Tom Goodenough

Listen: YouGov’s Joe Twyman defends shock election poll

Listen: YouGov's Joe Twyman defends shock election poll
Text settings

Can we trust the pollsters? Bruised by Brexit and caught out by Trump, the psephologists claim they’ve finally learnt their lesson. If so, that’s a big problem for the Tories: today’s YouGov poll predicts that the party is on course for an electoral upset which could see them lose their majority. YouGov have been busy defending the numbers behind the headline this morning, and the company's Joe Twyman has been doing just that on the Spectator's Coffee House Shots podcast. So, given YouGov failed to get it right in 2015, why should we believe them this time? And will the pollsters end up tweaking their assumptions? Here's Joe Twyman:

'We are confident in the power of our data and the smart of our data scientists, that we have something that works.'

Twyman said that today's poll - which shows the Tories could be down 20 seats come June 9th - was a 'snapshot' of YouGov’s current modelling based on assumptions about turnout. And here's his response when asked on the podcast by Fraser Nelson about their 'horrendously wrong predictions at the last election':

JT: You are right that last time, in 2015, we weren’t where we wanted to be and in light of that we’ve done various things. The first thing is we’ve identified the type of people we need to contact in greater numbers to be more representative of the population as a whole. And we’ve spent hundreds of thousands of pounds going out to recruit those people to make sure they are in our surveys.

The error last time around was not enormous. The average error was 3.3 per cent. The other thing I would say is that last time around, we ran this same model and we found that Scotland would go overwhelmingly to the SNP,  the Lib Dems would be wiped out and the Conservatives would do much better than the existing poling was suggesting.

At the EU referendum, it (the same model of polling) never showed Remain was ahead. With all polling you make adjustments based on historical precedents, based on the data available and that’s what we have (done) here.

You can listen to the full Coffee House Shots podcast using the player above, or click here for details on how to subscribe.