The Labour Leader Race Poll

So the Labour Leadership race rolls on, five candidates are duking it out for the prized position of Leader of the Opposition. Today YouGov released a poll of 1,184 Labour Party members, and 1,102 Labour supporting trade unionists to see what their views were. The results were fascinating:

The results are quite surprising. The fact that David Miliband is ahead amongst party members is no no surprise, though his lead amongst trade unionists is. The media narrative of the two Miliband race appears to be true and so does the perception that David is the frontrunner, though his lead in the run-off is not insurmountable. However I have several qualms about this poll.

Firstly, there is the MPs and MEPs. YouGov had no way of telling how the MPs and MEPs would break in the final round so they simply assumed that supporters of Abbott, Burnham and Balls would break 50/50. Now this is probably the fairest way of doing it if you want an overall winner, but I can guarantee that supporters will not split so neatly. I would expect almost all of Diane Abbott's left-wing support to move to Ed, and I would expect most of Ed Balls Brownite support to drift to the less tribal, but more Brownite than Blairite EM. I would also expect Burnham's support to drift to EM more or less. Indeed this has happened with the other colleges in this poll, as we can see from the smaller lead attributed to David at the end.

Secondly, it is not clear to me that YouGov has a representative sample. Whenever one polls one attempts to get a representative sample. If the population is 40% working class we try to get a sample that is 40% working class because we know that class affects voting behaviour. So where are YouGov getting figures for the demographic make-up of the Labour Party and Trade Unions from? Did Labour and the TUs give YouGov the figures? I doubt it. Another related issue is the nature of YouGov's polling itself. YouGov uses a panel of registered people on its website, rather than a random sample. For normal polling this works pretty well, but for something like this YouGov's internet savvy, youthful membership is going to make a difference, as it is likely that they will be using social networking sites, and blogging. Nonetheless, this is not to say that the poll is not interesting – it is, but its predictive value is probably a bit thin.