The first time I looked at TradeSports.com -- an Irish sports gambling site that operates through an international exchange of contingent contracts -- they were putting the chances of President Bush's re-election at 60% (i.e., a $100 BUSH WINS contract was priced at $60). The clients of TradeSports.com spend real money on real contracts, taking real risks in pursuit of real reward. The virtual prediction market site, NewsFutures.com also offers contingent contracts on President Bush's re-election chances, but both purchases and pay-offs are done with play money. Their "market" gave Bush only a 49% chance of re-election. This was in the spring of 2004. One obvious empirical test would be to compare how close the different sites come to predicting the actual election results -- will the local knowledge of the local American website outperform the profit-driven predictions of the international gamblers?

As I write this, in the summer of 2004, I can't yet know. What I do know is that an 11-point discrepancy in contract prices couldn't last if both markets were based on the risk of real money. If both markets really paid a hundred dollars for their $100 contracts -- and if they were both active enough to allow a large number of such contracts to be traded -- then arbitrageurs could guarantee a certain return in November. By buying the cheaper contract in each market -- the $40 BUSH LOSES in one market and the $49 BUSH WINS contract in the other -- they could guarantee a $100 pay-off for an $89 expenditure (ignoring transaction costs). Note that the arbitrageur doesn't need to have any opinion of Bush's re-election chances: he is, after all, betting both ways. All that is necessary for the arbitrage profit opportunity is a disagreement between the two markets: a discrepancy in their prices.