Posts by Andrew Robertson
Last ←Newer Page 1 2 3 4 5 Older→ First
-
Hard News: Decidedly Undecided, in reply to steve black,
Hey Steve
Yip, you've got the correct numbers, but they happen to be two of the lowest bases we've seen since I can recall.
Yes 813 is those people who expressed a party voting preference (sometimes with a second prompt) and were likely to vote. And yes there were other people who were asked and answered, but are excluded from the table because they weren't likely enough to vote.
All the numbers are unweighted, and all the percentages are weighted. It's a MR convention to report unweighted numbers, because weighted numbers can mislead when it comes to determining the reliability of the results.
-
As an aside, my perspective on polling is aligned with the 'Total Survey Error Approach.'
Below is a link to a great textbook on this approach. It is very readable for the non-statistics-minded person.
-
Hard News: Decidedly Undecided, in reply to steve black,
Actually we've only seen it dip below 800 a couple of times. Those unlikely to vote are still asked the party vote question, but excluded when calculating the party support results.
Yes, I expect these methodological differences may contribute to some of the differences we see between polls. And sampling approach, and sampling success rates, and interviewing standards, and the demographic make-up of interviewers, and intereview durations, and weighting schemes, and shift days/times, and sample size, and area stratification, and success in identifying new telephone number banks....
Any pollster, ever, who tells you their sample is perfectly representative either doesn’t know what they’re doing, or is lying. Yes - the data are flawed, and they have been since the dawn of population surveys. This is nothing new. No pollster can get a representative sample of eligible voters. It’s simply not possible.
The pollster’s job, in my view, is to try to understand why they can’t, and to attempt to disentangle the signal from the noise. It’s the job of a good pollster to spend hours thinking about sources of error, and considering ways to reduce it, cancel it out, or otherwise adjust for it. They can’t always get it right, but that’s the nature of measurement in a context where there are so many variables.
-
Hard News: Decidedly Undecided, in reply to Sacha,
I don’t think so. The problem there was interpreting that NZ First was ‘out’ because they were on 4.8 or 4.9%.
In our May poll we had NZ First on 4.8%, so we reported two alternative seat conversations – one where they cross the threshold and one where they don’t. Both alternatives were reported in the media story too.
-
Hard News: Decidedly Undecided, in reply to Sacha,
Sorry I don't get what you mean there.
I'm pretty much against reporting party support percentages to 1dp, because that suggests a level of precision that no poll can provide. Actually, we've had people tell us that we shouldn't even report to whole numbers - but provide only ranges. If we did that though, we couldn't convert to seats and it would be hard to tell how close things are under MMP.
The only reason we report <5% parties to 1dp is to show how close they might be to the 5% threshold.
-
The only intent counted is very or quite likely to vote - the others are quite or very unlikely to vote, or unsure.
-
Hard News: Decidedly Undecided, in reply to Sacha,
It differs between each poll - in the link above you can see how much the base reduces from the total.
-
Hard News: Decidedly Undecided, in reply to Sacha,
If you check out Page 7, Question Order and Wording, that might help clear it up.
-
Sorry – I probably wasn’t very clear.
Very or quite likely refers to likelihood to vote. Anyone else is unlikely to vote, so is not included in the party support analysis.
Weakly decided refers to those who haven’t made a decision, but say they are likely to vote for a particular party. They are included in the party vote analysis.
So you can be weakly decided, but still likely to vote. Which, as it happens, is me! Sometimes. :)
-
No, weakly decided likely voters are included too.