If you've been reading Wonkblog, you know that most election models predict that Republicans will maintain control of the House, even as Obama retains the presidency. There are a lot of things that explain that. But one under-discussed factor is redistricting following the 2010 Census.
Every 10 years, that process winds up pitting members of Congress against each other and results in bizarre, geographically incoherent districts (the term "gerrymander" comes from Massachusetts governor and future Vice President Elbridge Gerry drawing a district that looked like a salamander). But because Republicans control more statehouses than Democrats, that process tends to favor the GOP.
Sundeep Iyer and Keesha Gaskins at the Brennan Center for Justice took a stab at quantifying that effect. They argue that the 2010 redistricting will result in a net gain of 11 seats for Republicans. The main reason is that 173 seats were redistricted by GOP legislatures, while only 44 were redistricted by Democratic legislatures; the rest were either set by independent commissions or courts. As you'd expect, the GOP legislatures redistricted in ways that produced more Republican-majority districts, the Democrats did the opposite and, interestingly, the commissions tended to favor Democrats as well:
It's possible the effect of those commissions was random, since it does seem to be milder than the effects of partisan legislatures; it could also be that they're correcting for past pro-Republican districting. In any case, the result was that where 230 seats disproportionately favored Republicans before redistricting, 241 did so afterward:
Given underlying party trends in districts, Republicans are favored to maintain their current majority, within a seat. Given that, it's remarkable that Democrats are favored to even make the modest gains predicted in Sam Wang's model (the one seat pickup projected by Eric McGhee, et al is more in line with Brennan's findings).