Kiwiblog

Rob Salmond at Pundit looks at the polls, and linear projection of where the parties may end up on election night, being:

• National 50.8%
• Labour 24.8%
• Greens 14.6%
• NZ First 3.8%

He comments further:

Using the analysis above, we project that New Zealand First will get 3.8% of the vote. We can also project with 95% confidence that the party will get between 2.3% and 5.2% of the vote. Importantly, we project that there is only a 4% probability New Zealand First will get more than five percent of the vote.

We project that National will get an absolute majority of the vote, but a pretty small one (50.8%). Based on this analysis, we think there is a 74% chance that National will get over 50% of the vote.

Of course this is all based on the trend of the last few weeks continuing in the last week, and it may not continue if events get in the way.

I am staggered (but not surprised) that the SST devotes its front page to the results of the Horizon poll, despite the fact that its results are miles out of line with every other pollster, and that certain parties and activists openly recruit their supporters to sign up for the Horizon poll, as it is an Internet panel poll.

Even worse, the SST doesn’t even disclose that the Horizon poll is not a phone poll, but an Internet panel poll.

The spreadsheet above compares the Horizon poll to the weighted average of the five main pollsters. The differences are huge and massive. The SST would know this. Yet they still made this poll their front page, without even disclosing the difference in methodology.

Even worse the SST (at least online) doesn’t disclose the actual percentages for the big parties. I assume this is deliberate because they know if they said National is on 35% only, everyone would laugh.

Incidentially the HoS had a poll done this week over four days by Key Research. Their results are National 55.6%, Labour 26.2%, Green 11.3%, NZ First 3.6%. That is pretty close to the other five pollsters, but again shows how vastly different the Horizon results are.

Roy Morgan have just released their latest poll. Labour drop to a record low of 24.5%, National stay at 53% and NZ First drops to 3.0%.

Normally there is a week delay between a RM poll ending and the release of results. However this one goes right up until yesterday. It covers a 12 days period so around half is before and after the tape incident.

In the sidebar is the latest weighted average of the public polls (only counting the latest poll per pollster). National is projected to be on 65 seats and Labour 34. That means their caucus reducing by a quarter. Their average poll rating is 26.5% so really for Labour a good result now would be 30%.

NZ First on the five polls is 3.2%, but people will be looking for the first poll done entirely after his showboating.

For now I have assumed that ACT wins Epsom. The latest One News poll had Banks 11% behind Goldsmith but I am reminded that no public poll has ever shown ACT winning Epsom and they have won it twice before. However if you assume ACT do not make it, then National and the Greens will gain one seat each.

There were three polls out last night and this morning. The table below shows the date and time weighted average of the polls.

The first set of figures is averaging the results of the last polls by every public pollsters – One News, 3 News, NZ Herald, Fairfax, Roy Morgan.

The second set of figures is the average just for the three polls released last night or this morning, being One News, 3 News and the NZ Herald.

On this average, you need 62 seats to govern.

My Herald column looks at what happens if Winston decides again.

This means that Winston Peters will decide who gets to be Prime Minister of New Zealand for the third time out of six MMP elections. In 1996 he chose Jim Bolger over Helen Clark, in 2005 he chose Helen Clark over Don Brash and if he holds the balance of power in 2011, make no mistake he will choose Phil Goff over John Key, and there will be a Government that can only pass a law if it can get the Greens, Winston, Hone, and the Maori Party to all agree to it.

And imagine the blowout in spending and debt!

Again polls have shown a certain reluctance for National voters to tactically vote for ACT. I speak often to many National supporters in Epsom. To a person, they all want National to have a coalition partner to the right (economically) of National. The debate is whether ACT in its crippled state is worth saving, or whether you do the humane thing and put it to sleep with some electoral euthanasia, allowing a new party to arise phoenix like from the ashes.

The prospect of Winston Peters installing Phil Goff as Prime Minister should be sufficient to resolve that debate. If they do not vote for John Banks, then a change of Government becomes significantly more likely.

Epsom voters now have a clear choice.

