The problem with any poll is determining turnout. If you know the makeup the electorate, you can predict the outcome of an election pretty clearly.
The question for this election is whether Cankles will be able to reassemble the Obama coalition, which counted on a very high number of African Americans and Hispanics. I just do not think that minority voters are going to go out in as large of numbers as before.
All polls are weighted based on what they expect the electorate to look like. So most polls are expecting a Democratic leaning electorate (based on 2008 and 2012), somewhere between 4 and 8 percent D advantage. But what if the electorate is roughly even on election day? Then all of the polls are completely off base.
That is a good question and a reasonable hypothesis, but there are ways to test it with reasonable accuracy - for example by asking people in a poll if they're definitely voting, probably voting etc. Pollsters have experience polling for non-Obama candidates and they can get a reasonably good sense of the likelihood of people actually voting. In addition to lower turnout, I would have thought that 2016 might see a return to pre-2008 partisan affiliation for black voters - back when blacks "only" voted 80-85% Dem. At the moment, the polls seem to indicate that Trump might actually do worse with black voters than Romney did. I'm not sure I understand that - but I trust math more than hunches.
Agreed.