Public trust in election polls has never been lower. Recent media coverage has highlighted repeated failings of election polls globally. The Guardian’s Alan Travis recent opinion piece summarised this with the headline “Can we still trust opinion polls after 2015, Brexit and Trump”. His answer? “It’s complicated”.
Why polls get it wrong
Concern over the accuracy of polls has been well documented.
One of the biggest issues has to do with pollsters’ recruitment methods. They generally omit overseas and postal voters and remove anyone who refused to answer or said that they would not vote. It is a serious mistake to disregard the latter. These respondents should be used to forecast the possible voter turnout. Indeed, voter turnout has a great impact on election results, and is often the hardest element to foresee.
Others blame the respondents’ behaviour for polling inaccuracy. Some voters may not know who they support, some may later change their minds, while others may lie about their real intentions.
Voters can oscillate between parties, or move from undecided to decided, but this behaviour is often ignored by the polls. However, errors due to recruitment or respondents’ behaviour actually have a marginal impact. One should always assume a 3% margin of error for any poll.
Interestingly, not all of the recent polls were wrong – despite this being the common perception. Ipsos MORI did predict the outcome of the EU Referendum a week before the vote, giving 53% to Leave and 47% to Remain. YouGov later placed a Leave outcome ahead by two points, just two days before the referendum result.
What recent polls show
Pollsters had Theresa May twenty points ahead of Corbyn on the day she called for the snap General Election. But with a week left, May’s lead has considerably narrowed, with Survation (27/05) putting the Prime Minister ahead with only six points, YouGov (26/05) with seven, and Opinium (25/05) with ten.
How can this be?
Voters could be warming up to Corbyn, and the Conservative Manifesto could have put off some respondents who previously intended to back May. Another possibility could be that some Labour-supporting respondents lied to pollsters early on about their intentions as a protest against Corbyn, but have now come to terms with him, or see no other alternative.
Favourability in press coverage and in the polls may also encourage others to support Corbyn. This trend is likely to be sustained as long as May’s campaign countines to attract unfavourable coverage, such as her perceived ‘dementia tax’ u-turn.
Will the polls be right this time?
It is important to remember that polls show us ‘intentions to vote’, and are not ‘actual’ votes. Much of it will also depend on how accurately pollsters will be able to predict the outcomes for all 650 constituencies.
For while one candidate may be prepared to win the popular vote, First Past The Post still means everything comes down to how many MPs for each party gets elected. It is fair to assume that the race for Number 10 is far from a done deal.
Measurement and evaluation