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

Tuesday, 18 July 2023

pn985. A Non-significant Poll Result is an Insignficant result or No Result at all

The latest 1NewsVerian (was 1NewsKantar Public) poll conducted between the 8th to the 13th of July showed National with  35% party support, down  2 from the previous poll in May, Labour with  33% down  2, Act with 12%  up 1, and the Greens with 10%  up  3. 

This result would see National/Act able to form a government with 61 parliamentary seats with Labour/Greens/Te Pati Māori with 59 seats —  a very close result. 

 Varian noted that the result is a snapshot in time and not a prediction.  

On the whole Varian are excellent pollsters who fully reveal their methodology (see attached) for the full report) but I have some issues with the poll and its conclusions which I will explain later.

Events around the time of the poll that could have influenced the result included National's announcement that gang membership would be an aggravating factor in sentencing, and Labour announcing the results of a market study on banking. There were a string of Labour negatives including Michael Wood's resignation, Kiri Allen's non-disclosure of interests, Jan Tiretti not correcting incorrect statement, and two positives — the PM's trade deals with China and the European Union, though these came too late to influence all of these polled.

So what can we take from the poll which had a sampling error of ±3.1% for results around 50% and ±1.9% for results between 10-20%? The answer is very little. Most results were less than the required sampling error. 

They were also "weighted", i.e., increased or decreased in number for age, gender, region, ethnic identification and education level so that the sample resembled that of the voting age population as a whole. 

Add to this, the results were rounded so National's 33% could have been  34.6 and Labour's  33 33.4, a difference of 1.2: hardly a sufficient margin to make any  call.  

 Also, 12%, labelled undecided voters, didn't know or refused to say how they cast their Party vote.  These 120 people of the 1,000 polled were excluded in the results.

The poll asked those polled for their "preferred Prime Minister." Hipkins had 24% down 1% on May,Luxon 20% up 2 and Act's Seymour 7% unchanged.Twenty others received some support, none more than 2% and 39% were other, don't know, none and refused.  I'm unsure why this question is asked. Afterall, we never vote for Prime Minister. The winning party decides who that shall be.

I'm unsure also why, with an MMP system that gives us all two votes, the one for party and the other for electorate candidates, Verian only asks for the party vote. 

Many people, including myself, use their electoral vote to vote for one party and the party or "list" vote to vote for another party. 

The inclusion of this question would make for greater accuracy and could greatly change the poll results.

-- ACW

No comments: