The following map shows how far each state moved from its standard partisanship (in the 2000 and 2004 election) for the 2008 presidential election. In other words, in 2000, the country as a whole voted D+1 in the presidential race. In 2004, the country went R+2. In 2008, it went D+7. The average of the 2000+2004 race was R+1, so the country as a whole went D+8 from its average of the previous 2 races in the 2000s. The map shows in blue states that swung more than 8 points between that state's 2000 and 2004 average towards Obama in 2008. Those in red swung less than 8 points. States in gray swung exactly 8 points and represent the nation's average.
I believe the map indicates that if Obama does not perform as well as he did in 2008 (which appears likely), he will do proportionally worse in the darker blue states. IN, MT, ND and VT swung Obama more than 17 points more than their average for the 2000s. So did HI, but I believe the home state phenomena probably accounted for most of that. It also indicates that even if Obama's approvals fall to 22%, he is unlikely to do worse in OK, LA, AR, WV and TN. States with a candidate in the 2000s (AZ, IL, HI, TX, TN, MA) will probably have their results skewed somewhat.
Overall, I'd expect Obama to do proportionally much worse in the west than the rest of the country, provided he can't win the popular vote by at least 7 points. Given only this voting data, I'd also expect MT and ND to move back into the GOP blowout column and IN and NC to go back to the GOP column, even with a moderately less successful Obama 2012 performance.