Let’s also assume that, generally speaking, Obama’s approval rating rises and falls relatively consistently across all 50 states. This allows us to use the president’s national net approval and his 2012 margin, winning or losing, in each state to calculate what we call the president’s “expected net approval” for a state-level poll.

Our baseline is Obama’s national four-point victory in 2012, which is about what his net approval rating was on Election Day. There is no better “poll of approval” for a sitting president than reelection results. So if Obama was +4 in November 2012 and now he’s -5, he has suffered a minus 9-point swing overall. We can then apply this expected swing to individual states. For instance, Obama won Ohio by three points in 2012. Applying the -9 swing from then to now, his “expected” approval in Ohio would be -6.

Based on our technique comparing the actual state net approval to the “expected” approval, a large majority of surveys in 2015 find the president with higher levels of disapproval than we might expect if there were actually a uniform swing across the country. Of the 52 polls in our data set, 32 (62 percent) found Obama’s approval to be lower than the “expected” approval would anticipate. Only 17 (33 percent) rated the president higher than the “expected,” while three exactly matched expectations.