MMP is perfect for demagogues such as Peters. He selects who will be on his party list, and they become MPs based on his personal popularity, despite the fact 99% of New Zealanders could not tell you who the top six candidates on his list are. Their loyalty is purely to him, not to the New Zealand public.

STV will still deliver broadly proportional results, but candidates will have to actually be someone whom voters rank high enough with their ballots, to elect to Parliament. This should result in a significant improvement of the quality of candidates, if there is no backdoor through a party list.

I will blog more fully next week on the merits of STV.

KB reader Nik has sent in an updated version of his graph measuring the difference between the average of polls released in the two days before an election, and the actual results. His changes are based on reader comments.

You can see that the difference have been fairly minor – generally within 1.5% each way. National and Greens tend to do worse than the polls show by around 1% to 1.5%.

Despite what Winston claims, the polls for the last two elections have had him within 0.5%. In fact in 2005 his result was slightly less than the polls.

I’ve blogged at Curiablog, the results of tonight’s 3 News poll. Nats up, Lab down, Greens up, ACT down.

The weighted average of the five public polls released this week is:

• National 52.6%, 67 seats
• Labour 27.8%, 35 seats
• Green 11.1%, 14 seats
• NZ First 3,2%, 0 seats
• Maori 1.5%, 3 seats (1 overhang)
• ACT 1.1%, 1 seat
• Mana 0.7%, 1 seat
• United 0.4%, 1 seat (overhang)

There is no doubt NZ First has increased its support. It usually does when they get in the news, and ironically as much as Winston loves to pretend there is a media conspiracy against him, they in fact give him and NZ First far more publicity (but not scrutiny) than other parties polling at his level.

So they are up in the polls, but are they on the verge of 5% as Roy Morgan had them? No, or not yet, in my opinion.

Roy Morgan often has had NZ First higher than the other polls. They had them at 4.5% in May 2011, 5% in April 2011, 5.5% in Jan 2011, 4.5% in Sep 2010, 4.5% in Aug 2010, and prior to the last election had them twice at 6.5% in July and Aug 2008.

As always, I recommend looking at the average of the polls. This had them at 2.2% in October and rising to 3.1% in November. So as I said an increase, but still around 50,000 votes short of 5%.

The Press reports:

Waimakariri Labour MP Clayton Cosgrove faces a crushing defeat as a new poll shows his National rival, Kate Wilkinson, has opened up a huge lead in the seat.

A Fairfax Media-Research International snap poll in the seat put Wilkinson on 53.9 per cent of decided voters against 36 per cent for Cosgrove. The only consolation for Cosgrove is that the poll of 250 voters had a margin of error of 6.2 per cent and a high proportion of undecided voters – 23.5 per cent.

The poll also showed a huge lead for National of 71.9 per cent to Labour’s 17.4 per cent.

That compares with 2008 when National won the party vote in the seat by 49.4 per cent to 33.9 per cent for Labour.

Paul Epplett of Research International said his best explanation for the huge shift in support was related to the earthquake and polling on that had been positive for National.

This will worry Labour greatly, to be 18% behind in a seat they currently hold. Having said that, the party vote figures look a bit extreme to me, so I wouldn’t take the exact margin as gospel.

But what will be worrying Labour, is if one seat they hold now has them 18% behind, what about the others? Their next most marginal seats are Rimutaka, Christchurch Central, Palmerston North, Wellington Central, New Lynn, Hutt South and Te Atatu.

UPDATE: Whale has a post pointing out that the woman in the advertisements Cosgrove is running who says “Why I’m voting for Clayton Cosgrove” is not even registered to vote in Waimakariri – she is on the Te Tai Tonga roll. He couldn’t even find someone from his electorate to endorse him so he got a union official

The Dom Post reports:

The Fairfax Media-Research International mini poll shows the UnitedFuture leader, who has held the north Wellington seat for the past 27 years, could be ousted by Labour’s Charles Chauvel.

It put Mr Dunne on 37.4 per cent, less than two points ahead of Labour’s Charles Chauvel on 35.6.

National’s Katrina Shanks is a distant third on 19 per cent and the Greens candidate Gareth Hughes registered just 1.4 per cent.

