Support 100 years of independent journalism.

  1. Spotlight
  2. Elections
4 November 2019

What do the polls tell us so far about the 2019 election?

The election is only just begun, and we have very little to go on – yet.

By Stephen Bush

Opinion polls are like the clock on your oven. Even if the time is wrong, you can still use it to measure how long your pasta needs to cook, because if nine minutes have passed on the clock, it does not matter if the time reads 20:09 or 15:09 – you know that nine minutes have passed.

Similarly, it does not matter if an opinion poll shows Labour up from 24 per cent to 28 per cent, or 29 per cent to 34 percent: what matters is that it shows Labour support on the rise outside the margin of error. At pretty much every election, analyses based on measuring change, whether it be “the Conservative Party is gaining support at Liberal Democrat expense” in 2015, or “Jeremy Corbyn is getting more popular since the election was called” in 2017, aged better than aphorism-laden analyses based on what the headline polls showed.

Even if the polls are broadly right, you’re better off looking at whether or not they are showing change than fussing overmuch about the level they show, while remembering that the margin of error is plus or minus three per cent and that any change within that margin might just be noise.

That’s a good approach to reading the polls at this, or pretty much any other election. So, what do the polls show at the moment? Well, not much.

This election – and indeed throughout the 2017 parliament – we’ve had a big problem: just one pollster, YouGov, is doing polls more than once a week. Each pollster has their own different approach to overcoming the challenges of doing polling (among those challenges: that we can’t reliably predict who will end up voting, that only politically engaged people respond to polling, and that people lie, to pollsters and to themselves). It may be that when the dust settles on 12 December we realise that this was fine because YouGov were actually the most accurate pollster. But as it stands it means that we have to be a bit more reserved in our judgements.

Sign up for The New Statesman’s newsletters Tick the boxes of the newsletters you would like to receive.
I consent to New Statesman Media Group collecting my details provided via this form in accordance with the Privacy Policy

Since the election was called on 30 October we have had nine polls. One of those pollsters, by ORB, is so infrequent as to be useless for our purposes, as the last ORB poll was in April 2019. The political situation has changed quite a lot since then, so we can learn nothing of use from it.

We have had two polls from YouGov, and a poll apiece from Survation, Panelbase, ComRes, Opinium, Deltapoll and ICM. The first YouGov poll showed the two parties’ standings essentially unchanged and the second showed a big Labour increase outside the margin of error. Survation, Panelbase, Deltapoll Opinium and ComRes show the position of both parties unchanged outside the margin of error.

ICM show the Conservatives gaining outside the margin of error, but their most recent poll before the election was called is more than a month old and I am highly reluctant to attribute those gains to anything.

So it is far too early to say, with any confidence, what “the polls” show us – though Labour supporters are now in the odd position of hoping that YouGov, the pollster that has shown the biggest Conservative leads, turns out to be the gold standard, and that its outside the margin-of-error increases turn out to be followed elsewhere.