21 Jun Do polls accurately represent public opinion? Or do they help shape it…
Polling has long been used by government, businesses and individuals to identify the public’s opinion on different issues, including where they lean politically.
Recently in New Zealand, two polls differing widely in results while measuring the exact same thing have captured the public’s attention. The polls in question were intended to measure the support of New Zealand’s political parties and their respective leaders.
The first poll, conducted by 1 News and Colmar Brunton reported that support for the Labour Party was at 42 per cent, down by six points with the National Party claiming majority support with 44 per cent, up four points. In stark contrast, Newshub and Reid Research produced a poll stating Labour had 50.8 per cent support, up 3.3 points, while National only had 37.4 per cent which is down 4.2 points.
Other recent polls in New Zealand that have produced vastly different results have been on the upcoming Cannabis Referendum. Many polls show narrow public support for cannabis legalisation; however, the latest poll from Newshub and Reid Research paints a completely different picture. Other globally controversial polls include Brexit and the many polls incorrectly predicting Trump’s support prior to his election.
So, what gives? How can there possibly be such a major discrepancy between poll results? Well, the answer may lie in how the polls are being conducted. Polling through landlines started to become quite popular in the 1980s and they were reasonably accurate with this method. Unfortunately for pollsters, most people under 70 no longer have landlines.
The 2017 election was the last time Colmar Brunton used landlines exclusively, they now use a 50/50 split of landline and mobile, while Reid Research uses 75 per cent phone and 25 per cent online polls. Might the internet be at fault? Perhaps, when it comes to polling through phones, the public is prompted to share their opinion externally by the caller whether their views are strong or not. However, when it comes to online polling, people have to proactively choose to participate, without being prompted. You’re likely to get people with stronger opinions participating rather than the ones who are indifferent or don’t really care.
If polls are only estimates, why do they matter so much? The reality is that most people conform to majority opinion, so when these results come out undecided voters or those without strong opinions may shift their stance to that of the majority. There have been many instances of politicians using polls as another political tool to further their agenda. One such politician who seems to be immune to the effects of polling is the President of the United States, Donald Trump, who has had terrible results in various polls yet still manages to distract his supporters from those results.
Considering the inaccuracies of conventional polls and their impact on the political landscape, can anything be done to improve them? Apparently so. YouGov has recently used the statistical technique of Multilevel Regression and Poststratification (MRP) to produce very accurate results. Pollfish has also been using an ad-based mobile collection technique which has seemingly been even more accurate.
While polls will never be entirely accurate, advances in technology gives us hope the margin of error will continue to reduce.