Nate Silver has been reading the tea leaves and the Democratic pollster/analyst has found even more gloomy news for Obama.
Here are 5 reasons the wave may be even bigger than pollsters are saying:
1. Downballot and cross-ballot effects. Republicans are poised to win somewhere from 22 to 28 of the 37 United States Senate races on the ballot. There are also 37 races for governor; the picture there is a bit murkier, but Republicans will almost certainly win a clear majority, and could conceivably win as many as about 30.
2. Unlikely voters voted - and they voted Republican! Almost all pollsters apply likely voter models of some kind, which estimate how likely a respondent is to vote based on their degree of interest in the election, their voting history, and in some cases, their knowledge of things like where their polling place is. On average, these models show Republican candidates performing about 6 points ahead of their standing among all registered voters in these surveys.
3. The incumbent rule, or something like it, makes a comeback. The incumbent rule - the notion that undecided voters tend to break against the incumbent - is something that I've spent a lot of time debunking. There isn't really any evidence that it's been true in recent elections (the period I've studied in detail covers 1998 through 2008). Undecided voters in these elections were about as likely to vote for incumbents as challengers.
4. The Scott Brown effect. Here is a little pet theory of mine. Say that you're a fairly conservative Republican in Massachusetts. Your senators have been John Kerry and Ted Kennedy for many, many years. Your representative to the House is a Democrat. Your governor is a Democrat. Your state always votes Democrat for President. You feel compelled to vote out of patriotic duty, and you usually do. But deep down, you're resigned to the fact that your vote won't really make any difference, and the candidates you want to win never will. And to be honest, you've got a little bit of pent-up frustration about this.
5. Likely voter models could be calibrated to the 2006 and 2008 elections, which were unusually good for Democrats. In addition to wrongly excluding some Republican "unlikely voters" (see Point No. 2), it's also conceivable that some likely voter models based on past voting histories are overrating the propensity of Democrats to vote. The reason could be that some of them are based on past voting history, and a common question is whether the voter had participated in the last two elections.
All I can say is, "Cowabunga, dudes!"