Final polls vs the 2014 election actual result

How did pollsters go last election?

Pretty well. The actual results were well within their margin of error (see tables below) and close to the actual National and Labour vote.

But our average of the five final-week polls does reinforce two polling phenomena that have now been observed for several elections in a row.

One is that the Green Party vote is over-estimated.

As Green support tanked in a rash of polls after Metiria Turei resigned, some of the party’s supporters took to social media to complain that its youthful supporters lacked landlines, so were not captured in the surveys.

The counter argument is that most of the polling companies are now using a mix of online, landline and mobile phone responses, and all are weighting their samples to match New Zealand’s demographics at the last census.

They all have the right percentage of millennial voters surveyed.

The real problem is the under-30s vote in lower numbers than any other age group, so some of that Green support goes begging.

With NZ First, I suspect we’re seeing a “shy NZ Firster” phenomenon, just as there was a “shy Trumper” effect that saw some of Donald Trump’s supporters too sheepish to admit they would vote for him when rung by pollsters (but not internet surveys) in the run-up to the US election. As NBR observed pre-election, this is nothing new, tracing its roots back to the so-called “Bradley Effect” in California in 1982, when many respondents were too sheepish to admit which way they intended to vote in a racially-charged contest).

With pollsters being dumped on left, right and centre, it’s also worth noting while the Shy Trumper effect had its limits in the US, yes, it did distort some polls (to the degree that there were many) in a handful of rust-belt states that proved crucial to the outcome under the first-past-the-post Electoral College system. But Hillary Clinton actually carried the nationwide popular vote (analogous to our MMP list vote) by 48.0% to 45.9% – that is almost exactly mirroring the final day of tracking polls,  which had Ms Clinton an average 2.0% ahead.

The UK’s Brexit referendum in June last year didn’t look so good for the pollsters.

A BBC poll-of-polls showed a tightening race, with “Leave” ahead in four of the final 10 polls it quotes. Yet, nevertheless, the final day’s polling average was Remain 45%, Leave 44% – when, in the actual referendum, “Leave” won by four points.

The pollsters’ explanation was that the heat of controversy over Brexit drew in people who don’t usually vote, throwing their carefully-calibrated weightings out of whack as some demographics voted in larger numbers than expected.

However, the pollsters re-jigged, and at this year’s UK general election they correctly anticipated an increased youth vote as Labour ran a resurgent campaign.

The final BBC poll-of-polls was right on the money, predicting Labour would get a much higher vote but still fall slightly short of the Conservatives’ tally.

Voting at AUT today. LOTs of young people.

— Greg Treadwell (@hacademic) September 11, 2017

Advance voting opens
A final factor: Voting opened in our election today.

So any events from now – and any polls – will be coloured by the fact that some ballots have already been cast.

The Electoral Commission says that in 2014, 717,000 or 29.70% of registered voters cast an early vote.

This year, it’s anticipating that figure could be up to 50%.

Accordingly, it has boosted the number of advance polling booths from 295 last election to 485 – and many have been placed in areas where young people lurk in a bid to boost participation of the weakest group, the under-30s.

Related video: Robertson and Grant head-to-head, moderat

All content copyright NBR. Do not reproduce in any form without permission, even if you have a paid subscription.