re: Battleground polls: Will Trump defy the odds again?

There were some major problems with the polling from 2016 that likely led to the misses.

1. Limited state-level polling, particularly in the states that tipped the election. For example, RCP had roughly 17 polls from August through the election in Wisconsin. There are already over 20 polls with an entire month to go in Wisconsin.

2. Pollsters were largely not weighting by education, and due to nonresponse biases, the groups that tipped the election for Trump in the Midwest (whites without a college education) were underrepresented. So if they had weighted for this underrepresentation, then the polls would have been much closers. Most pollsters have addressed this issue and are weighting by education.

So while it's possible that the polls are still underestimating Trump in those states it missed in 2016, I think the margin will be much lower due to addressing the issues they can address. In fact, I think there a chance they could OVERCORRECT since that's often the inherent response AND the incentives favor estimating closer toward a toss-up.