If this story on HuffPost is true, this will be a huge triumph for the more theoretically informed approach to data.
The implication of the stories like formerly Obama counties in 2008 and 2012 flipping to Trump in 2016 is that it is conceivable that a significant number of voters who seemed “Democratic” voted for Trump. It is not necessary that their numbers be huge: the Midwest was, after all, narrowly won by Trump, but just large enough to turn the election. The naive belief inculcated by less than skeptical approach to political science theory is that all people simply vote party, but this is not likely to be the case, especially in 2016: the educated and the affluent probably voted party, but an electorally significant chunk of the working class apparently did not. Without a causal theory, you could not have foreseen it–but this is the problem that has been bugging me the entire campaign.
It is hard to believe that the higher ups in the Clinton campaign were completely oblivious of this: the so-called Rendell strategy was predicated on campaigning for the affluent and the educated, with the prospect of winning over formerly Republicans voters to the Democratic side, even at the risk of losing the low propensity working class voters. If you are giving up the working class voters, why bother running a GOTV operation targeting them in the first place? Obama, with his considerable appeal to the working class whites, could benefit from GOTV. HRC could not. If this is indeed true, this is yet another evidence that we need more science and less data.