The polling mistakes in 2017 were less serious than in 2015 because they resulted from clearly underestimating youth turnout, rather than because pollsters had found it impossible to build a representative sample of voters. In short, their data was right but the tweaks they made to it were wrong. In 2015 both had been wrong. ‘The problem in 2015 was sample,’ Wells explained. ‘The problem in 2017 was people trying to solve sample through turnout.’
