0330 in the morning and sleep deprivation is setting in. But if this election is showing anything, it’s that the existing polling models will need substantial work after this election. Ironically the final result may end up quite close to the exit poll, through chance rather than science!
The real flaw seems to be the breakdown of the relationship between national vote-share trends and local constituency results. This relationship was always a bit strained but now appears to have no relevance to the results at all.
So I reckon in future GE polls will have to look a lot more closely at sampling results within constituency boundaries. It’s going to make polling a lot more complicated…
An amusing side effect of all this during this evening has been that for much of the BBC coverage, many highly skilled (and paid) pundits and pollsters have been forced to cover up they don’t have a clue what’s happening, either. Priceless!