Written By:
Tane - Date published:
8:53 pm, October 23rd, 2008 - 47 comments
Categories: polls -
Tags: tns
Tonight’s 3 News poll has the election wide open with two weeks to go, putting the Left on 59 seats to the Right’s 60.
3 News says that with the Maori Party in the kingmaker position it’s too close to call, but when Colin Espiner is promising to print out his blog and eat it live on webcam if the Maori Party goes with National I’d say a broad left LPG+M coalition is the likely outcome based on these results.
The Greens will be happy with 8.8% and are in a good position to improve on it. For middle class voters who are regretting their flirtation with National but still hold a grudge against Labour, the Greens seem a natural choice.
New Zealand First still can’t be written off either. While the party wouldn’t make it back on these results, at 3.5% and climbing there’s a decent chance they’ll reach 5% – or that Winston will win Tauranga. That would almost certainly rule out a National-led government.
So, overall, good news for the Left. Roll on tomorrow’s Roy Morgan poll.
The server will be getting hardware changes this evening starting at 10pm NZDT.
The site will be off line for some hours.
Wow
And this is before the effects of Lockwood Smith’s comments. Until they were made I thought that National had done a really good job on their ethnic campaign but I think that those comments are probably worth a few percentage points on their own.
Arthur Anae’s comments in the Herald today were interesting. The original source is here (I must learn html)
[lprent: Fixed, and you should. Look under FAQ ]
The only parties that, that poll got right was NZ first and the Maori party, not a chance that the greens will get 8.8%, National will end up getting over 50%.
Well if we use an analogy of a marathon race, of which I have run a few, we are at about the 32K mark, and with 10K to go there is still a pack of 4 or 5 running neck and neck.
Who will crack first!
Great stuff really (not this result, but the sport that it is).
Oh and BTW, Brett, National doesn’t have a chance in hell of getting 50%. They’ll be lucky to get 42%. I’m picking the Greens to stay pretty static on about 9% with Labour to fall back a notch too and with Act, UF & Winston to pick up a little.
Of course I could be utterly wrong!
Greens will get 6.5% tops.
Labour between 32-35%
National will crack 50%
NZFirst may not get 5%, but Winnie will win his seat, it wouldnt be the same without him.
Act, United, Prog will have dismal results.
The Maori party will do well, taking votes from Labour.
National will Govern alone, with maybe getting propped up by Hide.
Hang on, I just totally flip-flopped on the Maori party.
Don’t assume Peter Dunne will go with National. He wants to keep his cabinet seat so will go with whoever is in the better position to offer him that. Helen, Michael and he go back a long way.
Don’t disabuse brett of his wee dreams gooner – that’s mean! As for who’s gonna crack first? A little birdie tells me team blue doesn’t have anyone coordinating a nationwide strategy and they’re starting to turn on each other…
This is an interesting result. 2005 showed that the polls bounce around quite markedly in the last couple of weeks. I would be careful about making predictions based on any one poll. The weighted poll average of all polls in 2005 was very close to the final result. It’s fair to say that this poll is out of kilter with the weighted poll average.
It’s probably pertinent also to note that pretty much the only polls that the writers of the Standard seem to rely on are those that are well out of whack with the poll average. I think that’s a hazardous interpretation.
Even including this poll result, the weighted average of polls hasn’t changed markedly. On the weighted poll average National+UF+Act is still a long way ahead of Labour+Greens+Progressives.
I would also point out that there isn’t any evidence that NZ First’s vote is climbing. In the past year, of all the polls published, only four of them have shown NZ First pass the threshhold. Counting on Winston to help deliver a Left majority is probably wishful thinking.
[Tane: Tim, I agree they’ll move around a bit. That’s why I was careful to say “based on these results”. The outcome on election day could be very different.]
Booo!
Tim Ellis
I do not believe that the poll average can be relied upon as the Colmar Brunton result amongst others is flawed. The CB election day poll result for 2005 was National 44, Labour 38. The actual result was Labour 41, National 39. CB was 8 points out.
Do not rely on them. CB’s and other’s reliance on landlines creates a distortion that is worse now than it was in 2005. Poor people are missing from the poll and this is skewing the result.
At the ground level things are just as good for Labour now as they were 3 years ago. After Lockwood’s comments they are looking even better …
I don’t think Winston will get 5%.
On the other hand, he’s just had a boost from the Electoral Commission tonight: they found in NZ First’s favour … and against ACT.
He’ll be rubbing that in Rodney’s face.
These latest poll results are very much spot on with what i predicted in July
Can’t be arsed to find the post
The Left always gets 46 or 47% of the vote why is that so difficult to understand ?
