The race between Barack Obama and John McCain was supposed to be about winning the middle. Both candidates embraced the theme of moving from partisan gridlock to seeking bipartisan consensus. Obama's speeches evoked a country that was "not blue states and red states, but more United States." McCain focused on his record of working on both sides of Senate aisle. In contrast to the 2004 election, in which both parties sought to motivate their bases, this campaign was set on the battlefield of undecided voters.

The election results, however, record the exact opposite happening. Most undecided voters swung to McCain, but he was buried by Obama's ability to spur enormous turnout among the Democratic base while McCain, in turn, was unable to bring out the Republican base.

The swing among undecided voters continued a trend that had occurred during the Democratic primaries. I had posited a "Social Effect"--a variation of what some call the "Bradley Effect"--where voters who say they are undecided are in fact highly unlikely to vote for the leading, better funded or more widely supported candidate.

Obama's strength on Election Day, winning 53-46 in the popular vote, has led almost all polling analysts to conclude that there was nothing resembling a "Bradley Effect" this year, and that the undecideds basically split.

I think those conclusions are flawed. Consider four polls and their last estimate of the white vote:

ABC News: Obama: 45; McCain: 52
CBS News: Obama: 46; McCain: 49
FOX News: Obama: 42; McCain: 50
Gallup: Obama: 44; McCain: 51

On Election Day, the

The pollsters were more accurate in their estimates of how black voters, and all non-white voters would split, but even there, two of the four pollsters slightly overestimated Obama's strength and underestimated McCain's support in the black community--though the sample sizes are too small to draw conclusions.

How could the polls underestimate McCain's strength among whites, and under- or accurately estimate his strength among non-whites--yet still overestimate his overall strength?

The polls I've looked at missed just how strong the African-American turnout would be and overestimated the percentage of whites that would be voting. The likely voter models that most pollsters had projected about 77 percent of voters would be white, while the networks' exit polls found that only 74 percent were white. Similarly, ABC, Gallup, NBC and others pegged 11 percent of the electorate would be black , while Fox estimated they would constitute 12 percent of the electorate. The exit polls found that fully 13 percent of the voters were African-American.

A one percent shift in black voter share may not seem much, but note that because McCain was leading among whites while getting less than 5 percent among black s, every one point shift from white to black increases Obama's lead by more than 1 point.

As a result, the election that was supposed to focus on the middle turned out to depend on the wings. Obama lost those undecideds, but won the election by ensuring that his most enthusiastic supporters--young people and African Americans--turned out. Meanwhile McCain may have won the center, but he didn't do enough to bring the Republican base to the polls. In fact, they stayed home.