Tuesday, October 05, 2004

Another poll

Stuff has a poll today . If an election was held today who would you vote for? As at the time of writing, 7220 had voted - which is one of the biggest Stuff polls - and here's the results so far:

National 41.9%
Labour 32.6 %
Greens 7.1 %
NZ First 5.2 %
Act 5%
United Future 3.9 %
Maori 3.8 %
Pregressives 1.5%

Interesting that there's no undecided category. Now all I have to do is go out and interview 60 of `em to see why people who read Stuff want a National coalition Government.

Gee, even ACT get into Parliament - and perhaps even Government - if this poll translates to seats in the House. The ACT / National and Act/ National /UFNZ votes are bigger than the Labour/Progressive/Green/ Maori vote.


jarrod said...

This sort of poll is genuinely dodgy, despite the number of respondents. To start with, it's self-selecting - there's no attempt to adjust for demographics, and anyone can vote, which means that the results can be easily manipulated. All that can be taken from a poll like this one is that this may be a representation of the opinions of those visitors to the site who could be bothered clicking the link.

Having said that, to a certain extent any poll should be taken with a grain of salt - you're dealing with a potentially skewed sample regardless. In the case of a telephone poll, the sample population "people who can be bothered providing answer to a telephone surveyor" may differ quantifiably from the general population. But how would you measure that quantifiable difference? Therein lies the problem. I've yet to hear of a surveying methodology that adjusts for individual psychology...

Anonymous said...


perhaps you can tell us how a randomly selected bunch of phone calls (to only 750 people and PAID for by a pro-PC media outlet) is any different to a bunch of random clicks (of over 7000 people) on a computer program that automatically collates info.

The only psychology needing testing is.....well.

Anonymous said...

Sorry, I tend to agree with criticisms of Stuff's straw poll. It is vulnerable to income factors and may end up inflating some demographics and associated partisan voting patterns, relative to others. I prefer NBR and Colmar Brunton, who are professional pollsters.


jarrod said...


I think I'm in love. What are you wearing?

Why don't you take the time to learn something about surveying techniques and methodology? For starters, you could try looking at this page on DigiPoll's own website, which gives some background.

The basic problem with this poll is the same as the basic problem with the submissions to the CUB select committee - respondents are self selecting. Therefore they are not a genuine random sample, and we have no way of knowing if they are representative.

You like science, don't you? Did they teach you the importance of robust study design? Perhaps not.

Anonymous said...

Hmmmmm.....self-selection, now there's a novel concept.

Well, then we'll have to stop that won't we? People exercising free will is just not good enough, we must control what they can and can't respond to. Bring out zee dogs.......

If people can't be bothered to write to Parliament, or if they can't be bothered to express a view on TV or wherever, then it's highly unlikely they have actually THOUGHT about the issue. Not because they are stupid, but because they are lazy. And that laziness is what the PC brigade count on, and it's what people like Craig exploit when they tell lies and make up weird and wonderful objections to poll's and committee submissions. I would attach more weight to any of his or your arguments if you just said, "I don't like what I'm seeing because it doesn't fit with my world view". At least that would be honest.

jarrod said...


I'm sorry if this argument is going over your head. I bet it's a cute head.

Thank you for your charitable assumptions regarding my motives for disagreeing with the methodology of the Stuff poll. It's just like the people at the NZ Herald paying a polling firm to come up with the results they wanted. You'd think they'd cut out the middleman and just make up poll numbers, wouldn't you? It's so much cheaper that way. Harder to conceal the liberal media conspiracy though.

Just to be clear: you show me a poll from any reputable polling firm that shows a similar result to the Stuff poll, or to the Newman and Mapp polls you mentioned in relation to the CUB, and I will accept that the poll might be a fair representation of New Zealand's opinions. I'll be waiting.


PS You don't need to be "informed" to write a submission to a select committee. About the only requirements are an opinion and an envelope - as I understand some of the submissions make abundantly clear.

Rich said...

I'm assuming that this poll is meant to be a prediction of a future election result, rather than just an attempt to find out the answers among people who read Stuff and can be bothered to vote.

Biases in this poll, it's a self-selected sample of people who use the net, read Stuff, read it today, can be bothered to vote.

An election is also a self-selected sample, but a different one - everyone (in theory) gets reminded, and a lot more people feel they should go & vote.

Among the biases in the poll are that people who are dissatisfied with the present government, or are partisans of one of the smaller parties, are more likely to vote in an internet "poll" than the population as a whole.

Greyshade said...

What's the difference between 7000 random voters clicking on Stuff's button and 700 random telephones being clicked by a polling company. Well for starters the 700 telephones really can be random. The subset of people who respond to the poll when rung is not-random but it's a high and (more importantly) known proportion. We have no idea what proportion of the people who visit Stuff's site (or who heard about the poll some other way) the 7000 represents. Nor do we know how many of them are New Zealand voters or how many of them voted twice or more. Or even if they were human - maybe 1000 of the votes were generated by a single bot-program written to automatically visit the site and vote for a right wing party.

jarrod said...

MSNBC have a nice summary of the reasons why the internet polls they use are not scientifically valid here. The same logic applies to Stuff's poll.

span said...

well i couldn't vote at all, although i did happen to see the link, because my party of choice wasn't on the list. the best way to ask the question is probably not to give people a list of parties, unless they request it, but to just say who would you vote for if an election were held tomorrow and let them say which ever party they want.

i am quite astounded Mr Tips at your failure to understand the really basic premise of polling, ie that it needs to be a random sample to be able to extrapolate with any reliability to the general population. jarrod and craig have tried to explain it much better than i can, but they don't seem to be doing much good either. i mean i did Bursary Stats (showing my age) and they really hammered the "self-selecting is not good" point in, over and over and over again.

it must make reading actual polls v frustrating for you...

anyway it's good that we're having a debate anyway :-)