I do find it interesting how consistently Clinton's lead has been coming from doing better than Obama with white voters, with the Hispanic vote barely budging from 2012. Trump is consistently at or over 20% with Hispanics, so I think we need to start taking that possibility seriously.
Most national polls don't do a very consistent job of capturing a good picture of how Latino voters will actually break on election day.
Distribution of partisanship within an ethnic group is well established. What is hard to predict at times is voter turnout. So if Democrats are successful in getting Hispanics out to vote in high numbers, then they can get more electoral successes. But this is not predictable.
...Barack Obama may be one of the slickest politicians ever, but he also got strong negatives from white voters in the Mountain and Deep South. The white vote for Obama was about normal for a Democrat in the northeastern quadrant of the US and the far and Mountain West, but far below average for a Democrat -- almost in McGovern-Mondale territory -- in Mississippi (the definition of the Deep South), West Virginia (the definition of the Mountain South), and Georgia (which straddles both regions). It will take a huge rebound for any Democratic nominee to make states in the Mountain and Deep South competitive again.
As this election approached I thought that Hillary Clinton would do about as well as Obama did in 2012 in the northeastern quadrant and the West, which is good enough to win. I also expected her to get rebounds to levels of support characteristic of Al Gore in the Mountain and Deep South, which would not be enough to win there. This assumed a reasonably-competent Republican capable of holding the recent GOP coalition together.
Of course we all know how assumptions go. She is doing about as well as Obama in the Northeast and the West. But Trump is simply awful.