Every assessment of the state of polling in the 2016 race, including my own, must include a key caveat: Big shifts in the electorate can mean that the likely voter models the pollsters were using are wrong. And if the pollsters are making incorrect assumptions about who will vote, they'll end up with incorrect assumptions about who will win.
This is one of the reasons that Gallup tracking polls in 2012 showed Mitt Romney with a lead. Their likely voter model missed in its estimate of who would vote. The polls included too many white voters, and the result was a poll that was the L.A. Times-USC poll of its time: Embraced by supporters of the candidate who went on to lose.
Now, as then, supporters of the trailing candidate have consistently argued that pollsters are missing something. In 2012, they were: Pollsters actually underestimated the extent to which supporters of President Obama would come out to vote. Donald Trump, his team and those who back his candidacy argue that a similar effect is happening in the opposite direction this year, with Hillary Clinton's support being overestimated.
Early-voting data suggests that may not be the case.
The graph below uses data from Daniel Smith, a University of Florida professor who has been tracking the early returns in the state for some time. It compares the composition of the early vote in 2012 (from left to right) with the composition in 2016 (from top to bottom). Anything above the diagonal line shows an increase in density since four years ago.
Only Hispanic voters are above that line. As a percentage of all early voters, the Hispanic vote grew 4.4 percentage points.
This mirrors what we've seen elsewhere. In Nevada, a big surge of Hispanic voters spiked turnout numbers in Clark County, the state's most populous. Long-time observers of politics in the state think that it may now be out of reach for the GOP.
What's more, Smith writes on his blog that the Hispanic voters turning out are heavily people who didn't vote at all four years ago. "So far, 36% of the 907k Hispanics who have voted in 2016 didn’t vote by any method in 2012," he writes. "That’s a full 12 points higher than whites, and will likely be the key to who wins the presidency."
That's important because there is much more room to grow in terms of turnout among Hispanic voters.
Historically, Hispanic registered voters have turned out to vote at much lower rates than white voters. (The chart above, using data from the United States Elections Project, includes only non-Hispanic whites.) In fact, Hispanics vote in presidential races at about the same rate as whites do in midterms. For years, whites turned out more than any other demographic group, which helped shape American politics — especially since whites are the largest demographic group in the country.
That changed in 2008, when black voters suddenly took the turnout crown. It's not the case that there were more black than white voters eight years ago, but as a percentage of registered voters within those groups, more black voters showed up to vote.
In 2008. The year the first black president was elected. Which is to point out that external motivations can shift voting behavior.
Pew Research already anticipated that Hispanic voters would make up more of the electorate now than in years past simply by virtue of there being more Hispanic Americans than in elections past. (This trend is one reason that the percentage of the electorate that was nonwhite in 2014 was equivalent to the percentage that was nonwhite in that 2008 election.) That's without considering a possible spike in turnout among Hispanic voters.
So let's go back to Florida. Smith estimates that the number of Hispanic votes cast early so far probably surpasses all Hispanic votes in the state in 2012. One thing we're not seeing in the state is a big surge in infrequent white voters of the sort the Trump campaign had promised. Smith told Bloomberg News that new registrants in Miami-Dade County, the largest in the state, had broken heavily to the Democrats in October — 41 percent were Democrats and 44 percent unaffiliated, with only 12.5 percent of new registrants being Republicans. That doesn't suggest an outpouring of unprecedented enthusiasm for Trump.
In fact, in the most recent national Washington Post-ABC tracking poll, the levels of enthusiasm between supporters of the two candidates are about equal. Half of Clinton backers and just over half of Trump supporters say they're very enthusiastic about supporting their candidates, and 87 percent of the supporters of each say they're "very" or "somewhat" enthusiastic. (That poll also showed Clinton winning early voters by 16 percentage points, while trailing Trump by two points among those who plan to vote on Election Day.)
If Hillary Clinton wins Florida, the race is over, save a stunning development like Donald Trump winning Michigan or California. If we're seeing a surge of infrequent Hispanic voters heading to the polls in the state — a group likely to heavily favor Clinton — but not an equivalent surge among Trump's base, it's hard to see how Trump ekes out a victory in a race the polls put as a dead heat.
In other words: Once again, the polls will have missed the demographic numbers that made the difference.