Should we trust the polls?
September 22, 2017
The recent election upsets of Brexit and Trump have placed doubt over polling reliability. Whilst New Zealand has a good track record for accuracy, this campaign has seen huge discrepancies between polling results. The most striking was between the recent Colmar-Brunton poll which put Labour 4 per cent ahead of National whilst Reid-Research had National 10 per cent ahead.
To work out how the pollsters could produce such starkly different outcomes, we spoke to the polling companies about their methodologies and what factors would likely cause an election upset.
Is landline surveying to blame?
With the exception of Newshub-Reid research, all the polling companies use landline surveying. The common misconception is that landline surveying skews the results towards older voters. Polling companies have quotas they must fulfill for age, gender and region, based on the latest census data.
Reid Research managing director Ngaire Reid said that in the past the company has struggled to find a 25 – 30-year-old male from the West Coast. The hardest demographic to meet was under 35s.
“It’s been very difficult to get young people, you just ring and ring and ring,” she said.
Newshub now solely uses online polling to survey 18 – 24 year olds and most 25 – 35 year olds. It’s all a matter of speed and responding to consumer research which says landline use is in decline, according to Dr Andrew Zhu, whose company conducts the internet polling for Newshub.
Landline polling has become hard to achieve as a third of New Zealanders no longer have a home phone. Dr Zhu said that online surveying wouldn’t change the result “too much”, but he would still be paying close attention on the 23rd of September to see if it was accurate.
Could a high youth voter put the polls out?
A high voter turnout in New Zealand makes the chance of a shock result unlikely. Both the U.S and the U.K have voter turnouts in the low 60s, whilst NZ’s voter turnout remains strong at 76 per cent.
UMR mitigates for voter turnout by excluding voters who answer that they are uncertain whether they will vote, whilst Colmar-Brunton is confident that high voter turnout is enough to avoid this issue.
Colmar-Brunton chief executive Jason Shoebridge said the polls would likely be thrown if older voters stayed home rather than a high young voter turnout.
Oftentimes when two polls appear outside the margin of error it is a case of when the polls were conducted. The starkly different Green party projections in August were largely influenced by when Green’s co-leader Metiria Turei resigned.
The Colmar-Brunton poll was conducted after her resignation which placed them on 4.3 per cent. Reid-Research placed them on 8%, their polling taking place just a day shy of the leadership fiasco.
Is a polling embargo necessary in New Zealand?
Many countries overseas have introduced bans on publishing polls prior to the election. A 2012 study by the University of Hong Kong found that 38 out of 83 countries surveyed had a blackout period for publishing pre-election opinion polls.
There have been unsuccessful attempts in NZ to do the same. Winston Peters introduced a bill to ban polling 28 days from the election in 1999. He believed that voters were simply voting for the party they thought would win or whoever was performing best.
UMR Executive Director Stephen Mills believes polls provide necessary information for voters to make strategic decisions.
“It is vital for MMP, voters need to know if their party will meet the five per cent threshold,” he said.
Voters also make strategic decisions in their electorate if they think a party needs extra seats or if the seat is needed to form a coalition, as in the case of the Ohariu electorate.
Victoria University professor Jack Vowles agreed that polls provided critical information for voters to make their decision.
“A polling embargo would be a restriction on free speech. Polls provide voters with information that can help them make a choice – for example whether or not a party is likely to cross the threshold to gain seats therefore indicating the risk of wasting their vote,” he said.
A polling embargo under MMP would also likely result in a rise of wasted votes. A 2015 Oxford study found the number of wasted votes increased in countries with highly fragmented party systems when pre-Election Day polls are restricted.