The RANZ Political Polling Code provides rules for RANZ member organisations who conduct and report on political polls, plus best practice guidelines for those publishing stories on poll results.

The 2020 code review was conducted by a team of polling experts and takes into account technology changes in data collection and reporting methodologies since it was first introduced in 2014.

DOWNLOAD PRESS RELEASE

Released 26 June 2020

By Nicole Satherley, Lara M. Greaves and Andrew A. Sporl
iNZight Analytics; School of Social Sciences, University of Auckland; Department of Statistics, University of Auckland

At the request of RANZ Fellow David Fougere, Murray Campbell, Fellow and RANZ Political Polling Spokesperson, and Geoff Lowe, RANZ Chair, were invited to provide feedback on this academic paper offering a guide to understanding and reporting on public opinion polling in New Zealand.

The paper references documents already available on the RANZ website, including the RANZ Political Polling Code 2020, and its message is endorsed by RANZ.

Abstract

Public opinion polling provides a platform to elucidate the public’s attitudes towards and support for various issues, and allows politicians to learn of and respond to these attitudes. However, a low level of understanding about how polls work, and a lack of communication and transparency about the methods used for a poll, can impede this function. Although ample resources on the topic of public opinion polling have been produced across international organizations, these can be difficult to navigate and piece together for a lay-audience. They also cannot provide information on political polling as it relates to specific contexts, such as New Zealand’s unique mixed member proportional electoral system. Here, we provide a guide to understanding and reporting on public opinion polling in New Zealand. The guide covers key information on how polls work, aspects of polls that speak to their quality, including sample size, error, and sampling methods, and how political polling relates to actual party representation in New Zealand parliament. By identifying and explaining key aspects of public opinion polling, and why they matter, we hope this guide facilitates improved poll transparency and standards of reporting among journalists and media, and overall understanding of poll results by poll consumers.

GO TO REPORT

In the year of COVID 19, the 2020 election was certainly one of the most drawn out and challenging to follow and predict for all stakeholders, including journalists, pundits and pollsters.   

All the polls had forecast a clear majority for Labour; the question was whether Labour could govern alone or in partnership with the Greens.

As shown below, the two media polls by ColmarBrunton and Reid Research conducted in the last week leading up to election day were remarkably similar, and close to the share of votes for the Greens, and ACT. Both polling companies also predicted correctly that NZ First (and other minor parties) would fall short of the 5% threshold.

But the fortunes of Labour and National were not so easily foretold. ColmarBrunton’s and Reid Research’s last polls in the days immediately before October 17^th predicted near identical results for Labour (46%) and National (31%).  In reality there was a greater gap between the two main parties.

The shared systemic underestimation of the gap between the two main parties’ level of support has, to a large extent, been attributed to the very high number of early voters. In the lead up to this election almost 2 million people had voted prior to October 17^th versus 690,000 on election day. In the previous two elections early voting was not the norm (1.24 million in 2017 and only 717, 600 in 2014).

As noted by the Newsroom on October 18^th the National Party's pollster, David Farrar, tweeted on Saturday night that Labour's party vote from advance voting was 51 percent and from the election day was closer to 44 percent (figures confirmed by post-election analysis run by Stuff https://interactives.stuff.co.nz/2020/10/election-2020-results-analysis-labour-day/).  National, on the other hand, was 26 percent from advance votes and 30 on election day.

So, the anticipated late swing in voter support away from the incumbent to the opposing parties may have occurred in the last week but its affect was heavily discounted by the high proportion of early voters who supported Labour.

By the time those last pre-election polls were in field, a significant number of respondents had already voted. Asking them who they would vote for may not be the same as asking who they did vote for. That might explain the difference between the final polls and the election night results. All pollsters may need to ensure, for future elections, they can calibrate their questions and analysis to take into account the significant impact of extensive early voting on their final election night predictions.

Murray Campbell, RANZ Fellow and Polling spokesperson, October 2020.

Why do people conduct polls? Why do election polls get it wrong sometimes? Should all polls be vetted by the government?
The ESOMAR Government Affairs and Professional Standards team has compiled a list of FAQs for journalists and opinion leaders on Opinion and Election Polls to give confidence in tackling the latest polls.

DOWNLOAD FAQ HERE

The lead up to any election is always characterised by unscripted moments, own goals and the occasional red card. The 2020 election season has been particularly dramatic, especially for the National Party. The ever-changing political landscape making it hard for the voting public and the pollsters to keep up with what’s coming next.

As shown below in the timeline, in mid to late May, the two public polls ColmarBrunton and Reid Research recorded very similar results for the Simon Bridge’s National Party (29% and 30.6%), which culminated in a leadership challenge and led to Todd Muller taking over as National Party leader. The next ColmarBrunton poll in 20th-24th June recorded a solid recovery for National, up 9 points to 38%. But the period from July 8^th to July 21^st was an especially turbulent period for the National Party, involving the resignation of a leader, the standing down of an ex Party President and a junior MP, the appointment of a new leader Collins, and then another series of stand downs, demotions and sackings.

Reid Research started polling on the 16^th, just as the centre of this turbulence was passing over the National Party. Given these events it’s perhaps not surprising that National polled lower than the pre-Bridges leadership change (25% vs 30.6%). Note that this result was still higher than its worst election night result of 22% (2002).

The CBR poll started 9 days after the Reid Research poll, with relatively clean air. Even so, National’s result with Collins as leader at 32% is only 3 points higher than National’s vote when Bridges was toppled (29%). Collins has to consider as possibilities that she hasn’t made much difference from Bridges' low point and is doing worse than Muller (6 points) or acknowledge that she has brought the party forward from the abyss of the events between July 8^th and July 21^st. The latter supports the notion that the Reid Research poll was an accurate and fair measure of support for a party dealing with a perfect storm of rogue waves between July 8^th and the 21^st. 
Released 31 July 2020

Historical polling statistics can be found on the Kiwiblog website HERE

Blogs by Gavin White (UMR) analysing polling data and the 2014 election results:
https://web.archive.org/web/20180123134029/http:/sayit.co.nz/blog/how-did-we-do

https://web.archive.org/web/20180123134042/http:/sayit.co.nz/blog/its-crunch-time
https://web.archive.org/web/20180123134055/http:/sayit.co.nz/blog/are-polls-getting-better-or-worse

https://web.archive.org/web/20171115092908/http:/sayit.co.nz/blog/what-political-polls-tell-us