For those of you looking for the bottom line, I predict that Barack Obama will carry North Carolina's 15 electoral votes by a vote of 2,263,022 to 2,258,945, a margin of 4,077 votes. In percentage terms, that means Obama will win by 49.58% to 49.50%. 27,232 votes will be cast for Gary Johnson and 14,761 votes will be cast for write-in candidates. There will be 45,388 undervotes.
Oh, and GOTV.
Obviously Obama's predicted margin of victory is FAR, FAR within the penumbra of uncertainty. As I see it, North Carolina is as much of a tossup as a tossup can possibly be. The primary measure by which I will judge the accuracy of this prediction by is the predicted vote margin of 4,077 votes for Barack Obama.
Also, I am going to try to create a magic spreadsheet which will project the results in North Carolina once the results start coming in after polls close at 7:30.
If FL and VA report results fairly slowly, and if NC dumps all its early votes shortly after the polls close at 7:30, then the early North Carolina results may be the first real indication, other than from exit polls, of how the presidential race is going nationally. And since the data is coming directly from the NC Board of Elections, it should be possible to get results that way as fast or FASTER than the AP. If possible and if time allows, I'll also try to do a projection for the Congressional race in NC-07.
More details on that at the end of the diary.
What should we look for on election night?:
1) If Obama wins early votes by about 150,000-200,000, he is in position to win NC, but it will be very close. If Obama wins early votes by more than 200,000 votes, though it will still probably be close. If Obama wins early votes by less than 150,000, then it will be tough (though not necessarily impossible) for him to win North Carolina.
2) If Obama is winning about 33% of white voters, he is in the position he needs to be to win NC, but it will be close. If he is much above that, it's looking good for an Obama victory in NC. If he is much below that, Romney will likely win NC.
-------------------------------------------------------------------
North Carolina is a tossup and Romney is Desperate
Let's put this very simply.
Conventional wisdom says that Barack Obama cannot win North Carolina. As soon as pundits started thinking about the 2012 election, they immediately began by assuming that Obama would lose Indiana and North Carolina. After all, he only barely won them in 2008, so if there is a uniform national swing to Romney, then those should be the first states that he picks up - and Romney should win them handily. So the story goes.
For Indiana, I won't argue with the conventional wisdom. But the conventional wisdom that North Carolina is a safe Romney state is neat, plausible, and wrong.
If you look at the actual data, rather than relying upon glib a-priori logic, you see a very different picture of North Carolina.
There is nothing in the early vote data - or in the data on registered voters who will vote on election day - that indicates that North Carolina is anything but a tossup.
Most of the public polls, and especially the most recent ones, show a tie, or maybe Romney up by 1 point.
PPP has polled North Carolina 3 times in approximately the last two weeks. Each time, they have found NC tied.
High Point University has found Romney up by 1.
Elon University has NC tied.
Yougov has Romney up 2. (I don't really trust internet polls, but nevertheless I'll count it...)
The only somewhat troubling result from a non-right wing pollster is from SUSA, who found Obama down by 5.
If we average all of those 5 pollsters, we find Romney up by 1.62 points.
Rasmussen and Gravis Marketing say that Romney has a large lead, but if they did a poll of Vermont, they'd probably find Romney ahead there as well.
On the other side, Grove Insight has Obama up by 3.
1.62 points, if that is really how things stand in NC, is not a very big lead for Romney. Obama has a number of advantages that are not fully taken into account in the polls, and it's not hard to imagine they could be worth more than 1.62 points.
Chief among Obama's advantages are his well oiled GOTV turnout machine. It's not exactly a well kept secret that Obama's North Carolina ground game is light years ahead of Romney's.
Obama's organizers are better disciplined. There are more of them. Obama's volunteers are better trained. There are more of them. Obama's ground game is battle tested from 2008, and has been built up meticulously over a period of four years, with painstaking attention to even the smallest of details.
Meanwhile, Romney has to keep pretending that NC is safe Republican, and keeps halfheartedly pretending that he is withdrawing from North Carolina and is expanding the map into some other state. Some days that other state is Michigan. Some days it seems to be Wisconsin. Other days it's Pennsylvania.
Romney knows that Ohio and Nevada are already lost, because he is already down by so much in early voting. He knows that North Carolina is teetering on a knife edge, again because of early voting.
Incidentally, Michigan, Wisconsin, and Pennsylvania all have something in common - they are heavily white and don't have substantial early voting. That means that, while Romney is going to lose them, he has not already lost them. And that is why he is forlornly tramping around in those states.
Romney knows that he is losing, and he knows that the only we he can hope to win is by creating a self-fulfilling media narrative prophecy that he is already winning, which (he hopes) will cause many people to vote for him at the last minute. The problem with that approach is that it is not grounded in reality, but rather is built on air.
Like Wile E. Coyote, as soon as Romney looks down at the foundation upon which he is (or is not) standing, he begins to fall. The only way forward is to simply refuse to look down, put on a brave face, and hope the media doesn't notice. It's a confidence game, and Romney is a confidence man.
The facts manifestly say that Romney is a desperate, losing, wounded animal, thrashing about blindly in the general direction of anything that sounds remotely like it is moving.
PPP Poll:
Late last night, PPP came out with one last poll of North Carolina. They found a tied race 49-49 (actually, it was 49.4% Romney to 49.2% Obama).
As dreaminonempty has ably demonstrated, internals of polls are not always reliable, to say the least - even if the top lines are reliable. Nonetheless, I will cautiously scrounge through the crosstabs of PPP's poll and see whether they fit with my assumptions.
First of all 62% of PPP's sample has voted early, while 38% of the sample plans to vote on election day. Since 2,738,922 people have voted early so far, that would suggest a final turnout of about 4,417,616, which is in the right general area. So PPP's likely voter screen is probably about right.
Among early voters, PPP finds Obama up 54-45, which is not far from my 53.1-46.9. Among election day voters, PPP has Romney up 57-41.
73% of PPP's sample is (or reports to be) White, 21% is African American, and 6% is "Other." That's about right, but in 2008, African American vote share was 22.3%. Among early voters in 2012, it is 27.4%. Overall in 2012, I expect it to be about 23.7%, with White vote share dropping from 73.2% in 2008 to about 70.6% in 2012.
In 2008, the exit polls say that Obama got 35% of the white vote. PPP finds him at 37% with white voters (with the caveat that probably not all respondents to the poll who claimed to be white were actually white). Averaging the White voter support rate from PPP, SUSA, Elon, and Yougov, Obama is at about 33.5% with white voters in NC. That's about what he needs to win.
So everything in PPP's poll is extremely consistent with the methodology I am using here. I think that PPP's poll is likely to be pretty accurate, and like PPP, my best guess is that both Obama and Romney will get under 50% of the vote.
Prediction:
Here is my prediction. It is very precise, in the sense that it gives specific numbers for everything. But in reality, there is a hazy undefined margin of error surrounding all my predictions. In the end, we just don't know.
And that, again, is why it GOTV is vital.
Let's start by looking at who has already voted - people who have their votes banked early:
About 2,738,922 people have voted early. There are still a few mail ballots that will trickle in, but that's basically the final total number of early voters.
55.2% of those 2,738,922 people are "likely voters" who voted in both 2008 and 2010. Of those, White Republicans dominate - they are 34.4% of the likely voters who have already voted. But we knew they would vote all along - it was just a question of whether it'd be early or on election day. 24.7% are African American, 24.0% are White Dems, and 1.6% are other minorities.
But among the remaining 44.8% of people who have already voted (a total of 1,227,251 million people), it's a totally different story. 689,528 "Sporadic" voters who voted in 2008 or 2010 but NOT both have voted. 149,559 "Unlikely" voters have voted even though they didn't vote in EITHER 2008 or 2010 and registered to vote some time before 2010. And 388,164 New Registrants have voted - including 70,058 who registered VERY recently - some time after 10/27/2012. Presumably these 70,058 are brand new voters who registered at One Stop Early Voting. Also, some 10s of thousands of additional One Stop Early Voting New Registrants are included among the 318,106 people in the regular New Registrants category.
