007 at 6:41 pm
A very thoughful analysis, Adam. You were also right about the Democrats and, more or less, the Greens.
December 9, 2007 at 6:53 pm
Thank you Adam,
thats one of the better explanations for the Narrowing, and for the lack of big swings seen in safe seats (apart from a few *star* exceptions).
Its like the “ReSet” button was pushed, back to the original socio-economic class base of the two major Parties.
When I was a young girl in junior high, during the 1972 campaign which was very ubiquitous even for one as young as I was, I asked my Mum and Dad about it – saying that I thought Gough looked pretty cool with all those great hippy-trippy jingle songs, but really, did it matter – what was the difference between Liberal and Labor anyway?
My Dad said “Simple , Liberal is for the rich bastards, Labor is for everybody else”.
Several years later during the Fraser years, I went on to Sydney Uni (when it was still free, and affordable for rural working-class riff-raff like me) and sitting in the Student Union bar one night, overheard a bunch of Young Liberal law students, and thought Dad was spot on
Its only been in recent times that it somehow became blurred methinx, perhaps starting in the latter years of Hawke/Keating?
December 9, 2007 at 6:53 pm
Well I remain pretty contemptuous of the whole “narrowing” meme.
Let’s get real here. Those commentators who pushed the “narrowing” line were not saying that Labor would end up a comfortable majority as opposed to a whopping majority. The clear implication was that the election would be close and the Coalition was every chance of winning. It wasn’t and they weren’t.
On the other hand, those who called the election for Labor months in advance were right. It didn’t mean, necessarily, that Labor would hold every bit of their large lead. It just meant that the gap was too large for the Coalition to haul in.
December 9, 2007 at 7:53 pm
You are too hard on Possum. It was he who mid way through the campaign said that the Liberals had a firewall strategy, which was to hold their safe seats while sacrificing their marginal seats. It looks like they did just this with the private school giveaway and, probably, the union scare campaign. The doctors’ wives care about the enivironment and all that, but they are easily persuaded that unionists are thugs, since they never get to meet any actual unionists in real life.
Incidentally,Ryan was supposed to fall to Labor not for post-material reasons but because the Liberals planned to put a big freeway through the electorate. It was reasonable to think that Ryan would go Labor for this reasons, and because the Liberal member Michael Johnson is a prize dill, and because Labor won Ryan in a by-election as recently as 2001.
December 9, 2007 at 8:41 pm
Two minor quibbles:
First, while my class groupthink makes me personally appalled a the Haneef case, I certainly take (and took) your worldy hard-boiled point. But the case was so badly handled it was possibly a lose/lose for the coalition ie 20% of the electorate seeing the government rushing to stomp on civil liberties for political gain and 80% seeingyhe government bungling an anti-terrorism case. So it was certainly possible to see the Haneef case as a potential liability for the government without that necessarily being (only) from a leftist position.
Second, your comment about Poss being wrong about all of the other seats mentioned is ambiguous since Flynn also fell.
December 9, 2007 at 9:06 pm
I think it was the Labor campaign that failed in the last few weeks and meant that the extra 1 to 1.5% swing that might have come actually failed to come over.
The Labor campaign was too negative particularly in the ads. The first whingeing wendy ad was good. The second was not. They attacked Howard too much rather than focussing on their positive message. They kept warning us about the mother of all fear campaigns which did materialise, but their warning was itself a negative message. They had a strong advantage on climate change at the beginning but the pulling of Garrett into line on the Kyoto 2 message meant many people couldn’t see a difference between the parties on this anymore (though the difference actually is very large). Labor was robotically on message, and this meant they failed on the ’sincerity’ test. People felt Labor was just saying things to get elected. (I have to admire Howard here. He was able to sound really sincere and concerned most of the time, when in actual fact everything he said was only to try and get re-elected)
Talking about taking a meat axe to the public service lost Labor a several percent swing in the ACT, and took away the slim hope of knocking off the Liberal Senator.
