There has been a great deal of back and forth about an Opinium poll for the Observer showing the Conservatives seven points ahead, and what it means for, variously, Jeremy Corbyn, the polling industry, the Conservative Party, Theresa May, Brexit and God knows what else.
In the real world of opinion polling, however, the result is just noise, statistically speaking. We’d expect, given that the pattern of the polls overall, for the polls to throw out the odd seven-point Conservative lead. I wouldn’t take that any more seriously than any single poll showing an equivalent Labour lead, ie, not at all seriously.
It is noteworthy that we seem to have a situation in which three of the pollsters are telling us slightly different things: YouGov is producing small Conservative leads as a matter of course, ComRes and Survation tend to show small Labour leads, and everyone else tends to produce ties.
Different pollsters tell us different results because they respond to the challenges of opinion polling in different ways. Polling companies have two problems: the first is the age-old problem that people are bad at accurately predicting how they will vote or behave at an as-yet-unspecified future date; the second is the comparatively recent problem of getting too many politically engaged people in their samples. (Most people do not follow politics that closely, but pollsters are increasingly struggling to reach people who only follow politics at election time or pay only limited attention. As this group often swings elections this is a big problem.)
Think of the problem that a pollster has as the same problem you have when you arrange a social event on Facebook. You know that everyone who clicks “Not going” definitely won’t be attending, and that anyone who clicks “maybe attending” probably won’t either. But how do you work out how many people who click “going” will actually show up? That’s a much harder question.
Added to that problem, we don’t know when the next election will be, or who the candidates will be. At the moment, saying who you’ll vote for is like clicking “attending” on an invite to “Jane’s Birthday Party” without a date or a venue. You’ll have some vague idea roughly when Jane will be having her birthday party, but she could move it forwards or backwards depending on events. Some people are so committed to Jane that they will go to her party no matter what, even if they have to cancel or rearrange their plans. Others are so opposed to Jane that they will do anything, even sit alone in their flat and stare at a wall, rather than attend her birthday party. Most people will make a decision nearer the time, when they know when the party is happening, where it is happening and how much it will cost.
The same is true of elections and this is the big problem that pollsters need to combat. Pollsters do a variety of things to combat this, which is why the polls are producing slightly different headline results depending on the polling company conducting them.
So what use are polls? Well, they still have a lot of use. In 2017, while the polling companies gave us different headline results, they all showed the same story: Labour picking up votes from Conservative Remain voters, Liberal Democrats, the Greens, non-voters, and a small but significant slice of Ukip support; the Conservatives picking up Labour Leave voters and the vast majority of the Ukip vote. They showed Theresa May’s popularity falling, and Jeremy Corbyn’s popularity rising.
Just as in 2017, I would pay far greater attention to what people are saying about leadership and how different groups of voters are behaving than I would to how they are saying they will vote – and even then, it will only be useful as a yardstick for where the political parties start from when the election campaign gets underway.