ICM show Boris and Ken neck and neck

ICM have a new poll on the London election for the Guardian, their first of this election (I think it's their first since 2000, since I can't find any ICM polls from the last mayoral election) and only the second phone poll of the election by anyone.

Unlike the double point Johnson leads recorded in the recent YouGov polls, ICM find the race

almost neck-and-neck. The topline figures are JOHNSON 42%, LIVINGSTONE 41%, PADDICK 10%, Others 7%. The second preferences of elimated candidates split slightly in favour of Boris Johnson, meaning on the second round support would split JOHNSON 51%, LIVINGSTONE 49%. The poll was conducted between Mar 28th and April 1st.

The obvious question that lots of people have already asked in the comments to the last post is how to explain the big difference between this and the YouGov poll. There is a five point difference in the two companies scores for Boris Johnson, a four point difference in their scores for Livingstone - it seems unlikely to be just random chance, more likely it is something to do with methodology. After Ken Livingstone's criticism of the YouGov poll the Guardian point out that (unlike YouGov) the ICM poll did include 29% ethnic minorities, which is the estimated proportion of London adults from ethnic minorities (though I'm not sure if it's the correct proportion of the London electorate). The couple of percentage points difference in the ethnic make-up of the panel is unlikely to be all the difference though.

There is also the question of how ICM have dealt with turnout, and whether there is the equivalent of their normal 'spiral of silence' re-allocation of don't knows. We'll really have to wait for the ICM tables to turn up before we can see. Even then, while it may tell us the reason for the difference, it won't tell us who is right - assuming that ICM and YouGov both continue to survey up until polling day, only the final polls and the actual result will tell us that.

(Edit - actually I'm told ICM did do a poll for Mind the Gap in 2004. It showed Livingstone at 50%, Norris at 29% and Hughes at 18%, but was very early in the campaign so shouldn't really be compared to the result: in YouGov's first poll of the 2004 campaign they showed Livingstone 46% and Norris 31%.)

UPDATE: The detailed tables for the ICM London mayoral poll are now up here. Looking at the details of what ICM actually did (if you're not interested in the anal details of polling methodology you can skip this post), the figures do indeed appear to have been weighted by likelihood to vote, which as usual increases the reported level of Tory support.

The poll was weighted by ethnicity. This could be some of the difference - certainly ICM found a massive contrast between the voting intentions of white Londoners, amongst whom Boris Johnson led by 19 percent, and ethnic minority Londoners where Livingstone has a mountainous 47 percent lead. Even with a contrast like that though, it doesn't make quite as much difference as you'd think - if you rework the ICM figures to see what the result would have been if it had contained only 24%

respondents from ethnic minorities (as Peter Kellner suggested the YouGov poll did), then at a rough calculation it would make only 1 point difference to Ken and Boris's scores, increasing the Tory lead by 2.

There would also appear to have been some sort of topline adjustment to the figures, since the filtered, turnout weighted figures were Boris 43%, Ken 41%, Paddick 8%. ICM did something to them that increased the Lib Dems by 2 points and reduced Boris by 1 point. Presumably they followed their normal methodology and added half the don't knows to the party they voted for at the last election.

Comparing cross-breaks YouGov has Boris doing better amongst women, ICM has Ken doing better amongst women. The two companies agree about Boris's lead amongst older people, and both have Ken Livingstone doing better amongst more middle aged people - but there's a huge contrast amongst young people, YouGov have a big Boris lead, ICM a big Ken lead.

In terms of social class the polls show pretty much the same as each other - that is, the pattern of support amongst ABC1s is the same as amongst C2DEs. ICM give a more detailed split, separating out AB, C1, C2 and DEs to show that Boris leads at the top and the bottom of the social scale, with Ken Livingstone ahead marginally with C2s and decisively with C1s.

So, small parts of the difference are probably down to different proportions of ethnic minorities in the samples and ICM reallocating don't knows to the detriment of Boris Johnson. Most of the difference though is less easily explained, while it appears to be with women and with young people, there's no obvious reason for that, and given the low sizes of cross-breaks it could just be meaningless chance. Who knows? We can't tell until the results come out who is right.