So in my view not a brilliant campaign. Competent, controlled, careful, largely mistake free, but the campaign did not excite hope, did not appeal to the idealists and so the very strong ‘its time’ mood was not fully translated into a bigger swing.
December 9, 2007 at 9:10 pm
It’s worth noting that Labor got close in Sturt because it is not a full blooded doctors wives seat. I live in Sturt and much of it is fairly working class (the seat is number 60 on income). It is only the bottom third of the electorate that is blue blood. This part of the seat saved it for the libs. The ritzy booths hardly swung at all while the working class ones produced huge swings (one over 13%).
This then supports your argument about the class based election. The people who could afford to care didn’t change their vote and the battlers did.
December 9, 2007 at 9:41 pm
Thanks for all those comments. I agree Possum was perceptive about the Libs’ firewall tactic. If that’s what they did, it worked. I doubt however that Howard was ever persuaded that the election was lost and all he was doing was campaigning to save some seats for the post-2007 election.
Zack is quite right about Sturt – it only just scrapes in as a high-income seat, which is why Labor came as close as it did. I’m sure that booth pattern, which is the reverse of 2004, was repeated in many seats across the country. It certainly was in the seat where I was working.
December 10, 2007 at 10:51 am
I also think that there were not many people for whom civil liberties would be a vote-changer who didn’t vote Labor in 2004 as well, and probably 2001. I live in Kooyong, and if anything it might have swung back to Petro Georgiou on the back of his standing up to Howard on child detention.
December 10, 2007 at 9:12 pm
Adam,
One shouldn’t be too hard on the optimists who predicted a sweeping Rudd victory in the vicinity of 87-90 seats (of which I admit to have been one [having predicted 90 seats for Labor]). Whilst some of the criticism you have directed towards us stands, we were at least half right in recognising that the polls were telling us something, and that Labor’s victory would be a comfortable one. The real suckers are those who chose to ignore the opinion polls, and in spite of all the indicators, chose to believe in a Liberal victory or a ‘too-close-to-call’ outcome. For many of them, faith in Howard’s supposed infallibility, blinded them to the reality of the situation – and come election eve, came the surprise, shock and even embarrassment of a Rudd Labor victory.
December 11, 2007 at 12:31 pm
Mark, I agree that overoptimism about Labor’s chances had a better basis in evidence that did arguments that the Coalition would or could win or that it would be a “cliffhanger.” I would not have been at all surprised if Labor had won 90 seats. My quarrel was with people who tried to over-analyse the available data and came up with ridiculous findings like Labor winning Wannon, a seat which has NEITHER Howard battlers NOR Rudd liberals – just plain old conservatives.
December 12, 2007 at 11:37 am
Good summary and a good read. Thanks for taking the time to record your views.
Good observations on issues like Haneef. I found it extremely frustrating that Rudd did not attack the Coalition for their gross abuse of power . In hindsight I have to concede that Rudd’s measured approach to limit damage was the best approach. As badly as the coalition handled the Haneef issue – it was probably a net positive.
I think you are a little harsh on the Pollbludger’s forecasts. By my calculation the forecasts can be summarized as (round to the nearest seat):
Mean forecast:87 seats (arithmetic average)
Medium:86 seats (1/2 above, 1/2 below)
Mode: 80 seats (most common forecast)
Standard Deviation (11 seats)
The median forecast is probably the best aggregate measure as it is less affected than the mean can be effect by extravagant forecasts.
So, from a distance of a month from the election the median forecast was within 2 seats of the actually outcome. The median estimate was significantly lower than the number of seat implied by the polls at the time. In other words most bloggers were forecasting a narrowing and overall nailed it to within 2 seats – which, by any reasonable analysis, is very accurate.
In aggregate the PollBludgers correctly forecast a narrowing and only slightly under-estimated it.
The other key statistic is the standard deviation of 11 seats. Without getting too technical, this means the forecasts were all over the shop. In many cases forecasts were a form of wishful thinking (Ashley 146), an act of bravado (Tabitha 0) or an emotional hedge (LETP 68). The accuracy of the “consensus” forecast is even more remarkable given this level of noise.
