Glenn Greenwald compared the pre-election polls to actual results in the battleground states and discovered Obama did better than the polls predicted in every single state.

As Greenwald explains, this data disproves some popular theories of the right, left, and center:

Just as was true in 2006, various factions from across the political spectrum insisted this election that the ultimate vote totals for the Democrats would be substantially lower than what polls predicted. Many on the Right claimed that this would occur because pollsters and the media organizations which sponsor them are biased in favor of liberals and therefore manufacture anti-GOP polls. Some on the Left (including some here recently) claimed this would happen due to GOP control and manipulation over electronic voting systems, which enable GOP operatives to switch large numbers of Democratic votes to Republican votes. And many pundits and others predicted this would occur due to the "Bradley effect," whereby voters intent on voting for McCain would lie to pollsters and say they were voting for Obama (or, relatedly, that McCain voters would be more reluctant to speak to pollsters at all).

As I wrote before the election, the unprecedented number of pre-election polls combined with rigorous aggregation of that data by sites like FiveThirtyEight.com meant the final poll averages would be very close to actual results.

FiveThirtyEight.com hasn't done a state-by-state analysis yet, but the accuracy of their national projection is scary:

Right now, Barack Obama has 63.7 million popular votes to John McCain's 56.3 million, whereas third party candidates have roughly a collective 1.6 million. That works out to 52.4 percent of the vote for Obama and 46.3 percent for McCain ... conspicuously close to our pre-election estimates of 52.3 percent for Obama and 46.2 percent for McCain.