From my experience, the generic congressional ballot polls are often worthless. 2010 was the first time since possibly 1994 with the exception of maybe 2002 that the GOP had a lead in that polling. In several of those intervening elections, the GOP held the House (1996, 1998, 2000, 2004). Only 2006, in which the Dems had a wider lead, did control change.
State by state generic congressional ballot has a limited track record. I don't even recall that being measured before.
-2% nationwide is not enough to flip the house. Frankly, it's not even enough to prevent a small net gain of 2 or 3 seats with the improbable scenario of everything else falling completely in line (candidates, fundraising, issues, scandals).
You seem to underestimate just how stacked the deck is naturally (this is the result of demographics and the current set number of seats), how stacked it is from 1991, 2001 and from 2011 redistricting cycles, and the impact of the local campaigns in each district.
It is huge. If a slight advantage in the generic ballot was enough for a huge Republican gain in the House in 2010, then that has been reversed. To be sure, it is always possible that a district can have a PVI of R+5 and have a Democratic Representative (Minnesota 7) and that a district can have a PVI of D+3 and elect a Republican. Nobody questions that an unusually adept politician can survive a political tide.
Tide elections sweep out weak incumbents and fill open seats. Reverse-tides also happen. Some of the Republican gains of 2010 have resulted in the election of new Representatives with promising careers. Some will not be so distinguished. In wave years, some unusually-strong incumbents get defeated and come back strong the next time. Some may seem very decisive and well-connected to excellent fund-raising machines except that they badly match the political norms of their districts.
No incumbent can escape his voting record. It is clear that someone running in an ultra-safe Democratic district (NY-15 or NY-16, both of which are D+41) faces vastly-different expectations of a Representative than a Representative from a district best described as R+29 (AL-06 or TX-13). That's not to say that an out-and-out Marxist can consistently win a district best described as D+40 or that an out-and-out fascist can consistently win a district best described as R+25. That said, districts best described as R+3 and D+3 are likely closer to each other in their world-views as they are, respectively, to districts better described as R+25 or D+25. A Representative best described as fitting a D+25 or R+25 district has to be unusually effective to win re-elected in a district best described as near-even.
Americans elected a bunch of Republican Representatives, most of whom would better fit a district like AL-06 or TX-13 than something close to even either way, in 2010. That was enough to shift the House like this:
http://en.wikipedia.org/wiki/Cook_PVIWe know how the new Republican Representatives have voted so far, and it is very far to the Right. Can someone far to the Right convince moderate voters that the Hard Right which offers economic ideology more characteristic of the Gilded Age than of the 1990s?
Recent polls suggest how out-of-touch Congress is. An approval rating near 10% itself suggests failure on the whole... but ordinarily Congressional Representatives are capable of fitting the cultural and ideological norms of their districts enough to have personal approval ratings around 50%, which is normally good enough for re-election almost every time. I'm going to suspect that approval ratings for those who represent "safe" districts -- districts consisting largely of anti-Establishment minorities or those whose economic interests are largely ranching and oil extraction -- are still very high.
When the approval rating at large for "Your Representative" is 41%, then the average Representative is extremely vulnerable to defeat in the next election. Incumbency is an advantage, but not enough to save a misfit in his district. I have tried to adapt an old study of Nate Silver (incumbent Governors and Senators with approval ratings around 44% at the start of the campaign season win about 50% of the time and that the drop-off is steep below 43% and that those with approval ratings around even 47% win almost 100% of the time) applies to both the Presidency... and probably also to Congress.
Incumbents can hold lots of town meetings and make plenty of appearances in county fairs... but not enough to hide an extreme voting record by local standards. What fits an area whose economic activity is largely ranching and oil extraction is very different from an area whose economic activity is largely dairy production and Rust Belt manufacturing. Incumbents can choose to run on their records and win -- or run from them and lose.
2010 really was a freak. It was a one-time reversal of trends almost certain to resume in 2012. If anything hold of the Democrats on the Senate is far shakier than normal only because the Democrats won about every Senate seat up for grabs in 2006 and have more opportunities for losses than gains... but every House seat will be up for re-election.
Who is most likely to lose a bid for re-election?
1. Someone involved in a well-publicized scandal. Representative William "Cold Cash" Jefferson was the only Democratic member of the House to be defeated in the 2008 Democratic wave. Mercifully such pols are rare. Corruption takes time to develop, so I expect few of the new Reps of the 112th Congress to fit this pattern.
2. Representatives simply not up to the job. These can be the ones who badly under-perform as freshmen or the ones who fade due to senility or crippling ill health. Democrats were obviously more vulnerable to that in 2010. Now it is the turn for Republicans.
3. Representatives who misuse the perquisites of office. Watch the Congressional junkets. If your district is a Rust Belt district in which unemployment is high and times are hard and your Congressional communications include photos of your Representative and spouse is of fine times on the Champs-Élysées, then expect that to be your last term in office.
4. Extremists, meaning those out of touch with the sensibilities of their voters. 100% voting which shows neither conscience, flexibility, nor independence, with one Party might not be good for someone in a district split roughly 50-50 D-R.
America is not on a rightward trend except as shown in the 2010 election. "Occupy Wall Street" shows that Corporate America is not winning the hearts and minds of America by treating people badly and insisting that the rest of humanity exists only to enrich and pamper elites. The Tea Party is now past-peak. Some high-profile elected Republicans are making fools of themselves. The Religious Right, which consists of people largely in middle age, is not winning over youth. America is becoming less white and Anglo, and Republicans are doing badly even among successful members of minority groups. The only discernible group clearly trending Republican over the long term is under-educated, impoverished white people in Appalachia and the Deep South, probably corresponding to the Religious Right. (Advice to liberals: spare the "trailer-trash" and "redneck" references in discourse in politically-charged websites. Such people as the under-educated, impoverished white people especially in the Appalachians, Ozarks, and Deep South are getting hurt as badly as anybody else by economic policies that exist largely to enrich the well-connected).