In your analysis I think you confuse the high level of variability (which is true) as a being a high level of bias (which is false).
December 12, 2007 at 2:31 pm
Adam,
Really interesting and insightful analysis. I got it about right with an 85 seat call, but must admit to a bit of overoptimism as that was my bare minimum call. A case of me falling for the educated professional groupthink perhaps. I’m also an outer suburban mortgage belter, so that was probably why I didn’t get carried away with 100+ seat predictions, despite my sincere belief such a result was thoroughly deserved by the Coalition.
You’re analysis of the “doctor’s wives” is for me most compelling. I think this election will put to bed any lingering question as to whether Labor can really win these seats. They can’t. As you said, the class interest trumps all others. I once heard a quote that “the good folk of Mosman will vote for an ashtray, so long as it was the endorsed Liberal candidate”. An ashtray would have a lot to commend it compared to the member for Warringah, but he will be an MP for as long as he wants to be. If Labor is to gain seats next time (and I think barring disaster they will) it will be the Macarthur’s that they will pick up, not the North Sydney’s.
On your criticism of Possum though I think you may have missed the point slightly. The state based Newspoll analysis was merely based on what the numbers were saying at the time. We cannot know if an election held at that time would or would not have produced a result like that, merely that the poll results correctly broken down gives a certain number of seats. Your experience made you wary and you assumed a narrowing, which of course turned out to be correct, but this doesn’t make Possum wrong as he wasn’t making a final election result prediction, merely interpreting what the poll result would look like in terms of seats. Yes plenty of folks dearly wanted to believe this was an accurate indication of the final election result, but that is another thing entirely. If those numbers had been repeated then Dawson would not have been own it’s own in going to Labor. Your argument should not have been that this or that seat couldn’t fall, but rather that the swings shown in the poll will not be repeated at the election so they won’t fall. A small change of emphasis, but one that might have saved some angst.
Possum’s method however is sound as it merely replaces a single Mackerras pendulum with several sub pendula to predict movements in the measured sub sets of the poll. The general thrust of the argument was actually correct, with the biggest swings occuring safe Liberal seats (on average), the smallest in safe Labor seats, and the marginals somewhere in the middle. The greatest value of this wasn’t so much to say Higgins was going to Labor, but the refute the fallacy that the Coalition would hold on because the big swings would be quarantined to safe Labor seats. As you acknowledge, at least the over-excited Laborites had something other than blind faith to hang their hat on. Those predicting cliffhangers or Coalition wins even after the final polls had better odds with a lottery ticket.
December 12, 2007 at 3:10 pm
Adam,
My first post on your site, and I thoroughly enjoyed your lucid and compelling analysis, only pausing to note that you appear to have ALP experience similar to mine. Accordingly, I put greater stock in Graham Richardson’s prediction of 80 seats, given on the Sunday prior to polling day, than any other pundit or analyst.
Having recently SeaChanged to one of our safest seats, Cunningham, and
done HTV card duty here, I totally concur with your opinion of the North Shore safe Lib seats as I’ve seen their mirror image here as well as next door in Mega-safe Throsby.
These seats are going to remain safe for donkeys’ years, and if Team Howard Ministers didn’t “feel any anger” in the electorate, it’s because they never campaigned in the Illawarra.
Please keep up your scintillating commentary here and on Pollbludger threads.
December 12, 2007 at 3:53 pm
My first post here too.
Bearing booth by booth variations this is a great analysis. It’s disappointing but realistic and inevitable that we’re a nation of hip pockets as opposed to bleeding hearts, as though it’s impossible to be both.
Whatever way you look at it, this was a remarkable election. You rarely see such high levels of emotion invested in politics which is why I would argue that, whether polls are predictive or not, they served an important purpose throughout the year. I think the endless campaign and the possibility of change produced a strong need for certainty on both sides. Possum was probably personally responsible for keeping the consumption of Valium in check, particularly in the last couple of weeks.