Micky, the rolling poll of all polls in the election period, including the Colmar Brunton poll, was within 0.2% of the major parties. All of the polls during the election period bounced around considerably–including Colmar Brunton, which picked a 3% lead to Labour three weeks before the election. The Herald Digipoll showed a 4% lead to Labour a week before the 2005 election, and a 7% lead to National the day before the election.
You might want to pick and choose which polls you want to pay attention to, but there are sampling errors in every poll. The weighted poll of polls is far less likely to contain sampling errors.
As the election gets closer, the sample sizes will get smaller, and the polls will be conducted over much shorter time periods. This is why the individual polls will jump around so much. I know there is a tendency to dismiss a whole polling company’s results, based simply on one poll against the final result last time, but that’s not credible.
Brett, give us a pint of what you are on
Tim if I remember correctly the weighted poll of 2005 had National+UF+Act winning didn’t it? Also leading into 2005 election the average of most polls were wrong as well weren’t they?
If that were the case, it may have been close but they were still incorrect and IMO it would be reasonable/acceptable that polls used by the writers of The Standar should/could have as much bearing as the TV1, Herald, Colmar Brunton etc…….
If not I stand corrected on my 1st two questions but they are still as relevent
Polls heading the way I expected them to. National dropping votes and Labour picking them up. Greens are looking good and I still think they have a chance at a full 10% of the vote. Will certainly be interesting if Labour break 40% of the vote as it could conceivably result in a Labour/Green government.
hear hear, we should encourage such an optimistic view on life even when it is totally uncalled for. Reminds me of that old saying Love is Blind 😀
Not just poor people but young and single people as well.
why are the commentators saying the balance of power will lie with the MP? Surely it’s more likely the MP will stand back from the fray (unless one or other of the co-leaders scores a knockout in their current spat) and that on these figures (or thereabouts) it’ll be the saintly Mr Dunne?
(captcha = “do jokesters”!)
I apologise I didn’t read this earlier. I have read this before, but I haven’t seen any evidence of it. LP and I had a lengthy discussion about this about a month ago, in which he asserted low land-line use skewed poll results.
I haven’t seen any evidence that this is true. Land line use is actually very high in New Zealand–about 98% of households have land-lines as of the 2006 census. Anecdotally, I believe that many of the domiciles without landlines are inner-city apartments; hardly a bastion of the Left.
All of the published polls weight by gender, age, location, and often ethnicity. So if the population is 50/50 male/female, and a 25/25/25/25 split for 20-30yrs, 30-45yrs, 45-60 years, 60 years+, then the pollsters will be required to get responses from this breakdown. That dramatically reduces sampling error.
It’s just not true that young people are under-represented in polls. I think the evidence that landline access is limited among low-income people is sketchy at best.
Another major objection to polls is that they don’t include the refusals and don’t knows, which typically run at about 30%. It so happens that the non-enrolled population is about 10%, and the enrolled but non-voting population was 18% in 2005, and has been growing over the past 20 years. I haven’t seen an in-depth analysis of this, but I think it’s reasonable to assume that many of the non-responding people polled are unlikely to vote anyway.
They’re coming for you Lockwood.
Jokes.
Anecdotally, I believe that many of the domiciles without landlines are inner-city apartments; hardly a bastion of the Left.
Don’t be too sure. Those without landlines tend to be the younger ones, hardly a bastion of the Right.
Polls reach a particular subset of the population and are full of noise. Some of the issues are discussed here.
TE: Incorrect. Sorry about the length, but I won’t have time to really discuss it due to coding and family pressures until sunday. So I figure I’ll just run through it and let other people argue.
I’ve had more time to look at this while playing with canvassing strategies. As you’re aware, I’m a labour activist with a very strong programming bent, a background in operations research, and a massive lazy/efficiency streak. So I’m really interested in this stuff because it means I can get more effect for less work (much like The Standard does). The same issues apply for pollsters as for canvassing.
The question you are referring to is reported as
. The key word here is ACCESS (which I will keep repeating for a while).
Ummm… I don’t have time to dig out the country wide percentages. This is the prelim report for Hawkes bay. It shows phone access at about 90%.
Access to a land-line is not the number of households with landlines. Those of course could be in the house, they could be next door, or they could be 2 km’s down the road, or they could be at work.
The question is asked to find out how many people could call out. It is totally and utterly irrelevant for pollsters or canvassers wanting to call in.
Now figure out the polling problem if you didn’t have a directory of numbers. The land-line phone numbers are 10^8 for NZ (7 digits + main area code). There are about 1.5 million households, so lets assume that there is a phone at every house. Now lets add in all of the business land-lines numbers, all of those PABX’s using ISDN and VOIP circuits. I don’t have figures on how many phone numbers are in use. But it is a fraction of the available numbers, and it is likely that there are a very high percentage that are not at residences.
