To reinforce that, I can point to an interesting pair of graphs from the New York Times' Florida poll that was released on Monday, conducted with Siena College. The first graph shows how views of the two major-party candidates vary by the likelihood that a voter will go to the polls -- and the second weights those results by that likelihood.
Hillary Clinton has big leads among a lot of those infrequent voters, but Donald Trump's lead with those most likely to vote makes a much bigger difference.
The most recent Post-ABC News poll offered a similar result. For Clinton, 80 percent of her supporters who were registered to vote identified themselves as "certain" to cast a ballot, with another 9 percent saying they would probably do so. For Trump, the figures were 93 percent and 5 percent, respectively.
Part of this is due to the fact that demographics that tend to support Republicans also tend to vote more often. Retirees, people with steady jobs and people who own their own homes are frequent voters because they don't have to re-register to vote after moving as much and they can make time to get to the polls. Those demographics overlap with income, and therefore race: It's usually older, wealthier and whiter voters who vote more reliably. And those are voters that often vote Republican.
There's variation within the numbers, though, that can affect this race in particular. Let's look at data from three Post-ABC polls: One taken in Sept. 2012, our poll from August and our most recent poll. We've broken out several demographic groups by age, race and education, looking only at responses among registered voters.
Some things to note in that:
- Between the August and September Post-ABC poll, there was a 5-point jump among whites without college degrees in their saying that they were certain to go vote. It had been at the same levels as in Sept. 2012.
- The increase since August among white men without degrees was 11 points -- while the percent of white women with degrees who said they were certain to vote fell by 8 points.
- Among nonwhite voters, the percent that said they were certain to vote in Sept. 2012 was 77 percent; now, it's 72 percent.
- Nonwhite voters with college degrees are the most likely to say they're certain to vote, but they're a much smaller percentage of the voting population.
Again, this is largely in line with the long-term trend. Hispanics turnout in presidential elections at about the rate whites turn out in off-year races. Only in 2012 and 2016 did black voters turn out at the same levels as white voters. That discrepancy is one reason that Republicans have recently fared better in non-presidential election years. But it's not entirely in line with that trend, particularly given the fluctuations you see above. If college-educated women (who like Clinton) decide to stay home and white men without degrees (who love Trump) don't, that changes things, even if white voters overall are about the same percent of the electorate.
The inherent turnout advantage enjoyed by Republicans is one reason that the Democrats often have much stronger voter-turnout operations, which Clinton will need this year. When your voters regularly vote, as the GOP's do, you have less of a need to build processes that will ensure they do so. Donald Trump's campaign doesn't seem to be putting much emphasis on turnout, opting instead to hand things over to the Republican Party. The party is confident that it has a good system in place, though it hasn't been tested in a presidential election year.
Underlying all of this is a broader question: Can Donald Trump's candidacy somehow reverse the trend in the diversity of the electorate? A few weeks ago, we made an interactive showing that with the electorate that turned out in 1980, the 10-point margin between Trump and Clinton (as it existed at the time) became a margin of zero. If Clinton turns out more black and Hispanic voters in Florida, the 1-point race that Siena and the Times found tips in her direction. If Trump turns out more of those white men without college degrees, it tips the other way.
Pollsters spend a lot of time trying to predict that turnout for precisely this reason. In 2012, a lot of polls underestimated the number of nonwhite voters who would come out, and so underestimated the extent of President Obama's victory. Knowing that the race comes down to turnout isn’t the same as knowing the result.