Do Prediction Markets Work?

I used to think so, but most real world prediction markets do not seem liquid enough to provide any useful predictions. I love Felix Salmon's take on Intrade's market for McCain's vice-presidential nominee:

Yesterday's close on the Palin contract was 4; today's open was 6. By 8:23am Eastern time she was at 90; by 9:13am, less than an hour later, she was at 12. An hour after that, at 10:14am, she was at 95...

There's simply no real-world justification for the degree of volatility we saw in the Palin contract this morning; this is not a consolidation of "the wisdom of those who traded in the past" but rather a crazed market resembling nothing so much as a bunch of six-year-olds chasing a soccer ball.

See also Stephen Gordon's response to Andrew Coyne. Coyne is Canada's smartest and best columnist, but Gordon wins this one on points.