IEM and other markets work by allowing investors to buy futures in particular events, in this case the outcome of elections. So, for example, you can buy the right to be paid $100 if Barack Obama wins the democratic nomination. If this Obama Futures are selling for $40 and you buy, you're saying that you think Obama is better than a 40% bet to win the nomination. If there are more buyers than sellers at $40, the price goes up until an equilibrium is reached. In this way, the views of people with money to spend can be turned into a supposedly accurate prediction of future events. Unlike polls which provide little if any incentive to tell the truth, with futures markets people are rewarded for sharing their views about future outcomes.
However in Iowa and New Hampshire, the electronic markets have not worked very well. If anything they has followed the (often wildly incorrect) polls rather than deviating from them. At midnight the night before the Iowa Caucuses Clinton futures on IEM were trading at almost $60, Obama futures at under $30. After Obama surprisingly won the caucuses, his stock rose to over $50 while Clinton's dropped to $32. Then the day after Clinton's comeback in New Hampshire they switched again: Clinton is once again the favorite, Obama trails. (I've found the best way to read the markets is here.)
I'm not doubting the power of markets like IEM to aggregate opinion and predict future events. But so far this election cycle the markets have trailed performance rather than predicting it.