US President polls

A bit has been made of a 538 forecast suggesting a Trump win in the presidential election. But 538 run three separate forecasts:

Now-cast: Who would win an election today?

  • Clinton 45.8%
  • Trump 54.2%


That is straight after the Republican convention but before the Democrat convention.

Polls-only forecast: What polls alone tell us about Nov. 8

  • Clinton 53.2%
  • Trump 46.8%

Polls-plus forecast: What polls, the economy and historical data tell us about Nov. 8


  • Clinton 59.6%
  • Trump 40.4%





UMR polls

More has been posted (at The Standard) on the UMR poll that Martyn Bradbury went crazy over – see “BREAKING EXCLUSIVE SECRET POLL” – by Swordfish, who is a fairly reliable source and who analyses polls far more sensibly than Bomber.

Note however that we still don’t have any actual details published of UMR polls, we are only given a few raw numbers.


The UMRs tend, on average, to be slightly better for Labour and the Greens and slightly worse for National compared to the main Public Polls.

I’d take issue with Bomber’s assertion that: “The latest UMR internal Polling has shown a massive drop in support for Key.” (by “Key”, of course, Bomber means National).

It was typical Bomber bull and bluster.

In reality, the Nats slumped earlier this year (in the UMRs, in National’s Internal (Curia) Poll and, to some extent, in the Roy Morgans) – particularly in the wake of the Flag Referendumand the Panama Papers controversy – and since then it’s simply experienced a slow decline:

April 2016
Nat 43%
Lab 30%

June 2016
Nat 42%
Lab 28%
Green 16%
NZF 10%

July 2016
Nat 41%
Lab 33%
Green 12%
NZF 10%

So the latest UMR poll shows a the barest of declines for National (that could be a fraction of a percent and well within the presumed margin of errror) and a more significant shift of support from Greens to Labour Greens down 5, Labour up 4).

We’re in a similar situation to mid-2015 when a couple of the Roy Morgans and 3 UMRsplaced combined Lab+Green support a little ahead of the Nats.
(although – unlike the last few UMRs – the recent RMs still record a mild Nat lead over Lab+Green).

Then a swing back to National and the Right in the later months of 2015 / early 2016 – and now a swing back towards the Left (and, of course, NZF).

So these poll swings and round abouts are not unusual. Neither is Bomber’s frantic ranting.

Going by these numbers National should be getting a bit concerned but they shouldn’t need opposition polls to tell them that, it’s been obvious they have been having problems.

The next Roy Morgan poll which may be out next week will give us a better idea of how much weight can be put on the UMR poll results.

Journalists blame polls for their wrong guesses

This may seen repetitive, but again polls have been blamed for being wrong over the Australian election and for predicting the outcome incorrectly, this time by political journalist Tracey Watkins at Stuff, who should know better.


And journalists shouldn’t report on polls as if they are supposed to be predictors. Polls and pollsters should not be trying to predict unknown votes in the future.

In Lessons for Key in Australian election? Watkins wrote:

Australia has been plunged into political turmoil, its election result in the balance, Aussie prime minister Malcolm Turnbull fighting for his political survival. Sound familiar?  

Whether it’s Britain, the United States, and now Australia, voters are defying convention, the expert predictions, and even the polls.

Voters may well have defied the polls (whether voters are effectively telling both politicians and polls to get stuffed is another issue) but mentions the polls alongside ‘expert predictions’ when they are totally different things. Or they should be different things.

But a chill must have passed down John Key’s spine all the same.

If New Zealand follows the new world order he could be out of a job little more than a year from now.

Of course, the polls all say otherwise and Labour is hardly looking threatening. And yet.

The polls don’t say anything about how people may vote in over a year. They try to measure opinion accurately at the time they are taken, and by the time they are published that is always in the past.

And polls have known margins of error and even those have a lack of confidence in being 100% approximately accurate. They have a statistical 1 in 20 chance of being further askew.

The polls predicted Britain’s Brexit vote would go the other way.

The polls didn’t predict the Brexit vote incorrectly, they didn’t predict anything.

It is journalists and pundits who use polls to try to predict election outcomes. This is lazy and it is prone to a much higher margin of error than statistical polls.

And when journalists and pundits get their predictions wrong they blame the polls.

They are using polls as scapegoats for their own failings if they are themselves trying to predict the outcome of an election or referendum.

Polls may give us a useful indication of trends in public opinion, but the closer you get to an election the greater the chance of polls being seen to ‘get it wrong’.

