I am in complete suspense over who will control the House. As far as I can tell, things are about as much on a knife edge as one could imagine – despite a Democratic lead in polls. Why?
Unlike the Presidential and Senate races, the outcome of House control depends critically on events in the coming weeks. Voter enthusiasm will almost certainly be the determining factor. I agree (!) with Bill Kristol that the Obama-Romney margin will be strongly predictive; 3% or greater will suggest a likely Democratic takeover of the House.
I have evaluated incumbency and redistricting advantages quantitatively. I now believe that these two phenomena contribute approximately equally to an overall advantage of R+2.5+/-1.0%. In other words, Democrats need to win the popular vote by 2.5+/-1.0% in order to achieve near-parity in House seats. I was wrong about thinking there was little tilt in the playing field. It’s substantial.
Today’s prediction uses the generic Congressional ballot median since the conventions, currently D+2.5+/-1.0% (18 polls). Between now and the election, it is likely to move a few points in one direction or the other. Undecideds are at 10.5+/-0.8%. This month, a very large shoe will drop.
It is critical to pay attention to the uncertainty here: +/-11 seats.
House prediction, October 6th:
Two-party popular vote share: D 51.3 +/- 1.5 %, R 48.7 +/- 1.5%.
Seats*: D 217.5 +/- 11 seats, R 217.5 +/- 11 seats.
Uncertainties are 1-sigma (68% of outcomes).
Technical notes on this calculation are here. I would rather you not focus on the party-control probability, which is basically 50-50.
In the coming weeks, a better prediction will be possible two ways: (1) the freshest available generic polls, and (2) using district-specific polls, as collected by RealClearPolitics or Pollster.com.
A few days ago I met Bill Foster, physicist and Democratic candidate for Congress. He said that the DCCC was implacable. If a candidate is on the edge but has a fighting chance, he/she gets DCCC financial support. Too far ahead or too far behind? No money. Exactly what I would recommend. Here they are, along with knife-edge Senate races. And here is Crossroads GPS, which does the same for Republicans.
*The post originally said D217, R218. However, the actual median prediction is 217.5 each. Therefore I have changed it to reflect that. Normally I don’t like to give half-seat predictions. In this case, it has the advantage of conveying the situation to people who do not notice the error bar, which is +/-11 seats – not horrible, but the race is so close.