Claiming victory from inaccuracy

I rarely criticise other polling companies, due to professional courtesy. NZ is generally well served by the various research companies and we saw Reid Research and Colmar Brunton get some very accurate pre-election polls.

Some other companies were less accurate, and that is also fine. All companies have polls that are less accurate than others. Curia has. Sometimes your poll is a bit dated. Sometimes you’re at one end of the margin of error. So long as one is constantly self-assessing your results and methodology, there is no shame that some polls are less close to an actual result than others.

But what annoys me is when a company tries to spin an inaccurate result as accurate. That just insults the intelligence.

So I can’t let past this release from Roy Morgan, which I quote from:

The final Roy Morgan New Zealand Poll of the campaign titled: ‘NZ Election on a knife edge but Labour favoured’ predicted a tight result with Labour favoured to form Government over the incumbent National-led coalition.

That poll predicted Labour/Greens would get 48.5% and National 40%. In fact National beat Labour/Greens by 2% rather than trailed by 8.5%.

Roy Morgan predicted a potential coalition of Labour/Greens/Maori Party would hold 62 seats in Parliament compared to 58 seats for the National-led Opposition. As it turned out the Maori Party failed to hold their two seats in Parliament which instead went to Labour and the combination of Labour/Greens/NZ First will govern with a majority of 63 seats compared to the National-Act NZ opposition which holds 57 seats.

This makes it sound like they were just slightly out. Actually the difference were large. Their poll and the results were:

  • National 50 seats, got 56
  • Labour 49 seats, got 46
  • Greens 11 seats, got 8
  • NZ First 7 seats, got 9

As I said there is nothing wrong with not getting the result right. They were disadvantaged by finishing their poll on 10 September, 13 days before the election. What I object to is the press release which selectively quotes their poll to make it look more accurate than it was.