A massive 34.6 per cent of voters were undecided.

The poll, conducted on Wednesday night, surveyed 163 voters and has a margin of error of 7 per cent.

Research International spokesman Paul Epplett said that means the race for Ohariu is too close to call.

Definitely too close too call. The probability that Dunne is in fact ahead of Chauvel is only 61%. Add onto that the 35% undecided, and either Dunne or Chauvel can win.

What will be interesting to see is if the 19% voting Shanks, now vote tactically.

At Stuff I blog on who would be in and out of Parliament based on the average of the polls this week:

If National gains 65 seats, they will gain many new MPs. Highly placed candidates Jian Yang, Paul Goldsmith and Alfred Ngaro were always going to make it in, as were electorate candidates Simon O’Connor, Maggie Barry, Ian McKelvie, Mark Mitchell, Mark Sabin and Scott Simpson. Joining them would be candidates Paul Foster-Bell, Claudette Hauiti, Jo Hayes and Leonie Hapeta.

This would give National its most ethnically diverse caucus ever. They would have 11 Maori MPs, three Asian MPs and two Pacific MPs. They would also have a record 18 female MPs (but their proportion of the caucus would be unchanged).

What people may find amusing is I made a typo in the original, and it read “give National its most ethically diverse caucus”

I’ve blogged details of the latest One News Colmar Brunton poll at curiablog.

Labour have dropped to 28%. There are 16 days to go. I expect this will increase their Key Derangement Syndrome and they will try and blame him for the Pike River deaths, as one left wing blog tried to do today.

David Winter at The Atavism blogs on how different pollsters vary from the average of all the polls, to see which ones have a polling bias. I wouldn’t use the term bias, as that is a loaded word, but it is useful to see the variation. The better comparison will be after the election, against the actual results. However it is still interesting to see this graph:

David’s conclusion is:

It seems TV3 tends to get National a little higher and Labour a little lower than the rest of the pollsters, and perhaps the Herald goes the other way. I think trying to gain anything more meaningful than that from these results is probably the statistical equivalent of reading tea leaves, but feel free to stare at the graph and confirm your own political biases!

As I said, am looking forward to comparing after the election.

The graph below was not done by me, but e-mailed to me by a friend, so I can’t vouch 100% for it, but I took a quick look at the data and it looks right to me.

The graph compares the average of the polls published just prior to a general election with the actual result. It shows that the polls on average have been consistently close to the actual result for the last three elections. Not perfect, but generally within 1.5%.

Past performance is no guarantee of future results, but it is a fairly good indicator.

Stuff reports:

National is looking unstoppable on its way to a historic outright election win as Labour plunges in a Fairfax Media-Research International Poll.

The poll has National on 52.5 per cent, and Labour sliding to 25.9 per cent, 17 days from the election.

Labour would lose 10 seats if those results were repeated on election night, while National would get nine extra seats, and bring in a slew of new faces.

The Green Party has profited most from Labour’s slump, rising to 12.6 per cent in today’s poll – which would give it seven more seats in Parliament and nine new faces because of retirements.

Labour have dropped 5.4% since their last poll, taking them halfway to the 21% National got in 2002. The poll covers the disaster of a debate for Goff last week, but would not cover Labour’s pledge to give beneficiaries with a dependent child an extra $70/week.

What is interesting is that Labour are now at only twice the support of the Greens. In 2008 they got five times as many votes.

On this poll Labour gets 33 seats. They could well win 24 electorates, which if so means only nine list MPs. This means they would lose Rajen Prasad, Raymond Huo, Carol Beaumont, Kelvin Davis, Carmel Sepuloni, Rick Barker, Stuart Nash and Steve Chadwick.

At that level of party vote, Labour would also have to worry about some of its marginal seats. If Phil Twyford does not win Te Atatu, or Chris Hipkins loses Rimutaka, or Iain Lees-Galloway loses Palmerston North, then they would not come back in on the list as they are ranked too low. Instead Rajen Prasad would get back in.

A result on this poll would be bad enough, but if Labour loses another 3%, then Shane Jones and Andrew Little miss out, as well as Darien Fenton and Moana Mackey.

