Hard News: Poll Crunch
173 Responses
First ←Older Page 1 … 3 4 5 6 7 Newer→ Last
-
Gio, I would indeed say that. I'd also mention the large error ratio. But in absence of information, and wanting a prediction, what other result would you pick?
But isn't the insistence on prediction sort of self-defeating? At that sort of level of undecideds (even at the current NZ levels mentioned in this thread) you're speculating without data, and that's not prediction, it's guess-work. At worst, it's highly misleading guesswork. I know polling tends to be more of an art than a science, but come *on*.
-
I have this very vivid childhood memory in which I overhear my father telling a phone pollster that he is going to vote for X, but I very well know he supports Y. I ask him why he lied to the man on the phone and he tells me "because it makes election night a lot more exciting". I think it was 1984.
-
Lucy, do we actually know that the assumption of proportionality is bad? It may be quite accurate. The accuracy may already be included in the margin of error calculation. I don't know.
-
Margin of error is a statistical concept that describes the statistical uncertainty associated with generalising the outcome of poll to a population. As Wikipedia notes "it does not represent other potential sources of error or bias such as a non-representative sample-design, poorly phrased questions, people lying or refusing to respond, the exclusion of people who could not be contacted, or miscounts and miscalculations." So problems caused by undecideds aren't effectively 'controlled for' in the reported statistics.
The other thing about margins of error is that they relate to generalising the results of the poll to the population (i.e., having an educated guess about how the whole of NZ would answer the poll) not generalising the poll to how the population will vote in an election. That requires us to also know about the statistical relationship that holds between poll results and voting.
-
potential sources of error or bias such as a non-representative sample-design, poorly phrased questions, people lying or refusing to respond, the exclusion of people who could not be contacted, or miscounts and miscalculations."
Another consideration, which is beginning to be acknowledged by some pollsters, is the dependence on land-line telephone polling--ignoring the significant shift to mobile phones. This further undermines the statistical validity of any findings.
Too many pollsters also ignore the 'don't know's' or refusals-to-respond. They tend to be treated as know-nothings, when they are perfectly valid responses to intrusive research methods.
Peter Dunne--I just wish he was done and dusted. He really is a political whore.
-
gio I'm not sure if we're talking at cross purposes.
I think Giovanni has it right.
To me, by divvying up the undecideds and apportioning them out appropriately, they're throwing away a useful piece of information.
If candidate A is on 43%, and candidate B is on 40%, candidate B is close. But if the number of undecideds is only 3%, then they need them all to fall for them to draw even - impossible. Or have candidate A outed as a paedophile, so that people who said they were going to vote for them change their mind.
Yet if the undecideds are 12%, they need to get 3/4 of them - difficult, but not impossible. Keep the paedophilia card in your pocket!
In NZ we throw this information away. To me, it'd be interesting to know, at this point, how many people replied "I don't know". If that's 12, 15%, wow, good moves by Labour could make this thing really close. If it's 5%... not so much. Then their only hope is the fact that it's a poll, and not an actual election, and coalition moves that they make afterwards.
-
The Herald arguments on Saturday: "Feeble minded or disingenuous".
I'm voting for disingenuous by the feeble-minded.
The Herlad clearly want to condition people for a return to FPP in the referendum they clearly expect from the national victory they yearn for.
As newspaper go.....they have moved over into propaganda.
-
Gio, I would indeed say that. I'd also mention the large error ratio. But in absence of information, and wanting a prediction, what other result would you pick?
Well, you'd be wrong then on election day, wouldn't you? And it's quality of information we're talking about. If a poll has a 20% proportion of undecideds, and doesn't acknowledge the fact but simply factors them out of the calculation, that poll is unreliable far beyond its MoE.
And, as Stephen notes, undecideds have nothing whatsoever to do with the margin of error. The margin of error is what accounts for the fact that you didn't ask every single New Zealander, but a sample of New Zealanders. Undecided voters are people who might vote but haven't made up their minds yet - knowing how many of them are there and what kind of profile they might fit is very interesting information, not only for people in the business of trying to persuade them to vote for them, but also for the public.
David Sedaris had a nice bit recently about undecideds in the current American election:
To put them in perspective, I think of being on an airplane. The flight attendant comes down the aisle with her food cart and, eventually, parks it beside my seat. “Can I interest you in the chicken?” she asks. “Or would you prefer the platter of shit with bits of broken glass in it?”
To be undecided in this election is to pause for a moment and then ask how the chicken is cooked.
-
procatalepsis
Sounds like something I suffered from after my third pregnancy.
-
And where would we be without Chris Trotter jumping the shark:
If media outlets sympathetic to National’s cause attempt to cast as illegitimate the formation of any coalition government which leaves National sitting on the Opposition Benches, they will be guilty of engaging in activity subversive not only of the New Zealand constitution, the rule of law, and our democratic Westminster heritage, but also of creating a political climate likely to result in the outbreak of serious - perhaps fatal - public disorder.
In short, they will be guilty of treason.
Of course, Mr. Trotter's respect for the New Zealand constitution etc, doesn't extend to that tiresome convention you actually have to hold a fair trial before convicting anyone of treason.
-
Commisar Trotter has a nice Animal Farm ring to it.
-
Heh... Comrade Trotterski wouldn't be bad, but it sounds too much like something Sarah Palin would say. But seriously, Trotter accuses Guyon Espiner of "he careful crafting of a political “reality” which will enable National to cry foul if, after the election, it finds itself unable to attract sufficient support on the floor of the House of Representatives to form a government." All on the basis of what I say was a weak story on Monday based on a rather poor poll question designed to fluff the debate. But treasonous whipping of the booboise into rapine and pillage in the streets?
