by Atul Hatwal
Earlier today Anthony Wells posted a response at UK Polling Report to a post of mine, “Is Labour’s lead actually 3% not 10%?”
Anthony makes a series of reasonable points defending the accuracy of YouGov’s 10 point Labour lead. He cites the high level of precision of YouGov’s final poll in the London election and the greater robustness of polls today compared to twenty years ago. Fair enough. But this misunderstands the purpose of my piece.
Polls measure how people feel at a point in time. There are adjustments and various psephological tricks to ensure respondents reveal their true preference, but the emphasis is on measuring how voters are actually feeling when the poll is conducted.
In mid-term, with an election some way off, many people might feel thoroughly hacked off with the government and be convinced they want rid of them. But when faced with the real choice, when it is days before polling and they engage with the arguments, a significant minority will change their minds.
By definition, this cannot be picked up by polling several weeks out from an election because voters haven’t switched yet.
The point of my piece was not to contest the accuracy of current polls, but to try measure this effect: to attempt to quantity the proportion of respondents who are relatively convinced Labour backers in mid-term but will then switch when the campaign ramps up and the polling booth beckons.
In fact, it is only because modern polling is so much more effective that we can be confident that a shortfall between mid-term polls and actual election results reflects voters who changed their mind rather than those who simply lied to the pollster in the first place.