The Herald each election does a Mood of the Nation survey when they get reporters onto the street and interview around 500 people. I think this generally is a good initiative. They report that it is unscientific, but it allows them to use actual quotes from people to use in their stories. So long as they stress the survey is unscientific, and the focus is more on what people said rather than the numerical results, I think that is fine.

But you can take these street surveys too far. Off memory, in 2008 a street survey of 100 people was reported by the Herald with great prominence as showing Judith Tizard retaining her seat. She lost of course.

This time we have a street survey of just 47 people in Epsom, that generates a story. It says:

John Banks has some support in the wealthy suburb of Remuera, but is less popular on the liberal fringes of the Epsom electorate, according to a Herald street survey.

A poll of 47 Epsom voters yesterday found the National candidate ahead of Act’s Mr Banks by 22 votes to 20.

It is ridiculous to do a story on a street poll of 47. First of all, you have the sampling problem – that in fact it is a poll of people who happened to be out on a street – in no way random.

But even if you overlook the fact it is a street poll, the sample size is ridiculously low.  The margin of error is 14.7%!

I generally regard 300 as the minimum acceptable for an electorate poll. That gives a 5.8% margin of error. A sample of 47 is close to useless.

Colin Craig’s Conservative Party has been claiming they will win Rodney, on the basis of this poll report, which is on their website. The key extract is:

47.2% of those who had decided who they were likely to vote for as an electorate candidate would vote for Craig.

This would position Craig in first place in the electorate, ahead of ‘The National Party Candidate/Mark Mitchell/Lockwood Smith (36.3%).

The polling was done by Research First. Whale recently revealed their director is a candidate for the Conservative Party. I commented at the time:

This does not mean that the poll results are or are not valid. As I said, validity is based on knowing the question asked and the methodology. You can poll for an organisation you are involved with. For example, Curia does an annual poll for the Republican Movement on whether people want NZ to become a republic when the Queen dies. Now I am on the Council of the Movement, but this doesn’t influence the results. The key thing is I have publicly disclosed my involvement.

Now in response to a request from Whale, Research First have released details of the questions they asked. It is good they have done so, because as I said the exact questions asked are often vital to interpreting a poll’s results. Their response says:

 Relevant questions included the following, in order of being asked:

1 For your party vote, have you decided who you will be voting for in the election?

2 Which party do you currently intend to vote for?

3 Have you heard of the Conservative Party?

4 Have you heard of Colin Craig?

Participants were read a brief preamble to provide context: ‘Colin Craig is the leader of the Conservative Party of New Zealand. In the Auckland mayoral election, Colin came third with over 40,000 votes’. Then asked…

5 If Colin were to stand in as a candidate in your electorate, what would be the likelihood that you would vote for Colin to be your member of parliament (on a scale of 1 = definitely; 2 = likely; 3 = neutral; 4 = unlikely and 5 = very unlikely)?

Those who identified they were neutral or unlikely to vote for Craig were asked:

6 Who do you intend to vote for?

Okay, let’s take this step by step. The first two questions are pretty standard. Then there are two specific question asking awareness of the Conservative Party and Colin Craig. Then a statement was read out which puts Craig in a positive light (mentioning his votes in the Auckland Mayoral election), and then they ask people how likely it is they will vote for Craig, and only if they say they are neutral or unlikely to they even ask you who else you will vote for.

The results are no surprise, once you realise this is the questions that were asked, and in what order. You have a number of factors here influencing the responses, namely:

• The mention of the Conservative Party and Colin Craig first
• The description of Colin Craig provided to respondents
• The question only asked about voting for Colin Craig, with no mention of anyone else
• Only if you say you are neutral or unlikely to vote Craig, do you even get asked whom else you might vote for
• The other candidates are unprompted, so you are comparing unprompted results vs a prompted result.

I am surprised that Research First did not insist on these questions being included in their report, as in my opinion they are quite vital to it. I also think it is unwise to compare answers to a prompted question to answers to an unprompted question.

