The post-poll poll portal

Why were the polls so wrong?

Well actually the exit poll was surprisingly accurate, especially when it was updated following the first few declaraions. (But its methodology is unclear: how were the updatings done?)

But the pre-election polls were near-unanimously wrong. They all predicted a Lab-Con neck-and-neck, except Nigel Marriott below - he seems to have been the best. Eat your heart out, Nate Silver!

Here are some reasons why the pre-polls were so wrong:

1. They asked different questions (e.g. "How WILL you vote?", not "How DID you vote?")
2. Labour voters were less likely to vote. (Agreed: young and poor tend not to vote, especially in bad weather - which is why Labour knockers-up always used to pray for good weather:  Do they still?)
3. Voters changed their mind at the last moment
4. Tories are shy or ashamed at admitting their intention.

Post-polls might clarify some of the above e.g. by contrasting  "How WILL you vote?" with "How DID you vote?" and even "DID you vote?" for different subgroups.

Here are some useful links I have found:

• Nigel Marriott's excellent video on election polls. #BBC should learn from this and make him their television guru! Nigel's May 5th forecast is not perfect, but seems better than the others. It may even be better than the exit polls if you ignore his overestimation of LibDem and UKIP and underestimation of SNP - much of this can be ascribed to overestimating the inertia effect (by which incumbent parties are given higher %s than non-incumbents).

• This page describes an exit poll method from 2010. It may underlie the 2015 method used by John Curtice, the BBC guru.
• Nate Silver blames bad polling on weak British election forecasting http://politi.co/1PujUdv 
• How 'shy Tories' confounded the polls (Grauniad) http: //trib.al/InQVS9E 

Here are some statistical bits-and-pieces (not all independently verified):

Loss & Gain Matrix 2010-2015  (from Wikipedia 'Talk' page: 2015may8:1800BST

• Wikipedia infobox 2015may8:1800BST

• UKIP, LibDems & Greens collectively won one seat per 740k votes, Labour one seat per 40k votes and Tories one seat per 34k votes (Sources here and here. For nerds only, figures are 7.4 million votes/10 seats,  9.3 million votes/ 232 seats, and  Tory figure is one seat per 34k votes where 34k=40k*0.86 where 0.86=(35/48)/(30.7/36.6)

From the above Wikipedia infobox, the votes-per-seat ratios are as follows:

Popular vote / Seats won = Votes per seat

Conservative: 11,334,920 / 330 = 34k

Labour: 9,344,328 / 232 = 40k

SNP: 1,454,436 / 56 = 26k

LibDem: 2,415,888 / 8 = 302k

DUP: 184,260 / 8 = 23k