Polls don’t and can’t measure the opinion of ‘undecideds’. Which way undecideds vote can swing markedly in the weeks and even days leading up to an election. Polls are too slow to reflect significant late changes and are looking back anyway, not forward.

Understanding how polling works is basic stuff that political journalists should understand. But they are either ignorant of polling 101, or they choose to misuse polls.

Take a Financial Times Brexit poll of polls as an example of what they polls are telling us and what they aren’t telling us.


Some journalists may take that as a prediction of the referendum result but 48-46 is nowhere near as precise as it may appear.

Combined polls like this are usually weighted to give greater prominence to more recent polls but even those are already history. Opinions may have moved on already and often will have.

48 versus 46 in a poll doesn’t actually mean 48 versus 46. There are a number of inaccuracies.

  • It is rounded, so 48 could be anywhere from 47.51 to 48.49 and 46 could be anywhere from 45.51 to 46.49.
  • Typical margins of error are 3-4 % so 48 plus margins of error mean it could be anywhere between 44% to 52%, and 46 could mean 42-50%
  • Add the rounding and 48 could be 43.51% to 52.49%, 46 could be 41.51% to 50.49%.
  • Statistically there’s a 1 in 20 chance those results could be more inaccurate.
  • There could be typically up to 10% or more of ‘undecided’ and ‘refused to say’ responses.
  • Opinions could and often do change the closer you get to an election.

And this shows that some people vote tactically (differently to what they want), so may vote differently to what they would tell a pollster.

And there may be some people who deliberately give pollsters false opinions as a form of protest at polls or politicians.

Polls have known inaccuracies – and these are significant when opinion is evenly divided. And  they don’t predict election results.

Some journalists try to predict election results – and either ignorantly or deliberately blame polls for their inaccuracies.

Uk referendum too close to call

Voting in the referendum that will decide whether the UK stays in or leaves the European Union is on Thursday (in the UK). Results are expected to be known some time on Friday New Zealand time.

Newshub says partial results will be known about 1 am tomorrow but don’t make it clear if that’s NZ time or not, as that would be half way through the day on Thursday in England.

The latest poll suggests that the result is too close to call.

Newshub: Brexit polls tight on eve of referendum

“It’s very close; nobody knows what’s going to happen,” Cameron told Wednesday’s Financial Times.

The latest Opinium poll:

  • Leave 45%
  • Remain 44%
  • Undecided 9%

There’s nothing in that result. as it’s rounded it could be as close as 45.51%-44.49% or as wide as 45.49%-43-51% but regardless that’s statistically not a significant difference, well within the margins of error.

The result is likely to be decided by which way undecideds swing, if they vote, and which side is more motivated to vote.

More EU referendum polling

Another poll on the UK referendum on the European Union showing a swing to ‘leave’:

EU referendum: leave takes six-point lead in Guardian/ICM polls

Support for leaving the EU is strengthening, with phone and online surveys reporting a six-point lead, according to a pair of Guardian/ICM polls.

Leave now enjoys a 53%-47% advantage once “don’t knows” are excluded, according to research conducted over the weekend, compared with a 52%-48% split reported by ICM a fortnight ago.

That’s a statistically insignificant shift but reinforces indications that the vote may favour leaving the EU.

Recently  an ORB INTERNATIONAL POLL FOR THE INDEPENDENT 10TH JUNE 2016 had ‘leave’ at 55% and ‘stay’ at 45%.

Prof John Curtice of Strathclyde University, who analyses available referendum polling data on his website, noted that after the ICM data, the running average “poll of polls” would stand at 52% for leave and 48% for remain, the first time leave has been in such a strong position.

It’s going to be difficult for the ‘stay’ supporters to turn this around. Both major parties in the UK support and promote staying in the EU.

UK poll shows swing to exit the EU

Less than two weeks before the referndum an ORB poll in the UK shows a significant swing towards ‘Brexit’ or leaving the European Union.

  • Leave the EU: 55% (up 4 since the last ORB poll)
  • Stay in the EU: 45% (down 4)



The online poll of 2052 respondents, conducted from June 8th to 9th, shows a swing to Brexit with just 13 days of the campaign remaining. 

Other key findings include: 

  • Over 80% of respondents agree that leaving the EU would pose either some risk or a great deal of risk to the UK, vs only 19% for no risk at all. 
  • More than two thirds agree that the campaign has been too negative. 
  • The economy is a bigger issue than immigration for more than half of our sample (52%), with only 37% disagreeing. 
  • Two out of five believe that the result of the EU referendum will not have much effect on their everyday life. 