Random dialing is difficult because even after you cut out the start sequences that will not work, there are potentially hundreds of lines per residential household. Forget random dialing because even a machine having to work through analogue systems would be too damn slow. It would be a logistical nightmare. It is a lot easier and cheaper to go to http://www.whitepages.co.nz, pick numbers out of phones books, or pay yellow pages for a list of random list of residential numbers.
Finally we get to what I was talking about – listed phone lines. Lines where the phone is listed in the white-pages. It averages about 57-58% of the voter population when you assume that you can get hold of voters using any listed phone at the household. That is effectively access to a listed land-line in the household as far as a pollster or canvasser is concerned.
It is as high as 73% in Ilam, as low as 36% in Mangere. Auckland Central is actually 47%. The other major city centres are average.
There isn’t a major correlation with apartments apart from possibly Auckland Central, where you look at the number of bedrooms in a household as the indicator.
The strongest correlation is with mean household income for an electorate. The lower the mean household income of the electorate using the last census data (now available online to meshblock levels), the lower the number of listed landlines. The higher the average household income, the higher the number of listed lines.
When you look in detail in an electorate you find that areas with very high incomes, drop in the number of listed land-lines. But before you get too interested in that, the numbers of high household income voters are peanuts compared to the electorate as a whole.
The second effect is with age distributions in the population. Someone at age 60-65 is almost twice as likely to have a listed land-line as someone at age 25-30 across almost every electorate I’ve looked at. The graph is classic, the curve turns almost exactly at the 45 year-olds at just below the 60-65 group and flattens out. The age tip point has been moving up by about 3 years each election for the last 4 elections since I noticed it, and the lower slope has been increasing. It is the cellphone and computer generations.
There is no particular correlation with a listed household number and gender.
Ok – so what does this mean in terms of weighting. Well weighting doesn’t work when the sample you’re taking does not represent the population.
Take the extreme case. Lets imagine that we’re looking for someone in Mangere or another low income electorate and who is about 30. Just over a third of of the electorate’s people have listed landlines at their household. For someone aged 30 it is likely to about 20% of the people of that age in that electorate.
Now the main correlation is on income, so the likelyhood is that anyone you do contact aged 30 in Mangere will be quite wealthy compared to their peers. There is a strong probability that they will be conservative compared to their peers because they probably run a business, because that is what you see when canvassing.
But you’re short of 30 years olds in low-income areas so you weight them up. Hey – you just weighted someone up who was quite different to the rest of the peer population. Same thing happens when you take a nation wide poll. You’re likely to get more conservative 30 year-olds. They’re probably more affluent and likely to ve more conservative and techno-phobic. Why? You can tell, because they have a listed land-line.
Essentially the polls are crap because the pollsters now have problems getting hold of low-income, younger people, who use devices other than land-lines for communication.
They weren’t so bad 10 years ago because then even Mangere had close to 50% household phones. Now it is 36% and close to useless for polling. That is the extreme, but the same thing is happening across the whole country, even in the affluent electorates.
Damn I never got on to the self-selecting nature of the people who answer…. Oh well sleep time first. However take it from me that people aren’t that willing to spend time with pollsters, most won’t – in fact generally it is only the most angry/committed that will. They’re of those that do, a high proportion are undecided. Now apply that to the probabilities of finding a repesentative 30 year old in a low income area.
I’d hate to be a pollster. Then I see the treatment the media give these things and I have to laugh. Personally I think that all the MSM should be forced to do some systematic polling for a few weeks so they do know what they’re talking about. At present they take dodgy data that they don’t even understand the risk associated with. They spin it into being even more dodgy in pursuit of a headline and a rationalisation of some numbers.
This morning’s poll from Herald/Digipoll shows that National is comfortably in front and maintaining the 13% advantage that they’ve had for more than a year. TV3’s poll is already yesterday’s news.
Personally i favour Tim Ellis’ view – all polls will tend to have some bias depending on their audience. By taking a weighted average of those polls, you’re more likely to get closer to the real result.
LP,
You’ve made a number of very interesting points. Where did you get the data from the following ?:
Didn’t the TV3 poll have quite a close correlation to the results in the 2005 election?
Whatever… things have probably changed since then anyway, eg with landlines etc. I don’t take any poll results too seriously. The election will show which polls have been most accurate. Then there should be hard questions asked of polling companies that hvse been wide of the mark.
For the umpteenth time – “polls shmolls” wait until the 8th the we can all be wise after the fact. I wonder how much the polling companies are racking up in fees for these polls ?
Most bizarre comment “For middle class voters who are regretting their flirtation with National but still hold a grudge against Labour, the Greens seem a natural choice.”
The greens a natural choice for middle class voters – are you mad ?
“They’re coming for you Lockwood.”
What? small hand people or dirty bottoms?
HS. The Greens’ support base is middle class liberals. Pop along to a Green Party do some time and you’ll see.