If I was wanting to poll that seat, and get a result which was fairly trying to ascertain support, the questions I would use are either:

Which candidate, or party’s candidate are you likely to vote for with your electorate vote?

or

The candidates for Rodney are Colin Craig, Conservative; Beth Houlbrooke, ACT; tracey Martin, New Zealand First; Mark Mitchell, National; Terea Moore, Green and Christine Rose, Labour. Which candidate are you likely to vote for with your electorate vote?

One might also have a follow up lean question for those undecided. I very strongly suspect that the results to the questions above would be vastly different to the results of the poll commissioned by the Conservative Party.

There is never any 100% correct version of a question, and rarely 100% incorrect version. In terms of ascertaining potential support for Colin Craig, those questions may be legitimate if commissioned for internal use only. But what I think was wrong was to have them publicly reported as Craig being “in first place”. The questions should have been reported.

The lesson for media here, is to always ask for the questions. Those media who reported the poll, should be wary of doing so in future without checking.

At Curiablog I have full results from the Te Karere Digipoll of Tamaki Makaurau.

Sharples is 35% ahead of Shane Jones.  That will be a landslide, if an actual result. This was a seat Labour boasted of winning.

Four times as many voters cite Key over Goff as their Preferred PM. And this is amongst mainly poor urban Maori in South Auckland. The areas in the electorate which have more than 1,000 voters are:

• Clendon Park
• Mangerefour time
• Mangere East
• Manurewa
• Otahuhu
• Otara
• Papatoetoe
• Weymouth

So even in Mangere, Manurewa and Otara, four times as many Maori roll voters are saying they prefer the National Party leader to the Labour Party leader.

This is somewhat akin to a Republican candidate being four times more popular in Detroit as a Democrat. National only got 7% of the vote here in 2008, yet Key is four times the support of Goff to be PM.

Andrew Balemi (Auck Uni Stats Dept) at Stats Chat blogs on polls:

But let’s go back to the Horizon panel poll mentioned above. It claims that it’s to be trusted as it has sampled from a large population of potential panellists who have been recruited and can win prizes for participation. The Horizon poll adjusts for any biases by reweighting the sample so that it’s more like the underlying New Zealand adult population – which is good practice in general.

However, the trouble is this large sampling frame of potential panellists have been self-selected. So who do they represent?

To illustrate, it’s hard to imagine people from more affluent areas feeling the need to get rewards for being on a panel. Also, you enrol via the internet and clearly this is biased towards IT-savvy people. Here the sampling frame is biased, with little or no known way to adjust for any biases bought about from this self-selection problem. They may be weighted to look like the population but they may be fundamentally different in their political outlook.

Panel polls are being increasingly used by market researchers and polling companies. With online panel polls it’s easier to obtain samples, collect information and transfer it, without all the bother involved in traditional polling techniques like CATI.

I believe the industry has been seduced by these features at the expense of representativeness – the bedrock of all inference. Until such time as we can ensure representativeness, I remain sceptical about any claims from panel polls.

I believe the much-maligned telephone (CATI) interviewing, which is by no means perfect, still remains the best of a bad lot.

In my Herald column I look at why Labour did not rise in the weekly Herald Digipoll:

I think the reason is because the public are not convinced Phil Goff and Labour believe in them (their policies) themselves. This is hugely important to voters. They will vote for a party if they believe the party is sincere and believes in its policies – even if they themselves do not agree with them all. …

So if you change you policies and your position, especially from just three months earlier, you need to make the case for why you have changed your stance.

The same applies to Labour’s stance on the Government’s books. Labour spent two and a half year attacking every single decision the Government took to reduce spending to get New Zealand out of the projected permanent deficit that Treasury projected at the end of 2008. They not only opposed every spending cut that the Government made, they actually argued that the Government should be spending massively more like Obama did. Right up until 2011 they were arguing that it was more important to borrow and spend to (temporarily) boost the domestic economy than to keep a lid on spending.

Then a few months ago, they realised that the voters were becoming very worried about debt, in light of the on-going troubles in Europe, and suddenly Labour stopped arguing that the Government should spend more, and said that they would spend and borrow pretty much the same as National was.