The Independent adds in EU Referendum: Massive swing to Brexit – with just 12 days to go

The campaign to take Britain out of the EU has opened up a remarkable 10-point lead over the Remain camp, according to an exclusive poll for The Independent.

These figures are weighted to take account of people’s likelihood to vote. It is by far the biggest lead the Leave camp has enjoyed since ORB began polling the EU issue for The Independent a year ago, when it was Remain who enjoyed a 10-point lead. Now the tables have turned.

Even when the findings are not weighted for turnout, Leave is on 53 per cent (up three points since April) and Remain on 47 per cent (down three). The online poll, taken on Wednesday and Thursday, suggests the Out camp has achieved momentum at the critical time ahead of the 23 June referendum.

Differential turnout could prove crucial. ORB found that 78 per cent of Leave supporters say they will definitely vote – describing themselves as a “10” on a scale of 0-10, while only 66 per cent of Remain supporters say the same.

It could be hard to turn that lead and that momentum around.

Voters would switch to keep Greens out

I found an interesting poll snippet from the past (June 2002) in Boxing Bill loses charity bout:

Meanwhile, a new poll shows that four out of ten National voters are seriously considering switching to Labour to keep the Greens out of government.

I have found more detail in a Victoria University Press book “New Zealand Votes The General Election 2002”.

Apart from polling published by media agencies, political parties carry out their own research into the views of voters. UMR Insight is Labour’s pollster, as well as conducting a regular poll for the National Business Review.

A UMR poll commissioned by Labour in May found that 39 percent of National voters, seeing their own party in disarray, were willing to ‘do the unthinkable’ and vote Labour to keep Greens out of government.

This finding was leaked and mistakenly reported by other media as part of a National Business Review poll (Evening Post, 7 June 2002) but it supported anecdotal feedback from voters.

I wonder if Labour have done any similar polling lately. Nothing appears to have been leaked about anything like this to Martyn Bradbury.

That is a long time ago, 14 years, and much will have changed since then. It is now Labour that looks in disarray  (and has done to varying degrees for the last 8 years). Greens have changed considerably and are now well established with a significantly higher level of support.

But anecdotal feedback suggests that there are still a voter sentiment of ‘keep Greens out of government’.

And Greens have been kept out of Government, not just by voters but notably by Labour and NZ First in 2005.

So the possibility of vote shifting to Labour, or to NZ First, or to National is still there.

And now Labour of Greens have indicated a close campaign arrangement and an implied but unstated intent to jointly form a government if they get sufficient votes it raises another possibility – that voters shift from Labour to National or NZ First to keep Greens out of government.

Polls over the next few months may give some indication of this possibility – or not. Voter leanings this far out from an election can be quite different to campaign time when voters assess the possibly permutations for coalition arrangements and decide to vote based on who they want and don’t want to be in government.

I think there are two very notable aspects of this from 2002.

National’s vote collapsed to 20.93%, having been 30.50% in 1999, 33.87% in 1996 (the first election under MMP), 35.05% in 1993 and 47.82% in 1990.

Their vote had gradually declined through the 1990s and then plummeted in 2002.

Labour’s share of the vote has been declining in each election this century 20 last years 25%.

And that 39% of voters would consider voting for a party they wouldn’t normally support suggests that a large number of voters will potentially move their votes around.

There’s a lot that can happen in the next 15-17 months, especially in the lead up to next year’s election as voters seriously consider the keep options and decisions.

Proof of poll movement

With the latest One News poll Colmar Brunton revealed evidence of how much opinion – or the polled measure of opinion – can move over a short time.

One aspect that is usually ignored is that voters may think quite differently during an election campaign as they consider the governing possibilities and they decide how they want to vote – strategic voting has become more common – than how they might think in a spur of the moment poll mid-term.

The main poll question asked was “If a general election was held today, would you be eligible to vote?”

As usual One News show how the seats in Parliament would look if an election ‘was held today’.

But the polling in this week’s poll was done over 6 days, from Saturday 28 May to Thursday 2 June. And the poll was reported on Tuesday 7 June, 10 days after the first day of polling and 5 days after the last day day of polling.

How much could opinion change in that short a time? Quite a lot going by poll numbers split pre-MoU announcement and post-MoU announcement provided by Colmar Brunton:


The Memorandum of Understanding was announced by the Greens on Tuesday 31 May at 3.30 pm.

