North Carolina has become perhaps the most important state in the election. It has the potential to decide the Senate and the presidency.
Hillary Clinton has led every live-interview survey conducted there since the first presidential debate, even though Mitt Romney won it four years ago. She has a comfortable lead in the surveys taken after the third presidential debate, with Upshot/Siena, NBC/Marist, Quinnipiac, Monmouth and Elon polls showing her ahead by an average of four points.
It’s also a state where the election is well underway. Nearly two million voters — perhaps 40 percent of the electorate — have already cast ballots, and the data from early voting suggests that she has banked a considerable lead. The same data implies that pre-election polls are largely right about the composition of the North Carolina electorate.
This doesn’t mean the pre-election polls will be right. There are a lot of reasons the polls could be wrong besides the composition of the electorate. For instance, undecided voters or the supporters of Gary Johnson could break for Mr. Trump.
Continue reading the main story
What the Early Vote in North Carolina Means: A Daily Tracker OCT. 26, 2016
THE 2016 RACE
Why Republicans Don’t Even Try to Win Cities Anymore NOV. 2, 2016
Why Does Education Translate to Less Support for Donald Trump? NOV. 1, 2016
THE 2016 RACE
Get Ready for Another Swing in the Polls, but Not Necessarily a Shift in the Race OCT. 31, 2016
But if the polls are wrong, it probably won’t be because they assumed an electorate that was very different than the one that showed up to vote.
Clinton Leads Early Voting, but Doesn’t Gain Ground
No matter how you look at it, Mrs. Clinton leads among the voters who have already cast ballots. That’s true whether you look at the new survey data from Quinnipiac or NBC/Marist; it’s also true if you just look at the demographic characteristics of early voters, who are likelier to be registered Democrats than registered voters over all.
We can look even more closely with data from the Upshot/Siena polls, conducted in September and October. On average, they show Mrs. Clinton ahead by around three points among likely voters. The surveys were conducted using voter file data, so we can match our poll respondents to the state’s absentee data file — and figure out which of our respondents have already voted.
So far, nearly 600 of our respondents have voted early — basically a full poll’s worth of early voters. For this analysis, we’ve weighted this subsample of validated early voters to match the demographic characteristics of early voters by age, race, party, sex and 2014 vote history.
Over all, Mrs. Clinton leads among these early votes by 51 percent to 39 percent in the three-way race, and by 53-39 in the two-way race.
Most of these people were expected to vote, based on their vote history: We gave more than 70 percent of these voters a greater than 90 percent of voting.
In part as a result, these early voters aren’t doing much to move the electorate in her favor.
Ahead of early voting, we set up an “early voting tracker” that was designed to adjust our estimated composition of the electorate as early voting went on. Every day, early voters lock in their votes; the voters who stay home become a little less likely to vote.
So far, there hasn’t been anything about early voters that has changed our prediction for the composition of the electorate. There are basically no unusual trends: There aren’t any demographic groups in which low-turnout voters are locking in their vote at a particularly fast or slow rate. As a result, our early voting tracker still shows Mrs. Clinton ahead by 5.7 points — the margin implied by our poll.
Interactive Feature: 2016 Election Forecast: Who Will Be President?
This has been misinterpreted a little bit (and that’s partly our fault). The takeaway is not that the early voting proves that Mrs. Clinton will win by 5.7 points, or that our poll is right. The takeaway is that the early voting indicates that the electorate will pretty much have the demographic breakdown we’ve expected.
Early voting trends are validating our expectations for the eventual turnout. The polls might still be wrong — say, if undecided voters break to Mr. Trump. If so, our model will be wrong, too.
Early Voting Might Surprise Pollsters
Early voting hasn’t surprised our models, but it might surprise the pollsters.
Not all of our early voters said they were going to vote when we talked to them in September or last week. Only 81 percent of these voters said they were “almost certain to vote” or had “already voted” at the time we interviewed them. The other 19 percent of voters would have been missed in many “likely-voter” screens.
Mrs. Clinton had a big 53-33 lead among those voters who said they were less than “almost certain” to vote, compared with a 52-41 margin among those who said they were likely to vote.
In the compilation of data from our two North Carolina polls (which averaged to a three-point lead for Mrs. Clinton), she had a one-point lead among voters who said they were “almost certain” to vote or had already voted. She would lead by two points if one added the voters who were less likely to vote but have since cast their ballots in early voting.
