YouGov's sample is dodgy
Interesting poll in the Sunday Times:
Taking off my Blears partisan hat for a minute and taking a strategic view of the interests of the moderate wing of the Party in total, Alan Johnson moving into the predicted winner position is welcome news as he's the candidate closest to the one I'm supporting ideologically.
The 3 most moderate candidates - Johnson, Blears and Benn - have 57% support amongst members between them, and the 3 that have talked most left - Harman, Cruddas and Hain - 42%.
And only 18% of members would have voted for McDonnell if he'd got on the ballot.
I'm mildly bemused by the poll finding movement towards AJ as in both MP and CLP nominations he underperformed relative to expectations, and in canvassing I've done he's been losing support to Hazel.
I'd also like to see the figures for undecideds - all the canvassing experience I've heard is that the vast majority of members remain undecided.
Before everyone reads too much into the poll, I'd challenge the accuracy of their sample in the TU section. My hunch is that it is biased towards activists - who are likely to be more leftwing - because they are more likely to be internet users and to have signed up to the YouGov panel. Their sample would also be intrinsically biased towards white collar trade union members who have internet access at work and may not be politically representative of their blue collar colleagues.
My evidence for this - they say:
"YouGov questioned 1,676 people able to vote in Labour's leadership election. These comprised 1,097 members of the Labour Party and 833 members of trade unions who pay the political levy. 254 respondents belonged to both groups, and will be able to vote in both sections."
They've got about the right number of affiliated trade union members in their Party membership sample - 23% - I've checked a couple of CLPs and this is about bang on, but that implies that there are only about 40,000 Party members nationally who are also TU levy payers. This equals about 1% of the total number of TU levy payers in the country - 4.4m - maybe 2% of those in Labour-affiliated unions - yet YouGov's sample has an absurd 30.5% of its trade unionists also being Party members! If only this was the case - Labour would have 800,000 members, and 640,000 of them would be trade union members , if it was.
According to the 2005 British Election Study, only 45.7 % of current trade union members voted Labour in the 2005 General Election, 22.1 % Liberal Democrat and 19 % Conservative - so why would the union section of the electoral college produce results to the left of Labour Party Members and MPs?