From the outset, opinion polls have been under a cloud in this election. Two years ago, the final polls put the Conservatives and Labour neck and neck. In the event the Tories were seven points ahead. Many feel — much less fairly — that the polls got the EU referendum result wrong too.
When this election was called it looked as though it was not going to pose much of a challenge to the polling industry. True, working out how to remedy the errors of 2015 was still a work in progress. However, given how far the Conservatives were ahead, this appeared to be a contest where the outcome was going to be easy to call.
However, the considerable narrowing of the Tory lead registered by every single polling company has made the polling industry’s task much more difficult. Now, it seems, just a few percentage points either way could make the difference between a healthy Tory majority, a relatively narrow one, or maybe not even one at all. Establishing which of those scenarios will materialise on Thursday requires a degree of accuracy that it would be difficult for polls to provide at the best of times, let alone when they are trying to overcome past mistakes.
Indeed, one of the difficulties that bedevilled the polls two years ago looks as if it could be an even bigger problem this time around: working out who is and who is not going to make it to the polls. Inevitably, pollsters find it much easier to interview those who are interested in politics than those who are not. As a result, they can all too easily overestimate the likely turnout among key groups, such as younger voters, who tend to stay at home.
In 2015, younger voters swung to Labour while older ones moved towards the Conservatives. That made it more important than before to estimate correctly how many younger voters would make it to the polling booths. In practice, pollsters overestimated that how many would do so and that contributed to the overestimate of Labour support.
Advertisement
This time, the divergence between younger and older voters in how they say they are going to vote is even sharper. Most polls report that about three-fifths to two-thirds of 18 to 24-year-olds will vote Labour, and three-fifths to two-thirds of those aged 65 and older will back the Conservatives. That means that working out how many younger voters will actually vote has therefore become crucial to estimating correctly the eventual outcome. If young voters stay at home, Labour will do less well; if young voters turn out, the Conservatives are likely to suffer.
Because of the problems of 2015, almost every pollster has changed how they do their polls. Expensive and difficult though it can be, some are trying to ensure they interview more people with little interest in the election. Many are weighting the interviews they do obtain to correct any apparent deficiency in those with little interest. Meanwhile, some polling companies are relying on high-quality academic evidence on the level of turnout among different groups at the 2015 election to estimate how many younger voters will make it to the polls this time.
However, no two pollsters are addressing the issue in exactly the same way. That helps to explain why some polling companies have consistently been reporting higher Conservative leads than other companies. Trouble is, nobody can be sure which approach will prove best.
Still, not everything about this election has made polling more difficult. The best predictor of how someone will vote this time is how they voted last time. One might imagine, therefore, that pollsters should always ensure that how people say they voted at the previous election matches the actual result. Trouble is, people sometimes forget how they voted.
However, because the last election was only two years ago, such forgetfulness is less of a risk. Consequently, many pollsters are ensuring that how people say they voted last time matches the 2015 result. That should also help to reduce any pro-Labour bias.
Advertisement
The pollsters are certainly not simply going to repeat the mistakes they made in 2015. But we will discover only on Thursday night whether the numerous changes they have made have been enough.
John Curtice is professor of politics at Strathclyde University and president of the British Polling Council