From a sociological point of view, I can’t wait for the election survey results. Although that’s not nearly as stomach ulcer-worthy as the election was.
December 12, 2007 at 5:07 pm
Thanks for further comments. Prof Higgins is right about the intensity of class sentiment provoked in strong Labor areas by WorkChoices. In the seat I was campaigning in (outer suburbia rather than an area like the Illawarra) the candidate found real anger while doorknocking in the low-income suburbs, which produced large swings in election day. The middle class suburbs were “concerned” about issues like climate change, but not angry, and in the end it wasn’t enough to produce more than small swings. I think later analysts of this election will remark that Howard dragged Australian politics back 50 years, in the sense that he put class back in the forefront of people’s self-identification and political behaviour in a way that it hasn’t been for many years.
Ratsak says: “We cannot know if an election held at that time would or would not have produced a result like that, merely that the poll results correctly broken down gives a certain number of seats. Your experience made you wary and you assumed a narrowing, which of course turned out to be correct, but this doesn’t make Possum wrong as he wasn’t making a final election result prediction, merely interpreting what the poll result would look like in terms of seats.”
No, we cannot know, but I am absolutely certain that if the election had been held in August or September, when those polls were taken, Labor would NOT have won 112 seats, or anything like it. Polls must be INTERPRETED, not treated as holy writ.
Secondly, I gave Possum every opportunity to agree with me that these polls had little actual predictive value and that Labor would not win seats like Warringah, Wannon and Kooyong. He declined the opportunity, and insisted that the swing in the Coalition safe seats was real. People don’t lie to opinion polls, he said. Well, they may not deliberately lie, but they certainly say things which aren’t true.
December 12, 2007 at 5:57 pm
Adam, the reason I was trying not to answer your question the way you apparenltly wanted me too, other than saying it needs to be treated like a pendulum (so you would stop taking any given result for a seat literally), was simply because the question was a bit of a non-sequitur.Any answer I gave directly to the wrong questions you were asking would have lead you down the path of treating that breakdown as something it wasnt (which you went and did anyway).
What the polls were saying in September was 112 seats. Just like the polls during the last 2 weeks were saying around 91 seats, and the 2 days before the election, 3 of the 4 major polls were saying between 80-90 seats.
Polls arent holy writ (I certainly have never said that – and half of what I do is interpret the polls), but they are the best evidence we have available on the state of political play at any given time.That state isnt static, it’s dynamic – so of course it changes.
Just because the state of play at a given time is X seats doesnt mean that the state of play at future time will also be X seats, that’s just silly.
December 12, 2007 at 7:16 pm
Adam,
Poss doesn’t need me to defend him clearly, but I will take issue with:
“No, we cannot know, but I am absolutely certain that if the election had been held in August or September, when those polls were taken, Labor would NOT have won 112 seats, or anything like it. Polls must be INTERPRETED, not treated as holy writ.”
Actually had the election been held in September and all the polls were supporting a result that reflected the Newspoll breakdown then it is very very likely that Labor would indeed have won 112 seats or something like it.
What you could say with much greater certainty is that if Labor wasn’t actually going to win that many seats, then the polls close to the election would be showing much closer to the actual final result. Therefore the analysis would also be showing much closer to the actual final result.
Yes you were right that the “doctor’s wives” weren’t going to deliver blue ribbon coalition seats the polls were showing were in play a few months out from the election, and the spray of cold water on the more excitable was hardly unwarranted. It was based on good old fashioned experience and common sense. But, these were polls months out from the actual election. If these were actual poll results from the days before the election and you’d made the same comments, I’m sure you would have had to reassess your position afterwards.
I think the best we can say about polling is that the potential error from the final election result increases with distance from the poll, which is hardly an earth shattering insight.
So I suppose the interesting thing is to discuss what is the reason for that? Is it just that the actual election campaign really does have a significant bearing on the result? Is it that rusted on Liberals are using the polls to record a protest they have no intention of actually following through on? Or is it the imfamous undecided voters? Obviously there is elements of all three and more factors, but in what proportions? Is this somehow predictable? If it was it would make interpreting polls a much more useful exercise.
