Some polling observations.

Labour is 22.3% ahead of National on prelim results - a margin likely to grow.

How close did pollsters get to this?
UMR: 21% (1.3% off)
Roy Morgan: 19% (3.3% off)
Colmar Brunton: 15% (7.3% off)
Reid Research: 14.7% (7.6% off)

So UMR did best.

1/

These are big misses by the TV polls.

They’re akin to them having Labour ahead by 7 in final polls and then on election night National winning by a nose.

They’re way bigger misses than occurred w/ Trump/Clinton and Brexit. Hopefully they’ll look into why this happened.

2/

If you look at how close these four pollsters got to the 5 largest parties’ support levels, it’s the same order of performance.

Total error across 5 parties (from best to worst):
UMR 5.7%
Roy Morgan 6.4%
Colmar Brunton 7.8%
Reid Research 10.3%

3/

In general, the polls were too friendly to National, and too unfriendly to Labour. Compared to an average of these four companies’ final polls:

Labour outperformed 1.8%
National underperformed 3.1%
Act outperformed 0.6%
Greens outperformed 0.2%
NZF underperformed 0.2%

4/

We should revisit this after final results are announced. My guess is that the network polls will look even worse.

The party vote polls being off makes me nervous about the referendum polls by the same companies. Some mid-modelling of the electorate appears to have gone on.

5/5