As Yglesias explains:
[T]he Intrade betting market lacks the amount of liquidity and derivatives instruments that would be necessary to make it simple and worthwhile for an arbitrageur to spot this and take advantage of it. Austin Goolsbee saw it, tweeted about it, then I decided to write a blog post about it. If something like that happened on a really deep and liquid market, some trading algorithm would have long sense made the moves that eliminate the anomaly but it doesn't work on Intrade.That's correct, and it's a good reason to take Intrade and other such markets with a grain of salt.. However, I'd say overall it's not the main reason.
What the markets should do, if they work well, is give us the wisdom of crowds. This is one reason they might not do that; another is if they are deliberately manipulated by the campaigns (or even by supporters of the candidates). Does that happen? It should, if the cost is low and people believe that manipulation can help a campaign, both of which conditions surely apply to Intrade during presidential nomination contests.
The other question, however, is whether the crowds have any idea of what they're talking about. During GOP WH 2012, I think the answer to that was a resounding "not even close." That is: it seems fairly likely to me that Newt Gingrich having about a 30% of capturing the nomination was in fact what people thought back during the November Newt surge, but it was nuts! Intraders managed to keep Herman Cain below 10%, so that was a somewhat useful correction to some of the hype out there, but overall it seems to me that Intrade did very little to help make sense of the nomination contest while it was going on. What I suspect is that rather than sorting through the pundits and adding other information, what Intrade is doing is just reflecting what the best-known pundits are saying. And that's not really doing anyone any good.