Why were the national polls wrong?

[pbrower2a]

It is hard top predict who will vote, especially in populations whose prior behavior in voting is ill-known. Models may have assumed that the electorate would be much like that of 2008 (more D-friendly than justified, but generally right) or that the electorate would look much like that of 2010 (much more R-friendly than justified).

It all reduces to what narrative one wishes to believe.