So, among those 1,227,251 "Non-Likely" voters, the share of White Republicans drops from 34.4% (the White GOP share of "Likely" Voters) to just 25.1% of "Non-Likely" voters. Equally and oppositely, the share of African Americans rises from 24.7% among "Likely" Voters to 30.8% among "Non-Likely" voters, and the share of other Minorities increases from 1.6% to 4.7%. Obama very clearly expanded the electorate during the early voting period. There's a word for that. And there's a word for that - it's called GOTV.
Now, who is left that still hasn't voted?
Well, there are 5,951,357 Active Registered voters, so if 2,738,922 have already voted that leaves 3,234,140 who have not yet voted. In addition, there are another 709,744
"Inactive" registered voters. These are people who didn't return the confirmation mailing of their registration. They can still vote, but very few of them actually will, so we'll focus on the active registered voters.
Of the 3,234,140 Registered Active Voters who haven't yet voted, 909,336 of them are "Likely" voters. As with "Likely" voters who have voted, White Republicans dominate with 355,477 Republican Likely voters, or 39.1% of all likely voters. But among the other 2,324,804 "Non-Likely" voters who haven't voted, only 25.4% are White Republicans.
Simply put, the more that Obama manages to expand the electorate to include more of those "Non-Likely" voters who haven't yet voted, the more likely it is that Obama wins. And there's a word for how that's done - it's called GOTV.
Now, what percentage of the 3,234,140 Registered Active Voters who have not yet voted will vote on election day?
Overall, I predict that 58% of the 3,234,140 remaining voters will vote. That will result in 1,870,426 election day voters, which means that:
1) Total turnout of "accepted" ballots will be 4,609,348.
2) Total turnout of ballots cast in the presidential race will be 4,563,876, meaning that there will be 45,472 undervotes.
Both of the above total turnout numbers represent a 5.9% increase in turnout above 2008, which is equal to the increase in the Voting Eligible Population in North Carolina from 2008 to 2012.
But a lot of those 1,870,426 people won't vote by themselves. They need some help. There's a word for how that's done - it's called GOTV.
Now, how the heck did I come up with those percentages?
I averaged two other percentages to come up with those crazy percentages. The idea behind this averaging approach was to take into account two different countervailing effects.
First of all, if a demographic group votes early at a high (or low) rate, that is an indication that this demographic group is enthusiastic (or unenthusiastic) about voting this year, and so members of this demographic group will also vote at a high (or low) rate on election day. So, for example, the fact that Native American turnout was low in early voting is probably an indication that it will be fairly low on election day as well, among Native Americans who haven't yet voted.
But on the other hand, there's a countervailing effect. If a larger proportion of a demographic group has already voted early, then it is likely that the "best voters" among that group have already voted, and thus the people who are left over are the "bad voters" who are less engaged with politics. Because of this effect, turnout of remaining will tend to be lower on election day. In the case of North Carolina, this effect applies especially to African Americans, but also to other groups that have had high turnout rates like White Republicans.
I don't know how strong each of these effects is, so I simply averaged them.
So the first percentage that I included in the average is the % of Active Registered voters who have already voted:
Now, at this point it must be said - I spotted a math error. The first time I went through this process, I forgot to update the % of Active Registered Voters who had already voted.
I had gotten all the way through the rest of my projection, and had even gotten my final number and had taken my screenshot. This is what my final prediction was GOING TO BE:
That's right, many readers probably won't believe this. Many readers will probably think that I rigged my prediction - but that's really and truly not actually true. Because of a math error, I was on the verge of predicting that Romney would with North Carolina by 1,101 votes - or by 49.55% to 49.53%.
So anyway, I went back and updated the percentages of people who voted early, to include the last Saturday of early voting. Because Obama did well on the last Saturday of One Stop Early Voting, this made him do better in my prediction:
So, now we have the first percentage included in my average - and these are the corrected numbers. The second percentage that I included was given by me. These are constant regardless of race or party registration, with one exception - White Republicans get a 15 point boost. The justification for that is the idea that White Republicans have not voted at quite the same rates as African Americans so far, but they voted at high rates in 2008 on election day and likely will do so again.
