Written By:
Marty G - Date published:
11:57 am, October 21st, 2010 - 41 comments
Categories: act, greens, labour, national, polls -
Tags: why the right is worried
The poll numbers – Labour + Green equals 44.5% vs National + ACT at 50% according to the latest Roy Morgan. A year ago that was LG = 34.5% and NACT= 59%.
Those numbers suggest that 240,000 voters have shifted from Right to Left in the past year. The trend is quite clear now. The Right’s support peaked last year and the gap has been gradually closing ever since. A further 66,000, 2.5%, and the two sides are equal. As the economy falters and communities around the country see the mandates they’ve just given the new Left mayors around the country frustrated by National, expect to see those numbers shift Left more.
On current trends, Left could pass Right as soon as February (really early election anyone?):
I don’t think the gap will close that quickly but the pattern is undeniable: in time, that gap will close and then reverse.
Despite this, media reports continue to be laced with references to Labour’s inability to win and being ‘stuck in the doldrums’. That’s just dinosaur thinking from some journos who still think that the biggest party governs, not the biggest coalition.
It’s vital for the Right that Labour be written-off in the minds of the public. The risk for the Right is that you begin to believe your own spin and they’ll be caught with their pants down the day that the Left polls higher than the Right.
The server will be getting hardware changes this evening starting at 10pm NZDT.
The site will be off line for some hours.
Watch it go the other way as the disaster that Labour, EPMU and CTU have dealt the economy with the sabotage of The Hobbit sinks in.
All that good work by Goff at the conference undone.
Love the scatter-gun approach from grumps. Small union out of its depth has a fiasco, somehow it must be the EPMU and the Labour Party’s fault. And how dare the CTU attempt to salvage the situation by reopening dialogue?
An interesting insight into the feverish minds of the right .
You have to question why Warner and the Government have made this public instead of just negotiating.
Labour and the CTU needs to have its ducks in a row on this one. This whole project has gotten acromonious for a lot of different reasons, and a few hundred thousand dollars worth of minimum terms and conditions for workers will *not* be it.
Watch the NAT spin machine start the smear on unions and Labour though.
Will have to fight back fast.
This morning I heard the theory put forward that it’s Warners tactic to pressure the Govt for tax breaks… Makes sense to me!
Got to be careful not to pay to much attention to the polls. As always, Labour and Greens are going to have to win by knocking on every door and talking to every person, in person. Grassroots is where its at.
But yes, this is heartening.
Agreed. Only one poll matters, it comes up every three years.
the trend in Farrar’s monthly poll of polls has LG passing NACT within a year.
early election, mark my words.
April/May? Weather will still be good and turnout high. NATs won’t like that either.
What you ignore is the impact of a resurgent NZ First. Winston will contest. And he will hover upwards of 5%, no need for Tauranga or whatever. NZF could bleed support from either disaffected NACTS or equally angry purple voters who like neither the Goff Father nor Don Key. That makes Winston kingmaker, and I doubt any poll can measure which of the coalition arrangements he would (re)sell his soul to. I’d pick he’d take the popular “nice guy”, for they would likely offer him the nicer prizes.
Or, Winston could pull a repeat of the last election and get 4.5% and no electorate seat, sucking out a good proportion of the vote with him.
Last time, this was bad for Labour because it allowed Act and National to get more seats than they would otherwise have. If it happens in 2011, it could hurt National more if most of NZFs vote came from the right rather than the left.
Given the NAct prolonged attack on NZF in 2008, what does it say about their 4.5% vote in 2008 Election?
If Winstone becomes the kingmaker he’ll go with Labour. He’s got a score to settle if Hide does not get in he’ll go after Key for cutting him adrift by stating that he (Key) will not work with him (Winstone)
Winstones been in politics too long to realize he is going to be given a ‘second chance’ by Labour. He will get some bill passed for his support but in the short term he’ll support Goff just to take potshots at Key. Winston’s in it for the fun.
I wouldn’t put it past Winnie to go with Key just to make his life hell.
It’s also not too hard to imagine an election result which would allow Winston to punish Hide by going with Key.
Winston is definitely not the same guy as he was in 2002 and 2005. Will be interesting to see where his head is at in 2011.
As the finance guys say – past performance is no indicator of future performance.
the right is worried because they dont have a programme and they haven’t been able to organisea bogeyman.
and they they had their turn and couldnt do anything with it.
They’ve been given an opportunity to be the Govt and they’ve wasted it.
Wow this post is epic fail, if I did an analysis like that on a laboratory report I would surely fail.
1. Are all the polls equally valid? I.E. where they all well conducted and what were the sample sizes?
2. No statistical tests!
Inevitably labour will gain power once again, however your \”analysis\” adds nothing to the debate of when this will occur, and offers no seroius insights into the current political situation. There is no reason to believe that a polynomial line of best fit is a suitable methodology for forecasting political polls.
Wow, I mean, wow.
To answer you’re first criticism, they’re all Roy Morgan polls. So presumably, Roy Morgan being a responsible polling company, would try to make sure that all of their individual polls were “well conducted” and would generally strive for the same sample sizes each time so as to allow their polls to be comparable. Given that Roy Morgan themselves like to compare their own poll results over time, I don’t think Marty is being out of line by also comparing their poll results in much the same way.
