The tables for ICM’s poll in Crewe and Nantwich are now available on their website, and they reveal some brand new changes to ICM’s methodology which I’ll look at in the next post. First though, let’s dig through the entrails of the Crewe and Nantwich findings.

The narrowness of the 4% was indeed largely down to ICM’s normal spiral of silence adjustment. Taking just those people who actually gave voting intentions to ICM, the Conservatives had a solid 12 point lead. The narrowing of the lead was because a large proportion of respondents who told ICM they voted Labour in 2005 said they didn’t know how they would vote in the by-election: 61 out of 295 apparently – 21% of last time’s Labour voters, compared to 6% of last time’s Conservatives and 23% of the small number of voted Lib Dem last time in Crewe and Nantwich.

This means the vast majority of the don’t knows up there in Crewe are former Labour voters and ICM are making the assumption that those people will disproportionately end up voting Labour. ICM do this by reallocating 50% of don’t knows to the party they voted for last time, based on past research showing this is how people tend to behave at general elections. It is only an assumption of course, and people may behave differently at by-elections. If those former Labour voters actually stay at home or switch to the Conservatives the Tory lead would be much larger. If more than 50% of them end up voting Labour the Tory lead would be smaller than ICM’s poll suggests.

Another intriguing finding in the by-election poll was that the Conservative lead was much higher when people in Crewe and Nantwich were asked how they would vote in a general election – a 16 point lead in fact. On the face of this it is counterintuitive as we are used to bigger swings in by-elections than in general elections, not vice-versa. My best guess to explain this before seeing the tables was that ICM must have used the candidates names in the by-election question resulting in a “Dunwoody effect”. This was wrong – ICM didn’t use candidates names so this can’t be the reason. The actual reason seems to be that more people gave voting intentions for a general election tomorrow than for the by-election.

Comparing people’s answers in the by-election question and general election question, very few people actually changed their answers. Of 478 responses to how they would vote in the by-election, only 13 said they would vote differently if it was a general election. The difference seems to be almost entirely people who didn’t give a voting intention for a by-election, which does rather suggest that those don’t knows aren’t likely to break in Labour’s favour….


Comments are closed.