Labour’s problem again, is whether voters will believe them. It doesn’t really matter that much what numbers they announce today as their fiscal plan. The bigger issue is whether voters actually believe that the current Labour caucus really believes in spending restraint, when they have spent two and a half years arguing against it at every opportunity.

The full column is at the NZ Herald.

Labour told us how they kicked arse in the first two weeks of the campaign. Their opening broadcast was better, they won the debates, their policies were better, and people now hated John Key as much as they did.  National would be hammered for its unpopular policies.

So why did Labour drop to its lowest ever poll rating in the Herald Digipoll? A poll taken up until yesterday so doesn’t even include Goff in The Press debate.

And in the One News Colmar Brunton poll, National retains a 26% lead. This poll also taken in the last week.

What happened?

Two useful posts by Whale on a couple of polls.

First he comments on the Horizon poll that the Maori Party is encouraging their activists to sign up for the poll. It is quite possible they are not the only minor party to be doing so. This is one of the weaknesses in allowing people to sign up for a poll.

It can make a lot of sense for a minor party to do this. If the poll is of 1,000 people, then just 10 activists can get you an extra 1% if they are all selected for the poll. This is why it is desirable to have a huge sample size for your online panel (YouGov in the UK have hundreds of thousands), so you can minimise impacts like this.

Whale also comments on the poll done that alleges Colin Craig is leading in Rodney. I had been critical of this reported poll also as the report published by the Conservatives does not include the actual question asked. Anyway, Whale reveals that the director of the firm which did the polling is a candidate for the Conservatives. This should have been revealed at the time, in my opinion.

This does not mean that the poll results are or are not valid. As I said, validity is based on knowing the question asked and the methodology. You can poll for an organisation you are involved with. For example, Curia does an annual poll for the Republican Movement on whether people want NZ to become a republic when the Queen dies. Now I am on the Council of the Movement, but this doesn’t influence the results. The key thing is I have publicly disclosed my involvement. Incidentally we found this year a quite large drop in support for becoming a republic – I suspect a consequence of the royal wedding. So poll results are not always pleasing to the client.

I blog at Stuff on the October polls, showing the gap which had been widening for six months has started to turn. I comment:

If Parliament is 122 seats again, then a government needs 62 seats to govern. That blue line is at 69 seats. If it drops to below 62, then you will have a hung Parliament. A drop of eight seats is a drop of around 6.5 per cent, so that is effectively the target for the left – to have National drop 6.5 per cent in the last 23 days. Not impossible at all, but easier said than done.

For Labour to be able to form a government in its own right without needing the Maori Party, it needs to gain 12 seats, or around 10 per cent in 23 days. That seems less likely.

If the Maori Party does look as though it will hold the balance of power, could this affect the MMP referendum? Do you think having the Maori Party decide the government would be a fair result?

The issue is how much will National drop in these next 23 days. Here’s the last two months in 2008:

Sep 2008 – 50.6% N and 35.4% L = 15.2% gap
Oct 2008 – 47.5% N and 35.3% L = 12.3% gap
Result – 44.9% N and 34.0% L = 10.9% gap

Labour need the gap to close by 13% from 23.6% to around 10%, assuming all else stays constant (Greens at 10%, no NZF etc).

Te Karere have just released a Digipoll for Tea Tai Tokerau electorate. I have full details at Curiablog.

Hone Harawira leads Kelvin Davis by 7%. Margin of error is 5%, which means there is a 95% chance Hone is ahead of Davis.

The party vote is almost a three way tie – Labour 27%, Maori 27%, Mana 25%.

Key is Preferred PM and Harawira most favoured Maori MP.

Oh how the left howled cries of triumph. They said the Rena would crush National. No it was the credit downgrade. No it was John Key’s gesture in Parliament. No it was over-egging what Standard and Poors said.

And then today the Herald Digipoll reports National up 1.5% and Labour down 1.2%.

Mind you while a 23% gap is a good one, it is worth recalling that the important gap is CR-CL which is closer. It still only takes a 5% swing and then a Labour/Green/Winston/Hone coalition becomes a possibility.