My projection of the national popular vote is tied today. One outlier poll is weighted down because it is more than five points off of the average of the other national polls and the bellwether states of Florida and Ohio. Without the inclusion of that poll, Trump is actually leading nationally. I will update the state polls later this evening.
We are starting to see some of the pundits and analysts hedge their bets. None more notable than Nate Silver who confidently declared that Trump had no chance of winning the Republican nomination and then Trump won with more votes than any nominee in GOP history. His prognostications were so flawed with respect to Trump that the Daily Caller was able to point out seven times Nate Silver was "hilariously wrong about Donald Trump." Reading his latest analysis, one wonders if he is starting to fear an eighth.
Today Silver is highlighting the prospect of a split between the electoral college and national popular vote. It does seem more likely this time around, which underscores the error in placing great weight on national polls or averages of national polls, as Sabato, Chris Stirewalt and other establishmentarians smugly do each night. The reason a split is more likely is that Trump is not doing as well in red states as a typical Republican, but will do well enough to win them. Take Texas as an example. Trump will win the state but not carry it by as big a margin as usual, which hurts his popular vote total but not his electoral vote total. Hillary, on the flip side, is running up the score in blue states like California, Illinois and New York where many Hispanics reside. Those big margins do not translate to more votes in the electoral college.
Silver also notes that Clinton is doing worse than Obama in ten of the twelve battleground states, which is an indication that Trump can potentially take any of those ten. She is substantially weaker in Michigan in Silver's analysis, as compared to Obama in 2012. He notes that while Hillary is stronger with Hispanics than Obama, she is weaker with African-Americans, and there are more African-Americans in swing states in addition to more white working class voters, Trump's base.
To be fair to Silver, he has written that Trump supporters should not give up in the past when all of the analysts were saying the race is over, and that was when he was projecting that Trump only had a one-in-eight chance of winning. He is now projecting that Trump has a one-in-four chance, and that's saying a lot for this particular analyst. Silver cautions that there exists "significant disagreement" within the polls, and a "relatively high uncertainty" surrounds this year's polling. One car fairly interpret his cautions as hedges against being wrong, and it seems he's fearing that much more now. He should.
I think this high uncertainty and significant disagreement represents the clash between traditional pollsters using unjustified and hopeful assumptions, who happen to be anti-Trump on an emotional level that rattles their sense of objectivity, and the just the facts approach of newer online pollsters like LA Times/Daybreak (Trump +4) and UPI/CVoter (Tied 48-48). Those two polls bucked the trend last time and were very accurate. In addition to recent history, the state polls showing a very close race also suggest those polls are closer to the truth.
The high black and millennial turnout assumption pollsters are making that seem so obviously wrong to an objective viewer are throwing off the numbers. Focus groups and polls that establishment pollsters respect have shown us that Hillary is having great difficulty motivating black voters, especially younger ones, and also millennial voters. Traditional pollsters are ignoring their own data in assuming a repeat of 2012, which is where we see the anti-Trump bias manifest. That will be the deciding factor in 2012 about who is right, whether this data suggesting lower turnout among blacks and millennials pans out.
More than that, if Trump is drawing new voters who have never voted before, as we have heard anecdotally for months now, even the online polls will under-represent his voters. A voter who has never voted presumably just doesn't much like the process. That voter is less likely to sit through a twenty-minute survey on the phone or go through a thirty-question form online. While many of the online polls will not exclude them as a non-likely voter because they didn't vote before, those people are almost certainly less likely to sit through the survey.