Retesting the wisdom of crowds

A couple of years ago, I undertook a quick experiment, focused around the Iowa caucus of 2008. Using data from the Washington Post’s The Fix blog, I compared the best estimates produced by of political experts, opinion poll information and an average of the guesses of random punters on the blog’s comment thread. As it turned out, the “crowd” did rather well, beating most of the experts and one major news organisation’s polling efforts. Was this just a lucky break for the crowd? Or can they collective wisdom get it right again? In order to test this idea, I decided to embark on a quick experiment with the near to first year students, asking them all to predict the 2010 British Election result.

The method was really simple:

• Get all the students in the lecture theatre to predict the election result next year.
• Take the average prediction across the whole group to come up with a collective “best guess”.

The workings for the exercise are on this spreadsheet. First, you will notice that some of the guesses are deeply unlikely (100 per cent for the Liberal Democrats?). But this is not the point. They also act as an element of the averaging process. Second, you will see the collective average is:

For comparison’s sake, I have also added the YouGov figures from their Telegraph poll on the 27th November (further discussion of this poll here. I can’t actually find the “other” figure in any reporting, so have calculated it as the remainder of the other figures).

So what does our crowd suggest:

• The Conservatives will do rather less well than current polling suggests but should comfortably get the most votes;
• Labour will do better than current polling suggests, but still be in second place in vote share;
• The Lib Dems will achieve something akin to the upper-end of their current polling range;
• Others will likely go up on the 2005 figure of 10.3 per cent.
• While one has to be careful with seat predictors, Anthony Wells’s model at UK Polling Report suggests these numbers would lead to a House of Commons made up thus:
• Labour, 290 seats
• Conservatives, 270 seats
• Liberal Democrats, 58 seats
• Other, 14 seats.

The first two points are surely the headline from this exercise, as it seems the opposite of much current, accepted wisdom on the subject. And, if both the crowd and Wells are correct, these numbers could leave us with a full blown constitutional crisis on our hands.