Would Gord really bet the farm on ComRes numbers?
Let’s get real about a 2009 Q1-Q2 election
Everywhere, it seems, there is election date speculation and the new possibility that is coming into the frame is February 2009. But how seriously should we take this?
For one of the problems we have at the moment is that the only polls that are getting highlighted are the ones at the extreme. So YouGov’s survey for the Sunday Times showing a small increase in the Tory share and a small decrease in Labour’s was barely mentioned by the paper. Ipsos-MORI’s C41-L-36-LD11 hardly featured – instead what used to be trusted media bodies like the Press Association are putting out stories under misleading headlines like “PM takes lead over Cameron in poll”
I come back to what I have argued repeatedly – the only numbers that matter are on voting intention and here there is only one pollster which could give Brown confidence about going early – ComRes. On Sunday this reported unchanged figures of C37-L36 but with the LDs dropping three to 14.
This is out of alignment with the other firms and polling watchers such as myself and Anthony Wells of UK Polling Report have been trying to find out why. First thing yesterday morning we sent a joint email to the head of the firm, Andrew Hawkins, asking for explanations about ComRes’s weightings which on the face of it seem to be subject to a lot of movement.
For it appeared to both of us that the main party shares in the latest poll were being sustained solely by big movement in the firm’s past vote weighting numbers in the space of a fortnight and how this was worked out was far from clear. Thus the data seemed to show that in the end November poll the ratio of 2005 Labour voters to Tory ones was 6:5 – less than a fortnight later that ratio had moved to 13:9.
Without the change in the past vote weighting ratio the latest ComRes poll would have been reporting a less favourable position for Labour and the early general election speculation would have been much reduced.
Maybe this was down to us misreading the spreadsheets and I was hoping to have reported the firm’s response this morning. Alas Hawkins replied that “I’ll be happy to come back to you with a full response in due course but I’m afraid the combination of Christmas and ‘flu means that it won’t be for a few days.”
It looks as though there might be three or even four more polls before Christmas. ICM’s numbers should be out tonight; there’s another Ipsos-MORI poll that might come today or tomorrow, YouGov could be doing their Telegraph survey and there might be another ComRes poll in the Indy.
Mike Smithson