Their policies help working class voters and beneficiaries, but they tend to vote Labour for cultural reasons.
I think you’re right, Tane, that the Greens’ support is primarily among middle class liberals, and younger middle class people in particular.
I very much doubt, however, that there’s much flow between National and Green voters. The very liberal middle class just don’t vote National, as a general rule. They switch back and forth between the Greens and Labour.
This poll underlines for me the importance of talking about what Act might bring to the table in a National-led Govt. There’s been lots of talk about all the other parties, and what they might want for supporting Labour or National, but little about Act. IMHO that’s where the attention should go now.
Having done a wee bit of reading of their policies for our policy quickie series I’ve been quite startled by how similar some of Act’s policy is to the Libertarianz. eg around half of their Welfare policy is fixated on the DPB, and in particular promises that they will make those who would need a DPB type benefit draw down on their superannuation contributions first, and then the father’s super second (because of course we know that all people on the DPB are women, whatever), before a final last ditch safety net would be provided by Govt. That strikes me as quite nutty – how would that person then be supported in retirement? A lot of people just don’t seem to realise how extreme some of Act’s policy is, and I’m wondering how much influence they would have on National, seeing as how National would need them to form a Govt…
Most bizarre comment “For middle class voters who are regretting their flirtation with National but still hold a grudge against Labour, the Greens seem a natural choice.’
The greens a natural choice for middle class voters – are you mad ?
I personally think the Green’s choice to annoint Labour is a bad call. I get the feeling (unscientific) that a lot of urban voters saw the Greens as a kind of protest vote against the usual divisive Lab-Nat politics. Now that they’re well and truly entwined, I think they’ll bleed votes.
That said, it’s not immediately apparent where those votes will go. My hunch is, not Labour, as ‘time for a change’ seems like a strong vibe out there in voter-land.
BTW, does anyone else have the horrible feeling the entire Labour campaign has been pulled from the script of “Wag the Dog”? Remember, you don’t change horses mid-stream.
Tane
The vast amount of Middle class NZers will vote for Labour or National as will the bulk of the NZ public – suggesting that a very large proportion of the NZ public will be dissappointed regardless of the way the election goes unless we got a grand colaition of NZ and Labour which is the least likely coalition we’ll see despite the two parties probably being closer in terms of policy than any of the other.
TE: NZLP campaign data. I’m a NZLP activist and a programmer and have a MBA in operations. I get used for any of my available spare time that I’m willing to give up. I’ve been working on better strategies for the on-the-ground campaigns for years (actually more like decades).
Incidentally that is a major reason I’m starting to get pretty tired at present, less on here, and looking forward to the election. Holding a full-time programming job plus doing that plus The Standard chews time like you wouldn’t believe. We’re (partner and I) are both looking forward to the election being finished.
Unfortunately just after the election she is off doing one of her unpaid semi-voluntary things for 6 weeks. Looks like we’ll have time together after xmas.
Tim, have a look at the fine print on the Colmar Brunton polls:
I agree with lprent that TV1 just commission this poll to make news stories; it certainly does not have the predictive value that they ascribe to it, but I don’t think they care … they’ll present it as if it does because they need to be able to speculate on the election outcome with some semblance of authority, and the CB poll gives them that semblance.
Re: the class basis of the Green Party. I would definitely characterise it as a “petiti-bourgeois” party (i.e. representative of the small business class). Not to say that their supporters aren’t mostly working-class, of course (probably true of all NZ parties), but their organisational base in the petit-bourgeoisie is pretty solid. A friend of mine went to a national conference of theirs once and reported to me that every single person he met there was self-employed. He was quite surprised, but I wasn’t really; I think the Green ideology is quite a natural (good) reflection of the petit-bourgeois class.
I don’t say any of this in a derogatory way; I will be voting Green myself.
Tane, do they mention a) how many undecided? b) how many missed, actually attempts?
3 News didn’t – haven’t checked the TNS website, too much else to do.
In general media outlets try to pump up their polls as much as possible, so avoid mentioning any information that might cast doubt.
I seem to recall that 3 did mention the undecideds specifically on the preferred PM question, which was in the teens percentage wise (18% I think?) while Key and Clark were level pegging.
lampie – Anitfa, you fool. He’s not acutally that bad…. yet.
Julie
It would have been interesting (albeit a bit pointless) if they had included Obama and McCain in the question as to which leader you preferred out of Clark, Key, Obama, McCain just a gut feeling but I suspect neither of our two would have come out on top.
think I’ll stick with Goomer. that boy knows the score.
“lampie – Anitfa, you fool. He’s not acutally that bad . yet.”
HAHAHa (thumbs up) 🙂
“lampie – Anitfa, you fool. He?s not acutally that bad?. yet.”
HAHAHa (thumbs up) 🙂