ICM and the wisdom of crowds

ICM's poll in the Sunday Telegraph is intriguing. Rather than asking voting intention, it asked people to estimate what shares of the vote each party would get in an election tomorrow.

I'm not entirely sure what to make of it - largely because right now, all we have are some figures: Conservatives 31%, Labour 39%, Lib Dems 16%. Taken in isolation they are not particularly interesting, we already knew that people tend to expect Labour to do better than the Conservatives at the next election, and given that ICM tend to show high levels of support for the Lib Dems anyway we don't know if this is people predicting a higher level of support for the Lib Dems than they currently have... or a sample that would have shown that sort of level of support for them if asked in a conventional method.

What will be a lot more interesting is seeing how it works it over time and how it relates to conventional polling. ICM have been running it in parallel with their regular tracking polls, and have apparently found some consistent patterns of differences and tested it at the last election which it got correct. Martin Boon has a proper academic paper out on the method next month, so I'll be reserving judgement on what it can tell us until then.

Note that ICM are NOT abandoning their normal method. Their regular Guardian polls will continue in just the same way as usual. This is just complementary.