Why do polls keep getting it wrong?

Donald Trump has just been elected to become the new US president but looking at the polls this might come as a surprise as most were in favour of Hilary Clinton. This is not the first time this year that the pollsters didn’t accurately predict the likely outcome of the public voting. Going back to June, the UK was deciding whether to remain or leave the European Union. And whilst predicting it to be quite narrow, most of the polls were putting Remain in the lead.

Various organisations in the UK undertake polls which collect voting intentions data. From TV stations and newspapers to specialised market research agencies (e.g. YouGov, Comres, Ipsos MORI, ICM). Most of these market research agencies are members of the BPC (British Polling Council) which is an association that sets the standards in public opinion polling. The main aim of the organisation is to oversee that these companies produce representative data.

So, if polling organisations are using the representative sampling methods and weighting procedures, how come their polls are not accurate?

One of the factors could be the changes in polling methodology, how the data is collected. In the past, it used to involve going door-to-door and asking people face-to-face. Then the data started to be collected through phone, but nowadays data is increasingly gathered through online sources. And with that comes the issue of how accurate the data is and how representative it is of the population. It is known that some demographics are harder to reach via online methods and therefore to produce accurate weighting to ensure representativeness can be challenging.

Then there is a phenomenon of ‘shy voters’ (sometimes called ‘Shy Tory factor’ – firstly observed in 1992 parliamentary elections in the UK) where some voters do not want to admit or refuse to disclose that they will be voting for the candidate (or option) that is less ‘socially acceptable’. In this digital era where social media plays such a big role in our lives, it is much harder to go against the mainstream opinions, especially if it not supported by one’s peers.

These are just a few of the factors that can play a part in inaccurate prediction polls, similar to those seen this year. The challenges that the polling agencies are dealing with can easily be projected to market research as well, so we should take learnings from these too.

By Tereza Krtickova, Research Executive


This entry was tagged Politics, Polls, Election, Methodology and posted on November 10, 2016

<< Back