I. Gerrymandering
Conventional wisdom was that the Republicans had a "firewall" through gerrymandering.
It looks right now like the Dems will have around 52.4% of the national House popular vote (the Committee for the Study of the American Electorate says the Dems have 36.5 million, the Reps have 33.2 million, so do the math yourself). Assuming that the nine undecided races break in favor of the candidate who is winning, the Dems will have about 232 seats (or 53% of the seats). So the gap between the Dems vote share and their seat share will be at most around 0.6%- pretty unusual in a single-member system.
In the last couple of elections, did Rs do any better? Not much. In 2004 there was a 2% gap (Rs got 51.4% of national House vote, 53.3% of seats); in 2002, the first post-redistricting year, they got 52.4% of the vote and 52.6% of the seats. So redistricting doesn't seem to create a system that favors Republicans per se very much on the national level.
Having said that, seats/votes gaps are much narrower today that in the 1990s. Let's look at two elections after the 1990 redistricting:
1992 Reps gave 47.3% of the votes, only 40.5% of the seats- a 6.8% seats/votes gap.
1994 Reps got 52.8% of votes, 53% of seats- basically no gap.
In 1990s, Dem gerrymanders clearly favored Ds, giving Ds a big surplus before 1994, and limiting their 1994 losses.
In 2000s, results much more ambiguous, so there's no reason to believe that gerrymandering consistently favors Republicans.
At most, it may be the case that bipartisan gerrymanders, by limiting the number of swing districts, limited seat swings to either party, essentially creating a floor of 200 or so seats for each side.
II. Exit polls
After 2004, there was much wailing and gnashing of teeth about exit polls and their D bias. I compared the early evening exit polls (off wonkette.com) with actual results.
Surprisingly, there wasn't a clear pro-D pattern. In Senate races polled, Republicans did more than a point better than exit polls in Virginia (47% poll, almost 50% on Earth), Montana (46% poll, 48% Earth) and Arizona (50% poll, 53% Earth), worse in Maryland (46% poll, 44% Earth) and within a point of the exit poll showing in RI, Pa, Ohio, NJ, and Mo- and exactly at the exit poll showing in Tenn.
Looks like the kinks may be out of the exit poll system.
