Last polls!

All the final polls are through, including from some companies we haven't heard from so much, which has actually made the spread appear even worse!

(links to larger version on twitter)

• Top Labour performance comes from Qriously, who give them a lead of +2% (last time any pollster showed a Labour lead was April 2016). Now, they use a very different sampling technique to other companies (basically via smartphone ads), which at least gives them different sampling biases to everyone else (e.g. they called Brexit the right way, but overshot by several percent)...

• Top Conservative performance comes from BMG, who give them a lead of 13%, which puts us in Labour wipeout territory.

In terms of who we might consider the most reputable pollsters (long track record, lots of polls through the campaign etc) -

• YouGov, who have favoured Labour throughout, now show a 7% Conservative lead, i.e. significantly larger than the last few we've seen. Closer inspection shows they've made a tweak to the methodology (reallocating 'Don't Knows' on the basis of past vote) which accounts for some of this shift.

• ComRes, who have favoured the Conservatives throughout, now show a 10% lead, a little lower than they have been (actually, the lowest lead they've shown all campaign). This is probably just normal sample variation, but we can't know for sure.

---

All in all, a cautious bet might therefore be placed on Conservative lead in the high single figures. But let's take a closer look. The second graphic here (in a fetching shade of red) is what you get if you assume 'all the pollsters have made sensible choices, and so all are equally likely to be right'. I'm being a bit technical at this point, but basically, dark bits show you where the most probable outcome is. The bits along each side show you a more traditional plot of probability, where the peaks show what the most likely vote share for each of the two major parties is (just like in the first graphic), but looking at both together is more informative: if one share is higher, the other is much more likely to be lower as the two shares are not independent of each other.

What it's really telling you is that the most likely Conservative lead is 7% but that you shouldn't really be surprised to see anything between a Conservative lead of 2% and 12%.

(if it's outside that extremely generous range, then we can probably confidently say something went badly wrong again for multiple pollsters)

---

So what does it all mean for seats? Naturally this is difficult to answer with certainty, but taking the most probably vote shares for each party and doing a uniform swing calculation produces a Conservative majority of about 30 seats (in this scenario, incidentally, the Lib Dems get 5-6 seats). A hung parliament is however within the realm of possibility, although if we do end up there I'd expect the Tories not to be too short to form a government. On the other side of the coin, a majority of 60 would also be perfectly plausible.

As I've said before though, uniform swing is probably a dodgy method to use in this election. For what it's worth, the received wisdom is that Labour's vote share might be piling up disproportionately in safe seats they already hold. In that case the actual seat tally might favour the Conservatives a lot more than I've just suggested. Received wisdom does, however, not have the best track record in recent years...