Should we trust the general election 2017 polls?

The latest polls for the general election 2017 suggest support for the Conservatives is slipping. But pollsters have been wrong before. 

Sign Up

Get the New Statesman's Morning Call email.

At the beginning of the general election campaign, the polls predicted a Conservative landslide, and Labour’s worst election since 1935. But since then, some polls have suggested a very different story. A Times/YouGov poll predicted the Tories would fall short of an overall majority by 16 seats, leading to a hung parliament. 

Opinion polls have been wrong before. The Conservative majority in the 2015 general election came as a surprise after opinion polls had widely foreseen a very tight race between Labour and the Tories. And in 2016, both Brexit and Trump appeared to defy the opinion polls. 

General election polls have undergone hard scrutiny since those failing. A lot of effort has gone into correcting the general election polls’ tendency to overrate Labour’s chances. So do the pollsters deserve to be believed now, or are general election polls no better than looking at the latest general election odds for a glimpse into the future? 

One indication that is probably more reliable than general election polls is local election performance. According to that, general election polls could be underestimating the Conservatives.

Should we believe the general election 2017 polls?

Yes and no - and here's our guide to why.

How accurate have polls been in the past?

The 2015 general election 2015, the EU referendum and the 2016 US presidential elections all gave the polling industry a bloody nose. In those results, polls flattered the left and underestimated the right, but Glen O’Hara points out that in the 1983 and 2010 general elections, it was Labour that exceeded expectations.

But Anthony Wells of YouGov argues that the polls actually got Brexit right (calling a narrow result, veering towards leave). The US presidential result was a genuine upset, but again the polls were right to call a Clinton victory in the popular vote, which she won by 3 million. However, they were wrong about which states that vote would concentrated in - which meant that the electoral college skewed heavily towards Donald Trump. And in 1983 and 2010, the polls got the Tory victory right, but overestimated its size.

So while polls aren’t an exact guide to the future and need cautious interpretation, their record is stronger than it currently appears.

How can polls be biased?

When it comes to broad trends within the margin of error - usually accepted to be three percentage points - polls taken as a whole are fairly reliable. When it comes to detailed breakdowns of individual group behaviour, however, statistical noise creeps in.

Bear that in mind when you’re faced with predictions about swings between minority party voters, or leave vs remain voter behaviour – conclusions might be based on small subgroups that aren’t sufficiently representative.

Individual polls at national level are prone to the same effect: one outlier result might be the start of a trend, or it might be just that - an outlier. And be particularly cautious of “private polling” like the Ashcroft polls. They’re usually released to swing a particular narrative, and those who pay for them rarely reveal their methodology.

Should we believe the general election 2017 polls? Verdict

The only poll that counts is the ballot box. Only after the final count will we know for sure whether Jeremy Corbyn could become prime minister or whether Ukip is finished. But opinion polls are a fairly reliable guide to the broad shape of the general election. It would be a genuine upset if Corbyn won, based on Labour's current polling. 

Free trial CSS