As for the rest, if you read “I don’t think the gap will close that quickly but the pattern is undeniable” you will see that Marty doesn’t believe the projection here will hold out, just that it representative of the publics apparent shift in mood towards the current government.
If you did a straight-line fit of the polling since September 2009 you would end up with a fairly similar projection, to my eye.
Even well conducted polls have a statistical margin of error, that was more what I meant to ask. I’d like to see the graph remade with a error bars worked in, and then see someone try to make the case that polynomial extrapolation is appropriate.
Regardless of what MartyG said, he still put that ridiculous extrapolated trend line onto the figure without any justification. Its ridiculous to suggest that it is reasonable to use such an analysis in politics, the mood of the public is not something that changes at a constant rate over time, instead it is influenced by many complicated factors.
Thus this entire post is meaningless, one only has to compare the electability of Phill Goff versus John Key to come to the conclusion that the coming general election is National’s to lose.
Agreed JJ, poor statistics indeed.
they’re all roy morgan polls dickhead – go to roy morgan to find out more about them. they fact that you’re casting doubt on the polls without chcking them suggets you just don’t want to believe the evidence before you. poly or straight line trend since last year, you’re still getting a closing gap.
since you’ve clearly got the right rising and then falling and the left doing the reverse, I would have thought that he’s right to use the poly trend option.
the posts throughout the site offer insight into the politics, and this one does too. it says ‘the polls aren’t so golden for National as idiots might think’
No, we do not clearly have the right falling and the left rising – what we have is a meaningless graph without any mention of the statistical significance of such changes. The clearest thing I see is a sustainted gap between the blue and red lines.
JJ
Click on link in the post.
Read all the numbers. The fortnightly polling periods since the election, the party support levels, the “did not answer”, the sample sizes, etc. It’s all there.
If you’re still not satisfied, contact Roy Morgan directly and inform them of their “epic fail” (as you put it). I’m sure they’d love to hear from you.
I am not saying that the poll numbers themselves are epic fail, I am saying that the OP is epic fail. Extrapolation of poll data to future months is completely inappropriate.
All polls have margins of error, that is a fact of life. Look at the volatility from month to month, with such volatility it is foolish to extrapolate data.
“On current trends, Left could pass Right as soon as February”
What trend? Is the “trend” even statistically significant? Even cursory critical thinking produces the conclusion that this is a stupid statement. The polls have a margin of error of approximately +/- 4% – for all we know the supposed “trend” in these polls is an artifact of random error.
Once again, its not Roy Morgan that I am questioning – its MartyGs (lack of) analysis, and don’t misquote me it was clearly MartyG that I was referring to as epic fail, not Roy Morgan.
JJ, from your first comment:
1. Are all the polls equally valid? I.E. where they all well conducted and what were the sample sizes?
2. No statistical tests!
You asked questions, which had – at least partly – already been answered. In the link that Marty provided, and you ignored.
Hence, fail.
Ok, surprise surprise the comment I put together in the few minutes I had before the bus arrived was not perfect.
I have however made valid points abut Marty’s point, so no not fail.
JJ is absolutely right. A parabola ALWAYS bends, so entices the reader into making assumptions that just aren’t there in the data — in fact, I remember being taught that even-degree polynomial fits are never good ideas. Go for an odd-degree fit if you really must. And anyway, the two lines above are clearly better with a linear fit.
Sloppy work from a guy whose analysis I normally really respect.
On the other hand, the trendline would be perfect if we renamed it “Line of Best Wish”.
The Unreality is strong in this one…
I’ll only be truly happy when Labour’s consistently above 37% and the Nats are consistently below 47%.
Regarding NZ First, I suspect most of their supporters now favour Labour as a Coalition partner. I’d say most NZ First supporters who swung away from the Party in 2008 were those upset that Peters went with Labour in 2005. Certainly, NZ First voters in 2008 were far more likely to cast their Candidate-Vote for Lab than the Nats.
I did the same analysis using polls from 1999-2002. Remember Labour’s winter of discontent? Well applying Marty’s very in depth analysis, the graph for 1999-2002 showed National reached its nadir in 2001, and started recovering Labour and the Alliance. This analysis shows National won the election in 2002 in a landslide.
The same polls showed Labour at 53%, three months before the election.
They got 41% on election day.
So, smhead’s “analysis”, and mine, proves … the obvious. Voters are volatile, things change. All to play for, then.
The same polls show support from Labour shifted to parties that would support Labour (ie Greens and United). National’s support was at 30% a year before the election and lost to its worst defeat ever because Labour had a much more popular and competent leader remember and National hadn’t changed since the 1999 election and was hoping the public would suddenly just wake up and realise they’d made a mistake by voting National out in 1999.
Haha, that’s why Labour’s screwed. That and the unions tried to kill the hobbit to make the government look bad.
The polls are interesting, but not really that clear on what’s happening now and where things are going. There’s still a lot can change in a few months.
Marty – R2 for each of those lines of best fit?
Seconded.
3rded
Some actual statistical analysis would we greatly appreciated, not just a pretty graph.
Yeah regraphing showing the errors above and below would be good plus a bit on consideration of the limits of the polling — is it based on ringing people on landlines? does that introduce a bias to the results if people who only have mobile phones or no phone aren’t polled? But I think the approach of only graphing one polling company’s results is much better than the Farrar approach of assuming that polls from three different companies can be treated as giving equally valid results.