Who the Hell knows?
Two polls, two results. Both our esteemed Sunday papers tried to conduct polls in the last few days, to see what effect in any Brash’s dishonesty had had on public opinion. The Sunday Star-Times, polling only on Friday night, got a lead for National of 7%. The Herald on Sunday, polling on Friday and Saturday, got a lead for Labour of almost 4%. This is a difference of 11%, and would suggest that one or other (or both) has some rather major methodological problems.
As for the Sunday Star-Times poll: are the type of people at home on a Friday night an accurate representation of the New Zealand public? Presumably most of the young of the land and huge swathes of the general urban population would have been out drinking? Isn’t the point of polling over many days and at different times of the day so that everyone has a chance of being picked? As for the Herald on Sunday poll: is 400 people really an adequate sample size for a meaningful national poll? Couldn’t you have shelled out to pump it up towards 750?
So, I’m afraid we’re now left with our standard “who the Hell knows?” poll interpretation. Only a fool would try to read meaningful trends from these junk polls. There’s also a TVNZ poll out tonight, which I think will have been substantially taken before the Brash dishonesty. Who knows, perhaps it will add to the polling confusion by bringing in a result that fits in neatly between the two Sunday paper polls.
I think I’ll wait respectfully for the polls coming out in the last few days before polling day (TV3 on Wednesday, the Herald and possibly NBR and/or TVNZ on Friday) before even thinking about trying to call this race.
Suitably, adding these new figures into frogblog’s polling averager has Labour and National neck-and-neck, both with 50 seats. Though Labour has half a percent more party votes than National on our average figures, I haven’t the first idea what United and/or NZ First would do with this kind of election result. Half a percent isn’t really a meaningful difference, so perhaps all bets would be off.
As a result, I have for the first time abstained in predicting a “most likely government”. The Labour/Green/Progressive team has 59 seats, while the National/United team has 52 seats, which would be boosted to 59 if NZ First backed them. This would leave the Maori Party’s three seats as the casting vote. Who knows what would happen in this scenario? Would Helen Clark bite the bullet and call that last cab off the rank? Would Winston Peters say to the Prime Minister: look, you can have your Labour/Green Government and I’ll abstain from voting against it, but you’re going to have to be awfully nice to me with policy concessions? It’s just too hard to tell, so for the first time my “most likely government” is listed as “too close to call”.
NZ First’s numbers are still heading south, and a Herald on Sunday poll today has Winston 11 points behind in Tauranga. If there’s any Kiwi politician who could reverse such a deficit in six days, it would be Winston, but I wouldn’t count on it. NZ First looks to be headed back to Parliament with its leader’s tail between his legs and a radically reduced number of MPs.
The Greens’ position in today’s polls - of at or around 6% - is heartening. Still, we seem the only third party to have remained largely immune from the third party disease rampant in Election 2005.
The new data and charts are here
UPDATE: Well, the Sunday Star-Times poll showing National 7% ahead is even dodgier than I thought. Go have a read of political editor Helen Bain’s column on A10 (not online, sorry), and you’ll see that she refers to National being two points ahead of Labour in the latest Sunday Star-Times poll, not the seven reported. My only guess at explaning this discrepancy is that there are, in fact, two polls: one that was reported and one that wasn’t. The Sunday Star-Times did its usual poll, but sampling was completed before the Brash gaffe/dishonesty. It showed National ahead by 2%. So, the paper decided to do an additional poll on Friday night to see what damage National had sustained as a result of the Brash gaffe/dishonesty. Problem is: the new poll showed National extending its lead to 7%. This left the paper two options. It could report on both polls, and open itself up to ridicule (how on earth could National get a five point bump after its leader was exposed as misleading the country?). Or it could pretend like the first poll didn’t exist. It opted for the second option, except that it forget to edit Ms Bain’s column. If this suspicion is accurate, then today’s poll with National 7% ahead shouldn’t be taken with a grain of salt so much as with a truckload of salt. Also, it has to be said, if the Sunday Star-Times has just done what I suspect it has, then it has just indulged in a piece of pretty shady reporting one week out from a general election.
UPDATE 2: To be clear, my best stab at an explanation of the difference between the SST and HoS polls (National by 7% and Labour by 4% respectively) is that a Friday night poll is skewed in favour of the old, the wealthy and the rural. Indeed, Friday night is probably the worst possible time to try and conduct a poll - because next to no-one is home.
UPDATE 3: A frogblog reader has concurred with my suspicion over the Sunday Star-Times poll:
Yes, you are right. I was phone polled early last week (Monday I think) for the Sunday Star Times and was told that the results would be in the paper yesterday.
September 11th, 2005 at 11:29 am
Frog:
For once, we’re in perfect agreement apart from the bit where you said “The Greens’ position in today’s polls - of at or around 6% - is heartening”.
BTW, I don’t think it was a matter of “shelling out” when it came to the sample size. I was surprised enough to see the HoS’s pollster managed to get anywhere near 400 usable responses in 24 hours.
September 11th, 2005 at 11:57 am
this is totally unacceptable!
Anything claiming to be a politacal poll needs to be open about its process and have a process with sound methodology as judged by statistics experts.
These polls, as they get closer to the election may have a huge baring on how people actually vote.
Another reason why I have now become convinced that the introduction of stv for both party votes and electorate votes is essential to retaining our democracy.
- imagine a poll just before the elections that showed the greens under 5%.
There could be a rush to Labour to try and stave off the thought of a nat victory.
- This is just one example to do with the greens, but if this happened to any party it would be a tradgedy for our democracy.
Only by introducing stv at both party and electorate levels, can we correct this problem.
September 11th, 2005 at 11:57 am
this is totally unacceptable!
Anything claiming to be a politacal poll needs to be open about its process and have a process with sound methodology as judged by statistics experts.
These polls, as they get closer to the election may have a huge baring on how people actually vote.
Another reason why I have now become convinced that the introduction of stv for both party votes and electorate votes is essential to retaining our democracy.
- imagine a poll just before the elections that showed the greens under 5%.
There could be a rush to Labour to try and stave off the thought of a nat victory.
- This is just one example to do with the greens, but if this happened to any party it would be a tradgedy for our democracy.
Only by introducing stv at both party and electorate levels, can we correct this problem.
imho.
September 11th, 2005 at 11:59 am
oops I tried to stop the “add comment” from loading so that I could add imho.
could you please delete my first comment and this one
cheers.
September 11th, 2005 at 12:00 pm
frog..a tidy summary of the perils of the polls…but can i query your claims of ‘predicting a most likely government’..?
you aren’t engaging in the dark and perilous art of prediction….you are in the number-crunching business of post-mortem….of necessity ..your graphs look back..not forward..therefor you cannot claim to be predicting..eh..?..you are more in the arena of ..”if an election had been called last week..”…note use of the words ‘last week’..
(a small point..but a significant one..don’t want to claim any unearned glory..do we frog..?.heh heh..)
you have been consistant in refusing to ‘call’ the election..(the safer course..i know..)…unlike (ahem) foolhardy me..
and that’s lab/grn/prog…greens crack 10%..national vote collapses..act get back in ..peters is too hard to call..but he is certainly going through a ‘perils’ period..
yours in accurate use of the english language….
phil(whoar.co.nz)
September 11th, 2005 at 1:42 pm
I was polled for the SST Poll over a week ago - I would have expected it to be published last Sunday but there was no poll. I think it’s fair to say that any poll published today reflects public attitudes of a week previous.
Who knows what the Friday night influence is - it may have avoided contact with young single workers who intend to vote National because of the tax cut proposition, or young people intending to vote Labour becuase of the Loans policy?
I guess your statement of ‘Who the Hell knows!’ is the most accurate. Second on the accuracy scale is it is too close to call on a number of issues.
September 11th, 2005 at 1:42 pm
Come on Frog this is conspiracy theory-type stuff. And whilst banding around conspriacies can be fun in the analysis in the conspiracy vs stuff-up possibility, it is invaribaly the latter.
My guess is that the 2% was a typo not picked up in proof-reading rather than a accurate reflection of some earlier poll not picked up in proof-reading.
As for your query as to why this phantom second poll would show an increase after the exclusive brethren happenings, rather than assume the second one wrong, why not assume the first one under-reported National’s lead, which *fell* to 7% from it’s higher point. Or assume that most people didn’t give a stuff, or are tired on the emphasis on negativity, or like the fact that when Brash stuffed up he actually had the good sense to *apologise*.
September 11th, 2005 at 2:15 pm
On the Friday night issue, I agree you will not normally get a representative sample. However one can use direct pollsters to target certain demographics so the overall responses are still reflective of the population.
To know whether they did this, you would have to ask the SST what their gender and age breakdown was.