What I would like to do at some point (which shouldn't take long; I just haven't gotten around to it yet) is to look at what the actual tipping point seat was in the 2006 election. Like I said, if you looked at the 2004 numbers and assumed a uniform swing, then you would have expected that the Dems would need a 7.7 national PV margin to take the House. I think that's an overestimate of what they ultimately needed, but I should look at the margin in the tipping point seat of 2006 to see how big of an overestimate it is.
OK, to follow up on this, I just figured out the tipping point seat for 2006. Assuming I have the numbers right, it was Iowa-3, Leonard Boswell’s seat. Funnily enough, not a Dem. pickup, but a Dem. hold of an endangered incumbent. He won by 5.1 points. Since the Dems won nationally by 8.0 points, that means they only “needed” to win by 2.9 points in order to win the House. This contrasts with the 7.7 point lead that you would have expected them to need if you just took the 2004 numbers and assumed a uniform swing.
So the uniform swing assumption overshot what they actually needed by 4.8 points. For 2018, the uniform swing assumption would predict that the Dems need a whopping 11.4 point national margin of victory in order to take back the House. But if there’s a similar overshoot as there was 12 years ago (because swing seats are more elastic than safe seats), then you get 6.6 points.
So, TLDR: I guess it’s not as bad for Dems as I thought. Because swing seats are swingier, there’s a good chance that (barely) winning a House majority "only" requires them to win the national popular vote by about 7 points. Of course, that’s still a bigger structural disadvantage than the one they had in the 2000s. But if in 2018 they win the popular vote nationally by the same margin they did in 2006, then there’s a decent chance that they’ll just barely eke out a House majority.
Last I checked, the general congressional ballot was Dem +6 so they would already be close to reaching their minimum under your model's assumptions. Considering they're almost certain to gain on the general ballot as the administration progresses and the early unpopularity of the Trump administration, I'd contend the House shifting into Democratic hands is less of an uphill battle than some seem to think it is. People are slow to adjust their expectations to the historically constant midterm incumbent losses and also assume that gerrymandering is more ironclad than it really is and that low midterm turnout for Dems under Obama is the new baseline. If anything, reduced Republican turnout is a serious threat under a tumultuous and divisive presidency like Trump's.
Republican fears are well placed.
