Provisional results are in. Here’s how the polls have done so far.
You can find the detailed comparisons in this Google doc. Here are a few notes:
- In 2014 the official results changed things quite a lot. I’ll post an update when they’re out in a couple of weeks.
- Well done Reid Research – outstanding job and huge improvement over 2014.
- At this point, the Herald Election forecast has done the best of all the forecasts and polls-of-polls. They’re at least as accurate as the Colmar Brunton poll.
- Roy Morgan (the only poll to call cell phones) has not done so well this time, but they did not release a pre-election poll. Unlike the other public pollsters (who have developed bespoke election poll methodologies) Roy Morgan include their preference question in an omnibus survey which runs to a fixed schedule – their poll finished in field on 10 September.
- There is still no “polling crisis” in New Zealand. The established polls have done pretty well, once again.
Unfortunately, with only three public polls left in New Zealand, the forecasts and polls-of-polls don’t have much data to inform their models and calculations. It can’t help either that polling methods change, making house effects unstable over multiple elections.
While they don’t perform as well as the actual polls at predicting the election result, the forecasts and polls-of-polls are really useful for reducing the noise, and understanding if voter sentiment is changing over time.
UPDATE: DPF shared the pre-election results for some other polls I wasn’t unaware of, so I’ve added these.