And how relaible has such betting speculation been historically fugi? I have no idea, I'm just asking the question.
The Iowa Electronic Market, run by U of Iowa, has been extensively studied and has generally performed better than raw polls. Polls are an input to the market, in the sense that people trading in the market base their decisions on all the information available to them, which plainly is mostly the polling data plus whatever intuition they have about momentum. However--polls themselves are famously inaccurate sometimes, so take "better than polls" with a grain of salt, since the polls aren't always very good.
Sites like 538 attempt to integrate all the polling data in a different way, by using a formalized statistical model that the sites author attempts to back test. This takes the raw data and translates it into a prediction using all the available historic data the author was able to build into the model. In a sense, this is also an input to Intrade, as Intrade participants can view 538 predictions in making their trades--and probably do. There are probably a bunch of people who arbitrage the difference between the 538 prediction and Intrade's.
Whether a formal, professional, back-tested statistical model that integrates polls using formal methods is better, or whether an organic market is better at integrating polls, is a contentious topic that can probably be debated until economists give up on their profession and go home. I am certain there is no research out there stating that 538 is better than a prediction market or vise versa, but I believe that there is research demonstrating that both approaches are superior to raw, uninterpreted polling data.
However, again, polls themselves are not all that accurate, and so if all the polling is wrong--and that has happened before--then perhaps garbage in, garbage out.
As the famous saying goes the only really accurate poll will be conducted on election day.
Also note that even the most aggressive prediction, on 538, giving Obama a roughly 75% chance still has Romney winning one coin toss in four and with Intrade's odds he is winning one coin toss in three. Unless the markets were reporting Obama with over 90% certainty that is still a lot of leeway for inaccurate polls, voter shifts between now and election day, etc.