Opinion Poll Issues

Opinion polls in the UK 2015 General Election and 2016 EU Referendum failed to accurately predict the results, and in the US there is concern that they are nat accurately reflecting opinion in the 2016 Presidential election. Why might this be?

The British Election Survey conducted a poll of 3,000 adults after the 2015 Election accurately reported a seven per cent lead by the Conservatives over Labour, while pre-election polls predicted the two parties would get the same percentage.

The BES data was collected – at considerable cost – by going out and knocking on doors. It sheds valuable new light on who was undersampled in the pre-election polls.

The new thought is that opinion polls are not currently choosing representative samples.

• Online polling draws samples from large panels of volunteers. Not everyone puts themselves forward to take part, and those who do will not be representative.
• Landline polling is less useful as many households are increasingly reluctant to pick up their landline, and fewer households actually have a landline unless is it used for Internet access.
• Mobile phones are used by many people, and a growing proportion of young people in particular rely exclusively on them. There is no exhaustive database of mobile numbers to draw on, and there is no consensus among the pollsters on a comprehensive and cost-effective way to factor mobiles into the mix.
• The BES study identified the crucial missing group as non-voters, and particularly younger non-voters. It found 43% of the under 30s saying they didn’t vote, whereas among the less representative young people in the pre-election polls, 15% or fewer sometimes identified themselves as non-voters.
• Opinion pollsters are very good at making their samples reflect the general population. But the general population and the electorate are very different things, because around 40% of adults don’t vote. This means there are far too many young voters in polling samples, and not enough young non-voters. 

In the US there is concern among political reporters that the opinion polls on the Presidential election are similarly not representative. The polls still rely on landline phone polling - and as with the UK votes, younger voters increasingly use mobile phones.

• While poll modelling looks at people who might be under represented, who might not be picked up by telephone or online polling, many of the people who voted for Trump in the Republican primaries had been lost to the political process for years, so are probably not part of an opinion pollster's sampling group. 
• There is also the phenomenon of the 'shy' supporter, people who are unwilling to state a preference for controversial nominees.

Sources: The Guardian online, 13 Nov. 2015 and The Guardian online 10 July 2016 .