Most interpretation seems to involve little more than projecting internal bias (whether built on experience or wishful thinking) on the part of the commentator. Both yourself and Shanahan (and most others) predicted “the Narrowing”. And behold the narrowing did indeed arrive! But was it inevitable? Possum demonstrated that it was indeed not inevitable in all elections. Was there something special about the polls this time that made it so?
Unless every poll more than one week out from an election is utterly worthless then these are interesting questions to ponder.
December 12, 2007 at 10:54 pm
Adam, in support of your class identification thesis, here’s the astounding Throsby numbers from AEC :
GEORGE, Jennie Re-elected Labor 59,099 73.46 63.82 +9.64
WRIGHT, Stuart Liberal 21,348 26.54 36.18 -9.64
A redistribution had reduced George’s 2004 margin down 1.2% from 15% to 13.8%.
Throsby electorate includes the significant increase in the number of new McMansions in Shellharbour.
December 13, 2007 at 2:00 am
Adam
I have not read your blog before but have spent over 2 hours now reading your various ‘comments’ over the campaign
We have clashed very strongly over Hicks where I believe his treatment
was shabby and was counter to our rule of law & judicial practice.
I would have handled this idiot differently.
Howard’s & Bush’s abandoning what separates us from the terrorists
stupidly turned the idiot David Hicks into an unjustified martyr forever
Election Predictions
I started the day believing Possum’s qtrly Newspoll ‘ratios formula’ was brand new today
(your site in fact tells me it is a repeat of his October blog)
and further I thought this ‘ratio formula’ was a forecast seat predictor.
However when I read Possum’s blog today he argued his qtrly Newspoll ‘ratio formula’ was SIMPLY a translation of a poll at a POINT of time into the number of seats that poll represented at that same POINT of time
Later I asked Possum what were the number of seats Labor would have won using his ‘ratio formula’ based on the September qtrly Newspoll data
He replied 112 seats
The question therefore is : is the qtrly Newspoll ratio formula a forecast/predictor or just a translation of poll stats into seats at that time
Adam I believe you made the mistake of picking the former
Then possum has demonstrated why you were wrong
I believe the ‘ratio formula’ is both !!! …and both at the same time
which I feel is the flaw in possum’s ‘ratio formula’
WHY ? well over time I may be able to work out how to say this a little more precisely , but here goes
Firstly , if the resultant TYPE of ’seats won’ are patently unbelievable then the blame goes back to Newspoll…it was not done properly or is flawed
Secondly , if the number of ’seats won’ is unrealistic then the argument is well that would have been the result if the election had of been held then
(an argument that can not be disproved….or indeed proved)
Thirdly , the Quarterly Poll figures may be flawed because
a/ they are cum. quarterly with big time distances & factors between the
starting and finishing dates of the polls cumulated
b/ the state by state samples being small may be subject to more
inaccuracies than the “National” headline poll % which is based on a
large sample
c/ the problem of a small sample is worsened by breaking samples further
by state into 3 seats categories (safe ALP , safe Lib and Marginal
Four , the actual principle of applying a varying polled resultant % of each
of the 3 categories of seats and making it a factor of the polled result
% of the each whole state polled result % assumes a uniformity of %’s
which is contrary to election history of varying swing %’s between seats
of the same supposed type
and further is contrary to election history of varying % swings even
between various booths within each seat
and further the actual justification of factoring two different polling sets
of numbers together ie. state swings into the 3 seat categories by state
needs substantiation
do you have any view
December 13, 2007 at 9:46 am
and sixly any statistical measure which purports to merely translate poll
result % ’s into specific seats ‘to make the % poll result more
understandable /layman friendly ,
MUST SUBSTANTIALLY produce
ONLY realistic seats that would be won
and if it fails that fundamental test then it has failed its core purpose
meaning the statistical measure itself (not the poll data) is flawed