YouGov/Sunday Times - CON 44, LAB 31, LD 11, UKIP 6

YouGov's weekly poll for the Sunday Times has topline figures of CON 44%, LAB 31%, LDEM 11%, UKIP 6%, GRN 2%, down to a thirteen point Conservative lead compared to sixteen points in the week (and twenty-point-plus leads when the election was first called). It suggests that the mid-week YouGov/Times poll was picking up the start of a trend, rather than just a blip (though given that ORB hadn't done a recent poll and Panelbase changed methods, we haven't really seen confirmation from other companies yet).

If the Tory lead really has fallen, the next question is why. As ever, it's impossible to know for sure (though I will make my usual warning about assuming causality from petty campaign events - a few events like budgets, leaders speeches at conferences and so on can have an measurable impact on national polls. Calling someone a mugwump does not).

One thing that is interesting is Labour don't knows. Looking at the entrails of the YouGov polls, it looks as if some 2015 Labour voters who were saying "don't know" a week ago are now saying Labour. When YouGov were showing those twenty-point leads around 20-25% of people who voted Labour in 2015 were saying they didn't know what they would do at the election, in the last couple of polls that has dropped to 11%. The other thing worth considering is whether those twenty-point leads were real at all, or just the result of temporary enthusiasm? Were Tories all cock-a-hoop and itching to take part in polls last week, Labour voters too despondent to bother? Political weighting of samples should go a long way to counter-act any such biases, but it may not do so entirely.