Question: Can large groups of speculators predict elections better than I can?

Short answer: Don't economists wish.

Long answer: No. A cursory examination of the current lines at tradesports.com reveals that speculators as a group currently believe that a) Bush has a 52.3-53.0% chance of winning the election (bet PRESIDENT.GWBUSH2004) but b) Kerry has a 64.2-67.2% chance of winning Florida or Ohio or both (bet BUSH.OH&FL). This state of affairs is an enormous, ludicrous arbitrage, and I wish I had enough money to take complete advantage of it. (As it is, I fully intend to make a hundred dollars or two.) The reason this is ludicrous is that if Kerry wins either Florida or Ohio, he will win the election, period; no really, I defy anyone to come up with a reasonable winning coalition for Bush that fails to include both Florida and Ohio. The reason it creates an arbitrage (a risk-free investment) is that by investing in equal numbers of shares of PRESIDENT.GWBUSH2004 (sell) and BUSH.OH&FL (buy) one can virtually guarantee a 20% return on investment and even have a 5% chance (if Kerry somehow wins with neither Ohio nor Florida) of a 40% return. By my analysis, there is a trivial risk (under 3%) of total loss (in the unlikely to impossible situation of Bush winning while failing to hold both Florida and Ohio). What I'm saying is that everybody knows that PRESIDENT.GWBUSH2004 and BUSH.OH&FL are essentially the same bet (and my numbers bear that common knowledge out), and the fact that they are selling at drastically different prices is a red flag that people are being supremely stupid; and when people are supremely stupid in a free market, there's nearly unlimited money to be made risk-free for anyone with capital.

I strongly suspect that PRESIDENT.GWBUSH2004 and BUSH.SOUTH (predicated on Bush capturing the entire Confederacy) also produces an arbitrage, although I haven't run the simulations to tell just how risk-free it is (the only risk is that Bush will lose a Confederate state but somehow still win the election). Also, BUSH.ARIZONA is underpriced, BUSH.NEVADA is underpriced, BUSH.NEWHAMPSHIRE is overpriced, BUSH.NEWMEXICO is overpriced, BUSH.NORTHCAROLINA is underpriced, BUSH.OREGON is overpriced, BUSH.PENNSYLVANIA is overpriced, BUSH.TENNESSEE is underpriced, BUSH.WASHINGTON is overpriced, and BUSH.ELECTORAL+300 (predicated on Bush getting 300 electoral votes) is overpriced, all by at least $15. Of these, OR, PA, and WA are the worst-priced.

I ran my standard simulations today with some new polls and some new information about where Nader is on the ballot. I don't remember the results, but Kerry has a 61 or 62% chance of winning the election, with expected electoral votes in the 280s and an expected share of the two-party popular vote around 50.4%.