I find these fascinating. Note that this link takes you to the first of four pages. These are for House races.
It seems that of those who were polled, 46 percent recalled voting for Obama in 2008 and 45 percent recalled voting for McCain, which is much narrower than the actual margins. This could be (a) selective recall--because people now are less fond of Obama, they are less willing to tell a pollster they voted for him; (b) greater reluctance of Republicans to talk to pollsters [UPDATE--a commenter points out that this cannot possibly be the explanation--it goes the wrong way]; or (c) lower Democratic turnout. I vote (c).
The result that I find most interesting is a breakdown of favorable and unfavorable ratings of parties. Not surprisingly, among people who have a favorable rating of Democrats, the Democrats did really well. People who view the Republican Party favorably voted Republican--again, nothing to be surprised at.
But each party has a majority "unfavorable" rating, and there the results differed. People with an unfavorable view of Democrats went 88-10 for Republicans, but people with an unfavorable view of Republicans only went 76-22 Democrat.
That was the difference in the election. The "unfavorables" on net broke Republican. The way I would spin it is that there is a large block of voters who are negative on both parties, the Republicans captured a larger share of that block, and that block swung the election.
It strikes me that this block that takes an unfavorable view of both parties is potentially highly volatile. It suggests to me that a cautious moderate who does not tick people off might do well in a national election--but how could such a candidate survive in the primaries?
Note: I am not saying that I long for a cautious moderate. I long for competitive government, which is not embodied in any candidate.