There is no indication of when people who were polled heard about the MoU, how much they heard about the MoU or whether they heard about the MoU at all before being polled.

There was quite a bit of ongoing discussion and news about the MoU after the polling was complete, especially over the following weekend with coverage of the Green AGM where Green leaders and Andrew Little spoke about the MoU.

And Labour and Green leaders, as well as Winston Peters,  were interviewed about the MoU on Saturday on The Nation and on Sunday on Q+A.

So people who were polled in the last two and a half  days of the polling period, as opposed to the first three and a half days days of the polling, would at best have only based their poll decisions on very preliminary consideration of the implications  of the MoU, if at all.

It should also be noted that the MoU was not the only news over the polling period. Other news may have affected people’s opinions other than the MoU. Assuming that the MoU was the sole cause of a shift in opinion is baseless.

So as far as the MoU goes the before and after poll results should be viewed with a lot of caution.

As well as this single polls in general should be viewed with caution. Trends of one pollster over several months and aggregation of multiple pollsters are generally regarded as much better indicators of public opinion.

And another point – the before and after results show how much opinion measured by a poll can change in a short space of time, a matter of a few days.

NZ First support dropped from about 11% to about 7%, by about a third, a big variation.

Greens support increased by about a quarter, despite it being stated this wasn’t statistically significant I think it is notable.

And Labour support moved over 5%, from 26.1% to 31.3%. We don’t know whether support moved up a further 5% in the next 3 days, or dropped back again, or if the poll was an outlier poll.

All we know from this with any certainty is that polled opinion can change significantly over a short period of time.

Therefore the precise seating arrangements displayed by One News and others, and the ‘analysis’ of what a poll result might mean and why it might mean whatever they claim should be viewed with a lot of scepticism.

Reporting on polls by the mainstream media is usually awful and ignores the realities of political polling.

Single polls are no more than a rough indicator of opinion averaged over a few days.

One last point – a sample size of 1500 is unusual, 900-1000 is far more common.

As far as I understand it most polling results are usually obtained in the initial days of a polling period with the rest of the period used to fill the gaps in their demographic quotas.

So was a mid-poll decision made to increase the number of people being polled by 50%? Polling 628 people in two days seems unusual to me and may make polling variances more likely.

This latest Colmar Brunton poll demonstrates about how much opinion, or the measurement of opinion, can change over time, even over a very short time.

One News Poll – June 2016

One News Colmar Brunton poll for June 2016:

  • National 48% (down from 50)
  • Labour 29% (up from 28)
  • Greens 12% (up from 10)
  • NZ First 9%
  • Maori Party 0.7% (down from 1.1)
  • Conservative Party 0.7% (up from 0.3)
  • ACT Party 0.3% (down from 0.7)
  • Other 0.6% (up from 0,2)

Base(n=) 1,245

Despite the commentary on One News I think it’s far too soon to read much into this result in relation to the Labour-Green Memorandum of Understanding.

Polling was done between 28th of May and 2nd of June, after the budget and with the MOU announcement part way through.

Regardless of that, Labour+Green at 41% is still a long way short of National’s 48%.


For preferred Prime Minister:

  • John Key 39%
  • Winston Peters 12% (up from 10)
  • Andrew Little 7%

Before and after MoU:


That might surprise and worry some people but I still think it’s too soon to judge much from this poll in relation to the MoU announcement. I’d say that Greens will be a tad anxious.

Full report (PDF)

Double digit lead for Clinton

The latest poll (Reuters/Ipsos) gives Hillary Clinton an 11% lead over Donald Trump.

  • Clinton 46%
  • Trump 35%
  • Neither 19%

Clinton opens up double-digit lead over Trump nationwide: Reuters/Ipsos poll

Democratic presidential contender Hillary Clinton has opened up a double-digit lead over Republican rival Donald Trump, regaining ground after the New York billionaire briefly tied her last month, according to a Reuters/Ipsos poll released on Friday.

Some 46 percent of likely voters said they supported Clinton, while 35 percent said they supported Trump, and another 19 percent said they would not support either, according to the survey of 1,421 people conducted between May 30 and June 3.

Trump had briefly tied Clinton in support among likely U.S. voters in May, raising expectations for a tight race between the two likely contenders in November’s presidential election.

Only 5 months and about 500 polls until the election.


Get every new post delivered to your Inbox.

Join 4,514 other followers