How the pollsters did in 2020

PartyResultONCBNRRRM
PollDiffMOEPollDiffMOEPollDiffMOE
Labour50.0%46.0%-4.0%No45.8%-4.2%No47.5%-2.5%Yes
National25.6%31.0%5.4%No31.1%5.5%No28.5%2.9%No
Greens7.9%8.0%0.1%Yes6.3%-1.6%No9.5%1.6%Yes
ACT7.6%8.0%0.4%Yes7.4%-0.2%Yes7.0%-0.6%Yes
Maori1.2%0.6%-0.6%No0.6%-0.6%No0.5%-0.7%No
NZ First2.6%2.6%0.0%Yes3.5%0.9%Yes2.5%-0.1%Yes
TOP1.5%1.1%-0.4%Yes1.3%-0.2%Yes1.5%0.0%Yes
New Cons1.5%1.5%0.0%Yes1.7%0.2%Yes
Advance1.0%0.9%-0.1%Yes0.3%-0.7%No
Total Gap11.0%14.1%8.4%

There were three pre-election public , from One News Colmar Brunton, Newshub Reid Research and Roy Morgan. With final results, we can now compare how they did.

The table above shows the results, the poll, and the difference. The MOE column indicates whether the difference is within the 95% margin of error.

ONCB and NRR had Labour lower than the actual result and outside the margin of error.

All three polls had National higher than the actual result and outside the margin of error.

This is a fairly significant “miss”, but I think it is understandable. Polls can be self-fulfilling prophecies and I think a fair number of National voters simply didn’t vote or voted tactically. A poll is a snapshot at a point in time.

ONCP was spot on with the Greens, NRR was too low (outside MOE) and RM too high (but just inside MOE).

All the polls got ACT broadly correct.

All three pollsters had the Maori Party too low and outside the MOE.

ONCB and RM were spot on with NZ First. NRR were too high but just within the MOE.

All three polls had TOP correct.

The TV polls had New Conservatives correct.

ONCB had Advance correct but NRR had thenm too low.

So which poll was most accurate?

If you take total difference, then this time Roy Morgan did best with 8.4% vs 11% for ONCB and 14.1% for NRR. If you average that per result then RM and ONCB had an average 1.2% error and NRR 1.6%.

Another way of looking at it is how many of the results were within the margin of error. They are:

  • ONCB 6/9
  • NRR 4/9
  • RM 5/7

Now let’s look at the polling

CannabisYes ResultYes PollDiffMOENo ResultNo PollDiffMOEMarginActual
NRR48.8%40.8%-8.0%No51.2%59.2%8.0%No-18.4%-2.3%
ONCB48.8%44.6%-4.3%No51.2%55.4%4.3%No-10.9%-2.3%
Greens48.8%51.4%2.6%Yes51.2%48.6%-2.6%Yes2.8%-2.3%
Research NZ48.8%53.5%4.7%No51.2%46.5%-4.7%No7.0%-2.3%
UMR48.8%52.1%3.3%No51.2%47.9%-3.3%No4.3%-2.3%
Horizon48.8%52.5%3.7%No51.2%47.5%-3.7%No5.1%-2.3%

Note the results exclude informal ballots and the polls exclude undecideds so it compares like and like.

The only poll that was within the margin of error was in fact the internal polling released by the Greens. They were only 2.6% out from the actual result.

The two TV polls were the only ones that got the overall result of a loss correct, while the other four incorrectly forecast a win.

However in terms of the actual gap between the poll and the result the closest were the Greens and UMR and the furthest away was NRR.

EuthanasiaYes ResultYes PollDiffMOENo ResultNo PollDiffMOEMarginActual
NRR65.9%62.7%-3.2%No34.1%37.3%3.2%No25.4%31.8%
ONCB65.9%64.5%-1.4%Yes34.1%35.5%1.4%Yes29.0%31.8%
Research NZ65.9%72.1%6.2%No34.1%27.9%-6.2%No44.2%31.8%

For the euthanasia poll, the only poll within the MOE was ONCB. They had the yes result just 1.4% below the actual result. NRR was 3.2% below and Research NZ was 6.2% too high,. Note the Research NZ poll was in August though.

They all were right in clearly predicting a comfortable victory for yes, but ONCB got closest to the actual result.

Comments (17)

Login to comment or vote

Add a Comment

%d bloggers like this: