Tom Davenport has picked up on the question of prediction markets. He is looking for information on use and knowing Tom will do a good job of putting it together so if anyone can help, do! I have mixed thoughts on the subject. Part of that is cynicism about anything with that much hype, especially when the hype doesn't seem to match up with conversations I have with people in the claimed successful user companies. One of the problems for me is that in a market everyone is aware of other people's bets and the market therefore will tend to equilibrium. Prediction Markets are not examples of The Wisdom of Crowds whatever Surowiecki says. In fact he seems to contradict himself
His criteria for a crowd to be wise can be summarised as follows:
- You need diversity of opinion, each actor needs a private opinion,
- Their opinions should not be determined by others, i.e. they should not be following the herd
- Participants should be able to draw on local knowledge
- There needs to be a mechanism for aggregating the individual decisions to represent the collective.
Now if we look at these we can see that a market satisfies 3&4, to a degree 1 but fails on 2. In a market you are aware of and respond to market movements, so ipso facto you are not making a decision in isolation. So I can't see why he uses prediction markets in his book.
Of course this does not mean that markets have no utility. However there seems to be a bit of a Hawthorn effect in operation - the case studies seem to be of start ups but with no continuation. I could well be wrong here, but that is the impression I am getting. Markets are there as an way of handling trade, indirectly they may predict but it is not their primary purpose. However I think there are better ways of using crowd wisdom. Distributing part problems to lists of players, engaging social networks on the basis of engagement. Not everything has to be reduced to money surely?
Comments (4)
Interesting that the McKinsey article referenced in Tom's post gives Google examples where they find people underestimate the possibility of extreme events. I think this relates to the "Gaussianitis" post from the CE guest blogger Pierpaolo Andriani. Fat-tailed distributions. Intuitively they are quite challenging but they seem to be pervasive.
Posted by Wayne Zandbergen | April 18, 2008 2:25 PM
Posted on April 18, 2008 14:25
Dave - The question is less: "Are prediction markets free from bias?" and more: "Do they produce more reliable predictions than other approaches?"
You can knock them down on a conceptual level but do they work (better) in practice? I think the jury's out - there are lots of reasons why they haven't taken off.
And prediction markets are more a form or betting / speculation than handling trade - but the same can be said of the derivatives markets.
Posted by Matt Moore | April 19, 2008 2:44 PM
Posted on April 19, 2008 14:44
I think you can argue that if something is theoretically wrong then the chance of it sustaining itself is weak. From what I have seen so far Prediction Markets work until people realise how they work, then they game the system. Markets produce deviant basis around the measurement vector (trading options on coffee not coffee) as they mature. Its one of the downsides.
SO I am interested to see what Tom picks up
Posted by Dave Snowden![[TypeKey Profile Page]](http://www.cognitive-edge.com/blogs/dave/nav-commenters.gif)
|
April 19, 2008 5:16 PM
Posted on April 19, 2008 17:16
To quote a colleague -
"As an example of mis-pricing, in fall/winter (before Iowa) you could buy a 'not McCain/Romney/Guliani' presidential prediction contract for, say, $0.02. But the day they announced they were separating Huckabee out of the also-rans, his contract began trading at, say, $0.10 while the 'also-rans' sank only to $0.01 (or something like this). So Huckabee + others was worth $0.02, but Huckabee alone + others alone is worth $0.11, which makes no sense and gives some evidence of the limits of such tools."
Wayne
Posted by Wayne Zandbergen | April 20, 2008 7:10 PM
Posted on April 20, 2008 19:10