The way I came up with these "given" percentages was I simply increased them until they yielded a turnout number close to my target expected election day turnout of 1,870,426 people. When I was close to that target, I then multiplied each percentage by the constant necessary to exactly reach the expected election day turnout (the magic number 0.969059145 ...)
Then I multiplied those turnout percentages by the numbers of active registered voters who hadn't yet voted:
These are the voters predicted to vote on election day.
Then I added them to early voters to get total voters:
This implies that the turnout rates for ALL registered Active Voters will be the following:
With the number of total voters, I could then calculate how they would be predicted to vote:
This spit out the prediction that Obama will win North Carolina by 50.4% to 49.6%, or by a margin of 39,487 votes.
However, there were still a few more things that I needed to correct for. First, I needed to correct for the error in my predictions of 2008 (which I found in yesterday's diary). To do this, I multiplied my predicted 2012 Obama votes by the actual 2008 Obama votes divided by my predicted 2008 Obama votes. Likewise, I multiplied my predicted 2012 Romney votes by the actual 2008 McCain votes divided by my predicted 2008 McCain votes.
Second, I needed, there are going to be some under votes. To correct for this, I simply assume that the number of undervotes will increase at the same 5.9% rate as total turnout. I took undervotes away from Obama and Romney, in equal proportion to the overall predicted result.
Third, some people will vote for Gary Johnson or a write-in Canddiate. To correct for this, I simply assume that the number of Gary Johnson and write-in votes will increase at the same 5.9% rate as total turnout. I took Gary Johnson and write-in votes away from Obama and Romney, in equal proportion to the overall predicted result.
Fourth, I needed to calculate how many ballots would be "provisional" ballots. To do this, I simply multiplied the proportion of Obama's 2008 ballots that were provisional ballots by his predicted 2012 ballots, and multiplied the proportion of McCain's 2008 ballots that were provisional ballots by Romney's predicted total 2012 ballots.
Finally, I suspect that the number of votes Obama will get in early voting is higher than race and party demographics alone would predict, and likewise, I suspect that the number of votes Romney will get on election day is higher than race and party demographics alone would predict. So, because there is no time to do this in any better way, I am arbitrarily transferring 25000 early votes from Romney to Obama and transferring 25000 election day votes from Obama to Romney. This doesn't change the total amount of votes each candidate is predicted to receive - just how many of them will be from early voting vs. from election day voting.
All of this results in the following Final Prediction:
I predict that Barack Obama will carry North Carolina's 15 electoral votes by a vote of 2,263,022 to 2,258,945, a margin of 4,077 votes. In percentage terms, that means Obama will win by 49.58% to 49.50%. 27,232 votes will be cast for Gary Johnson and 14,761 votes will be cast for write-in candidates. There will be 45,388 undervotes.
Oh, and GOTV.
Obviously Obama's predicted margin of victory is FAR, FAR within the penumbra of uncertainty. As I see it, North Carolina is as much of a tossup as a tossup can possibly be. That is why it is important to vote.
And that is why it is important to GOTV.
The primary measure by which I will judge the accuracy of this prediction by is the predicted vote margin of 4,077 votes for Barack Obama.
Election Night Returns
Also, I am going to try to create a magic spreadsheet which will project the results in North Carolina once the results start coming in after polls close at 7:30. I have discovered that North Carolina seems to put a file (WARNING - BIG FILE) on its website with all the precinct level election night returns.
I presume, but do not know for sure, that the NC Board of Elections will be updating it frequently on election night as they process results.
If FL and VA report results fairly slowly, and if NC dumps all its early votes shortly after the polls close at 7:30, then the early North Carolina results may be the first real indication, other than from exit polls, of how the presidential race is going nationally. And since the data is coming directly from the NC Board of Elections, it should be possible to get results that way as fast or FASTER than the AP. If possible and if time allows, I'll also try to do a projection for the Congressional race in NC-07.