Most of the people who would have been considered unlikely voters were people who have a robust track record of voting in midterm elections or primaries. As a result, this hasn’t affected our model, which is based on vote history, even as it has the potential to throw off the polls.
Registration Shifts Drive Apparent Turnout Shifts
It’s understandable that people are comparing this year’s early vote with the early vote from 2012.
Those comparisons haven’t looked so favorable for the Democrats in North Carolina. The turnout among registered Democrats is down. The turnout among registered Republicans is up.
Well, there’s a pretty straightforward reason for that: The number of registered Democrats in North Carolina has declined since 2012. The number of registered Republicans has ticked up.
Over all, the number of registered Democrats has declined by 5.1 percent. The number of early voters who were registered Democrats has dipped by 3.1 percent — so the Democratic turnout hasn’t dropped as much as the number of registered Democrats.
Republican turnout has increased by more than the increase in Republican registration, but mainly because reliable Republican voters are turning out in greater numbers.
As a result, we expected the Democratic registration edge to decline from 11 to 8 percentage points heading into the early vote. That expectation has not changed based on the data so far.
Black Voters Turn Out Like Everyone Else
In 2012 and 2008, black voters were much likelier than white voters to vote early.
That’s not true this year. The black share of early voters has dropped to 23 percent from 28 percent in 2008.
There’s no question, in my view, that this indicates that black voters are less enthusiastic than they were in 2012.
But it does not suggest that there will be a huge collapse in black turnout — say, falling from 23 percent to 19 percent of the electorate. Instead, it indicates that black voters are basically voting like nonblack voters.
There are many reliable black voters — midterm and primary voters — left to cast ballots. Indeed, more than half of black voters who turned out in the 2014 midterm are yet to vote. The fact that they didn’t rush to vote on the first day of voting this year doesn’t change that they remain extremely likely to turn out, whether now or on Election Day.
What would it mean if black voters turned out at the rate implied by their vote history, rather than the elevated level of turnout from 2012 or 2008? They might represent something around 21.5 percent of the electorate, rather than the 23 percent they did in 2012.
This is an important shift. But it’s not an unexpected one. None of the pre-election polls suggest that black voters will represent 23 percent of the electorate, so there is no reason to assume that the polls are biased toward Mrs. Clinton on that basis.
Instead, they suggest that she would win because of strength among well-educated white voters.
The Unaffiliated Voter Mystery
Perhaps the most striking change from 2012 is the increase in turnout among unaffiliated voters. Their turnout has increased by nearly 40 percent in four years.
Again, registration changes account for a lot of the shift: The number of registered unaffiliated voters is up 21 percent. As with Republican voters, there has also been an increase in turnout among reliable unaffiliated voters.
The big number of unaffiliated voters, however, makes it more important to try to figure out which way they lean. Their demographic characteristics are conflicting: On the one hand, they’re disproportionately white. On the other, they’re disproportionately young, urban, newly registered and born outside of the state.
In our poll of early voters, Mrs. Clinton has a 49-39 lead among unaffiliated voters who have cast ballots. Mr. Trump has only a 10-point lead, 47 percent to 37 percent, among unaffiliated white voters.
More generally, Mrs. Clinton has a two-point lead among unaffiliated voters in our two North Carolina polls, which average to a three-point lead for Mrs. Clinton.
In fact, she has a lead among unaffiliated voters in all five Upshot/Siena polls — including those in Florida and Pennsylvania.
National polls have shown a steadily tightening race, mainly because Republican-leaning voters have returned to Mr. Trump. They have also shown a steady weakening in the standing of Mr. Johnson.
These trends could help Mr. Trump considerably in North Carolina.
In Upshot/Siena polls of North Carolina, the voters who aren’t backing a major party candidate are disproportionately white, Republican and male. Over all, 18 percent of white voters were undecided or supporting a minor-party candidate in our two surveys, compared with 11 percent of black voters. Similarly, 15 percent of registered Republicans were undecided or supporting a minor-party candidate, compared with 11 percent of Democrats. And 19 percent of men weren’t supporting a major-party candidate, compared with 15 percent of women.
If the polls are ultimately wrong, and Mr. Trump prevails or Mrs. Clinton barely wins in the state, undecided and minor-candidate voters are a lot likelier than turnout to be the reason.