Cogito, ergo sum. I think, therefore I am. (René Descartes, mathematician and philosopher,1599-1650)

Tuesday 16 March 2021

First poll of 2021: Pollsters take and record polls; the media need to do better in reporting them. (pn704)

The first TV1 News Colmar Brunton  poll of the year was published yesterday, which should give an indication of trends since the previous poll in December. 

The "trends" appeared to be there. Jacinda Ardern's was said to suffer a "personal blow", her popularity "crashed", while Judith Collin's career "survives thanks (for some reason according to Claire Trevett, NZ Herald) to the America's Cup".  Labour was down a little and National, Greens and TOP up a little.  There could be a  one-seat gain to National in the House.

 But what, in fact, has the poll shown us?   Why do I think it has shown us a confused very little?  

First, because the methodology was changed.  Whereas in December 60% were polled by cell phone and 40% by landline, this month's poll is 50% by cell phone and 50% "on line from New Zealanders who have signed up for market research studies"  who are paid for their efforts. Landlines reach more old people and have a better geographical spread; cell phones reach more young people, and possibly more Māori/Pasifika. 

Why the change?  Fewer people have landlines and polling by cell phone is quicker and cheaper.  Colmar Brunton says the results were "tested", and weighted as usual by age, gender, ethnicity and locality (TV1 did not say by how much for each variable) but I suspect far more weighting was needed from the market volunteer group, all adding to possibly increased error. 

A comparison of landline and the market group results would have enabled us to see how closely —or otherwise—they were related. Perhaps this "test" may be available later on the CB website. 

Secondly, because the figures do not add up. Add these percentages: Labour 49 (-4), National 27 (+2), Greens 9 (+1), Act 8, NZFirst 2, Maori Party 2, New Conservatives 1, TOP 1 (+0.7), Don't know 5, Refused to answer 4 (no indication of previous poll Don't know and refused).  Total 108. 

Colmar Brunton  says the "percentages do not add to 100 due to rounding". Partly. But the main reason is that they discarded the 9% don't know and refused to answer on the political parties question (and the 33% who answered the preferred PM and 29% state of the economy questions!) as if they were not an important statistic, Worse still, it was not reported that they had done so.

Not knowing the don't knows and refused to answer for December  makes proper comparisons even more difficult.    

Thirdly, with the  margin of error (±3.1%) unchanged, Labour's loss of 4% is not statistically significant. It could be due entirely due to polling error. And National, with a gain of 2%, is questionably significant. (Note, a MOE of ±3.1% only applies to scoring close to 50%; those with less have a larger MOE.)

The poll also asked about preferred PM and the state of the economy.   Jacinda Ardern was down 14%  to 43% and Judith Collins down 4% to 8% (but as noted a third, 33%,  of those polled didn't know or refused to answer). 

On the economy, some 42% (down 8) were optimistic about the economy and 29% (down 4) and a little less than one-third (29%) didn't know or refused to answer.   

My opinion is that these subsidiary polls are less reliable than the party poll, and the media fluster about Jacinda Ardern's popularity crashing is just that—media fluster. They have to have a story.

1 News'  Jessica Mutch-McKay thinks polls matter. I agree. They influence public opinion.

"I’m a big fan of polls," she writes. "I find them fascinating and love talking about them. Handy, given my current job. Polls are a snapshot in time and give us a flavour of how the public is feeling and allows us to see how policies [parties, generally; not policies--ACW]  and personalities are tracking. Political careers can live or die by polls. So, polls matter. Ensuring our polls are credible and accurately reflect the wider thinking of the general electorate is absolutely vital."  

Amen.

Colmar Brunton records results; the media reports them. The responsibility lies with the media to more fully explain what polls mean, may mean and don't mean. 

-- ACW

Useful link on poll methods


No comments: