Bugger the pollsters? Not this time
Newshub’s political editor Paddy Gower has been quick to brag about the accuracy of his network’s political polling in the wake of the election result.
The Newshub-Reid Research poll published two days out from the election got very close to the actual result (see below).
It is likely to prove even more accurate if most commentators' predictions of a slight decrease for National and a slight increase for Labour and Greens prove correct when all the special votes are counted.
Gower’s delight over the result is partly revenge, because TV3 copped it after the 2014 election when it had one of the least accurate final polls.
The winner in 2014 was the rival 1 News Colmar Brunton poll. This time, Colmar was off the pace and in sporting terms “had a shocker” with its second-to-last poll before the election.
The Colmar poll had Labour on 43 percent and National on 41.
At the time, Labour leader Jacinda Ardern commented that she was “surprised”, a clear indication that it didn’t match the party’s internal polling.
Former pollster, Dr Andrew Robertson, says the poll was almost definitely a rogue result.
“Looking back at it, I suspect the Colmar Brunton poll was an outlier, it is what happens when you are dealing with statistics – it is the random nature of sampling. It was bad luck for Colmar and 1 News that it happened so close to the election.”
However, the strong criticism international polling companies got after failing to pick up trends in the US and British elections is unlikely to be echoed here.
According to Robertson there is no “polling crisis” in New Zealand and “the established polls have done well, once again” (see below).
What does worry him is the decline in the number of polls.
“This is a big problem, there are now only two polls (Newshub and 1News) using bespoke methodology. Roy Morgan does one but they don’t follow the Research Association’s political polling code of practice.”
The two big newspaper groups, Fairfax and NZME did not run polls this election, instead doing forecasts based on the other polls.
“I think it is to do with cost and the Herald’s polling company Digi Poll closed down. Losing these polls is a problem - the forecasts and the polls of polls don’t have much data to inform their models and calculations.”
The way our pollsters gather their data is also changing.
This year Reid Research used an online panel to reach its quota of 18 to 35 year olds – 25 percent of the total sample. The remaining 75 percent were polled by the traditional method of calling landlines.
Dr Andrew Zhu who supervises the online sampling for Reid Research says the case for “online polling” is getting stronger.
“Twenty-five percent of people don’t even have a landline anymore and my research estimates that this will rise to 33 percent by the end of 2017. The reliability of the online panel we are using is getting better and better; therefore we have chosen to adopt online panels as a way to contact this hard-to-reach population.”
Andrew Robertson agrees that future polls will adopt a multi-method approach.
“Calling cell phones is not the answer so yes, I see a move to online as well as landlines and maybe even some door-to-door sampling.”
Both Robertson and Zhu feel criticism of the polls is largely unwarranted.
“I think unprofessional journalism is a bigger problem. Some media were reporting that National is making a comeback, whereas (apart from the one Colmar poll) National were remarkable steady in the polling. And it is often not made clear enough that the polls are taken at least a week earlier than when they are reported,” said Zhu.
Robertson’s view is that the polls provide a vital public service despite what some politicians think.
Former Prime Minister Jim Bolger once famously said “Bugger the pollsters” on the eve of the 1993 election and Winston Peters consistently rubbishes the established polls.
Robertson has some sympathy for Peters. “I think that 10 years ago he had a point, when the polls regularly underestimated NZ First support but voters have no other bellwether to measure politicians against – particularly when they say they are leading an electorate and it turns out they are not. There is no other way to get a view of what the public really think."