Fuck off. And the worse thing is, I don't believe Trotter believes a word of it either. But he's got his own political "reality" to construct, I guess, where if Labour doesn't win its because the media elites and Beltway insiders (shit, there's bloody Palin again!) stole it.
-
Of course, Mr. Trotter's respect for the New Zealand constitution etc, doesn't extend to that tiresome convention you actually have to hold a fair trial before convicting anyone of treason.
Or indeed, that parties should obey electoral law.
"Courageous corruption", anyone?
Reading Trotter's rant, I'm very glad we repealed sedition.
-
Reading Trotter's rant, I'm very glad we repealed sedition.
Sedition's for girlymen. It's treason or nothing.
-
Sedition's for girlymen. It's treason or nothing.
Pshaw... you're a fricking pussy, Brown. Burn 'em all at the stake and let Joseph Michael Savage sort them out. Forget all this limp-wristed namby-pamby nonsense about charges and so on.
-
Woo! New Pew!
McCain Support Continues Downward Spiral: Obama Leads by 19 Among Those Who Have Already Voted
And on the 10% who are undecided voters:
Undecided voters are less educated, less affluent, and somewhat more likely to be female than the average voter. Nearly half of undecided voters (48%) say they attend religious services at least weekly, which is same as the proportion of McCain supporters. Fewer Obama supporters (31%) say they attend religious services at least once a week.
-
Forget all this limp-wristed namby-pamby nonsense about charges and so on.
Send the judges out to work in the fields!
-
see - that's usefull information about undecided voters (both that they tend to be in the McCain demographic - and that they are undecided about McCain) - of course to be meaningfull you need enough undecideds to be sampled for stats about them to be usefull (easy if they are 20%, hard if they are 1%)
-
Send the judges out to work in the fields!
Oooh, I've seen this movie! Can I be the hot but sadistic warden who turns a tropical paradise into a hell of perverted appetites? Watch out Sian Elias!
-
Gio
Well, you'd be wrong then on election day, wouldn't you?
Who knows? I still think we're arguing at cross purposes. I'm not advocating getting rid of the information about undecideds - that's crucial data about the quality. But the extrapolation is not something to throw away either, and it seems to me you are advocating that.
Since you made up a totally fictitious example lets take it a little further. Suppose in your example you would live or die by your choice and you were forced to make a choice. Would you go with the 1 person or the 2 people in your tiny poll? In absence of any other information?
I see it like a football game. National is currently ahead and we've played 80% of the game. What's the likely outcome? Of course it is still uncertain. Very uncertain. But the odds are on the team that is ahead.
The technical term margin of error is a pretty simple calculation as I remember from school, but I really don't know what kind of statistical sophistication is added by these pollsters.
-
Who knows?<quote>
Well, I suppose, but a poll shouldn't be a shot in the complete dark, no? Unless they come out and say it, which they won't.
<quote>I still think we're arguing at cross purposes. I'm not advocating getting rid of the information about undecideds - that's crucial data about the quality. But the extrapolation is not something to throw away either, and it seems to me you are advocating that.
You bet I am. And I really don't see the problem in doing it, since it's done in other countries and the prospective voters there are able to compute the fact that the opinion poll results don't add up to 100 per cent without suffering from brain explosions. Undecideds are an important part of the snapshot of the electorate you're trying to provide, assimilating them to the other parties is lazy and amplifies the MoE. So why do it? I get that they want to tell us how many MPs party X is going to get, and who's best placed to form the next government, but they could still do it while leaving the baseline percentages as they are for all to see. Thusly
Today we found Labour om 42, National on 39, the Greens on 9, small parties bla bla bla and 6 per cent undecided voters. If the undecideds were to split along with the rest of the electorate, the seats in government would be...
-
but the concept of three tax tiers, income splitting etc are ones I support.
How would that be simpler than the current system? You'd have to file a tax return to split a couples income (and who would be allowed to do that? Would one be able to share income with flatmates?).
I think our system is as simple as can be achieved with fairness, given that it doesn't require tax returns from ordinary employees. (A flat tax would be simpler, as would arbitrary confiscation of assets from the rich - but probably less fair).
-
I'm sure "arbitrary confiscation of assets from the rich" would breed all sorts of tax shelters.
Obviously one would need to file if you wanted to file jointly - and the simplest solution ends up with a cheque at the end of the year (rather than your employer doing something subtle - and difficult - to make it come out right at the end of the year).
I've filed jointly in the US in the past - you just drop 2 SSNs at the top of the form - no one ever asked to see my marriage cert - though if my spouse was also male I'm sure questions would have been asked.
What's really hard for us at the moment is filing jointly in the US (for stuff we've left there) and individually in NZ and then trying to figure out how to handle the tax treaty stuff
-
Cheers for clearing that up Gio. How about we compromise, and YOU throw away the extrapolation for your own purposes, whatever they are. Everyone else who wants it can keep it. Personally, I want it. I also want the other data. I'm not gonna turn information-fascist and demand that various analysis be thrown away because I have no use for it.
-
The Herald ever so slightly redeems itself by broadcasting the doubting voice o Rudman - "When silver and bronze beat gold":
The sniping about MMP is a smokescreen for the critics' inability to adjust to the new rules. They're talking horse racing when the game is now chess.
Post your response…
This topic is closed.