When do markets work?

Barry Ritholtz is celebrating Michael Jackson's acquittal more than most, because it corroborates his view that prediction markets can be wrong; Intrade bettors were backing a guilty verdict on the charge of supplying alcohol to minors.
However, Barry is swimming against a strong tide here. There's lots of evidence (gathered superbly by Chris Masse) that information markets can be great predictors - for example, this presentation (powerpoint). They embody the wisdom of crowds.
This raises a question: can we get a list of conditions under which betting markets are or are not accurate predictors?
I'll kick off with two possibilities.
First, markets work best at aggregating dispersed information. This condition failed in the Jackson case because the information came from just one source - the courtroom.  But the condition held in the UK and US elections, where betting markets were more successful.
Second, markets can fail to account for individual quirks or eccentricities. They therefore work best at predicting the opinions of crowds (for example general election results where such quirks cancel out) than small groups. As Chris Masse put it in a comment on Barry's post: "Events whose outcomes are determined in a conclave by sequestrated jury members are among the hardest to divine."
Can we add to this list? And if betting markets are as good at predicting things as their advocates suggest, why do share prices so often seem systematically wrong*?  Why are gamblers smart but investors stupid? Is the short sales constraint a sufficient explanation?
* For example, in the last 12 months Shell has out-performed Royal Dutch by 5.8 percentage points. They are the same company.

Comments

[Barry Ritholtz]

Not so much celebrating, as pointing out an instance where the Wisdom of Crowds, well isn't.

I disagree with your assesment of the MJ case as being all in that courtroom. An Esquire article (from 1993?) laid out why MJ was subject to these ongoing extortion schemes: A kid's dentist father, divorced from his mom and jealous of their relationship with Jackson, performed dental surgery on his own son (unusual to say the least), with a general anethesia (unusual for that particular surgery) and used contraindicsted drugs with a known tendency to create heightened states of suggestibility.

That article colored my belief that Jackson (freakish tho he may be) was again being extorted by grifters with a history of litigiousness.

And what of Purcell leaving Morgan Stanley? Tons of external factors (i.e., outside a single courtroom) --

A 5% expectation that he would resing by June 30th -- that was even worse than the MJ contract!

http://bigpicture.typepad.com/comments/2005/06/purcell_predict.html

[Katie]

Surowiecki mentions exactly those two conditions, diversity and completeness of information and diversity of participants' expertise, but he also mentions that the data submitted by particpants must be suitable for aggregation, rather than one person "winning" outright. Doesn't apply here necessarily (especially as we've seen that the first two are not satisified) , I'm just saying that's his third requirement.

[Chris. F. Masse]

Hello Chris Dillow,

I like this spot-on comment. I have indexed it in the link below.

And thanks for the deserved link to professor Leighton Vaughan-Williams' presentation.

By the way, he has posted a HTML format of his latest paper:
http://www.ntu.ac.uk/nbs/spec/betting_research_unit/16580gp.html

Thanks,

Chris. F. Masse

[Chris. F. Masse]

Hello there,

I have come up with a long piece about Barry Ritholtz's 2004 political predictions versus the 2004 US presidential futures markets' predictions. (I made a link to this particular blog entry, by the way.)

How the 2004 U.S. political prediction markets outsmarted a Wall Street pundit, an esteemed economics professor and the commentariat as a whole!

Barry will respond to me on his blog on Monday evening.

Best regards,

Chris. F. Masse