Where prediction markets really showed their value was in the real-time aggregation of information. They did not favor Romney all night long, but were constantly digesting and aggregating all the recent and relevant dispersed information. Just as voting began, entrance polling showed that Ron Paul was doing well among the anyone-but-Romney crowd. Accordingly, the markets moved away from Santorum toward Paul. Yet, while the vote totals and pundits were still trumpeting Paul, prediction markets reversed their course in less than two hours moved toward a tight Romney and Santorum race.
The most impressive stretch of time for the prediction markets was as the last few results trickled in at around midnight Eastern time, four hours after voting began. Santorum was consistently winning in the vote total to the very end, but the markets knew what was left to be counted. At 11:54 PM ET I tweeted: "Markets still favoring Romney in IA even though he is down by 113 votes with 96% reporting? Almost everything left is college town of Ames." While the ultimate outcome was extremely close, the prediction markets were able to incorporate all available information to provide an efficient data point during the vote counting.