Just Peachy
One of my obsessions in this election's final stage has been Georgia. It's a very red state that has been tinged positively purple. Its senate race seems poised for a run-off, and could provide the 60th Democratic seat. And it's the swing-state in which I suspect the polling has been most dramatically underestimating Barack Obama's support.
That's crazy. So crazy, in fact, that SurveyUSA has taken the extraordinary step of offering a second set of data. It took its polling, and re-weighted the sample to project that black voters would account for 30% of the ballots. That's not terribly outlandish; they currently account for 29% of registered voters in the state, and this assumes that they'll comprise a hair under 24% of election day voters.
The re-weighted sample makes the race too close to call. McCain would lead Obama just 49-48%. Jim Martin would actually outpoll Saxby Chambliss, 47-46%, although the race would still head to a run-off.
So is that mere fantasy? Perhaps not. Simply increasing the percentage of black voters widened Obama's lead among early voters to 53-45%. I have to assume that when SurveyUSA re-weighted its sample, it used the documented 34.9% figure for early voters. To do otherwise would be nonsensical. So more than half of this swing isn't purely speculative, but an artifact of adjusting poll numbers to reflect an actual count of the ballots.
Here's a suggestion. When it comes to Georgia, disregard the polling averages and projections. On the strength of a remarkable and historic surge in black turnout, this race is simply too close to call. And watching the results roll in should provide one of the more entertaining and suspenseful spectacles tomorrow night.
If you've enjoyed this, please share it with other readers by clicking the 'recommend this' link. You can find more analysis on my blog, or subscribe by clicking "Follow Me" on the right. As always, I welcome your comments and corrections, and thank you for your feedback.
Advertisement
Check out http://www.talkingpointsmemo.com/talk/blogs/articleman/2008/10/obama-effect-update-obama-is-g.php
November 3, 2008 1:53 PM | Reply | Permalink
Hey Fly. As you may have seen, I've written about GA a bunch of times, and I think Obama's ahead in Georgia by a significant amount, I think more like 4-5%.
The internals in Georgia suggested pre-September 15 that Obama was winning just up to 24-25% of the white vote. Since then, he's hit 28%. Whites are most of these samples, and the subsamples are not too unreliable. These internal numbers certainly replicate across polls, and just as importantly, the little bump up coincides with Obama's national surge and make sense.
The internals I've seen also suggest that Obama is winning 98% of blacks (or more) in the NC, SC, GA, AL, MS arc. This too makes a lot of sense.
Thus, I quarrel with people who think that prediction here is guessing but that the polls out there are less of a guess. Any poll using 29% or even 30% is kinda dumb, IMHO. What we have seen -- and I think this is the critical point -- is the share of black votes within the whole of Georgia's early voting falling, and then essentially leveling. After 500,000, it was around 39%. Above 1MM, it was about 36%. Now it's 34.9% after 2MM votes. I predicted 32-33% before this started.
Let's assume that 2.25MM more votes are cast, and that they are only 29.2% black, the exact percentage of registered votes. Newly registered voters are more likely to vote statistically, and the new voters in Georgia are not even half white, so my suggestion is very conservative.
That would leave a pie of 4.25MM votes cast (about 78% of the total of registered voters), and the black share of those votes would be about 31.8%. In that scenario, Obama wins, unless the internals to which I refer (generally from Rasmussen, which has a tiny house lean to the red) are meaningfully off.
One of the interesting phenomena is the understatement of black support for Obama in prevoting polls. Without getting too nutty in analyzing it, any time a group of people is polled about something and the group actually is going to vote 99% of the time one way, anything that is in the nature of noise is going to be a downward departure. That's what I see when people model less than 95% of the black vote in for Obama, and it happens in many polls.
The two reasons people have had Georgia wrong is the silliness of predicting that blacks will vote exactly their 29.2% proportion in the GA voting pool (or wow, one percent more), so they're repredicting the Kerry election in a way, and the noise that says ten percent of that vote is up for grabs or with McCain, as some of the polls in the 538 average do.
Nate has GA at 5% likely for Obama. That seems facially incorrect. Thanks for all your excellent work here in your return to the Cafe. A
November 3, 2008 3:06 PM | Reply | Permalink
Thanks for your reply, Articleman.
Let me take a stab at justifying my numbers. We have two polls out today - SurveyUSA puts the white vote at 72-25%, and PPP at 70-28%. In the last round of polling, Rasmussen put it at 73-26% and R2K at 66-27%. So that establishes a decent range; I'd say that 28% is at the high end. It's possible, but I'm not counting on it. Similarly, while some polls have put Obama's support among black voters in Georgia in the high 90s, most have not. The level of black support in these polls is almost certainly too low, but I hesitate to put it as high as 98%.
But the more important point is that while black voters may account for 29.2% of overall registered voters, the fact that so many have already flocked to the polls means that they account for just 26.6% of registered voters who have not yet cast ballots. Bear in mind that the most committed and enthusiastic black voters - more than 705k of them - have already voted. The remaining voters are, by definition, a less engaged population. I'll be delighted if they vote in proportion to their registration. As I've written before, that's significant - black turnout is no different for white turnout, once you adjust for socioeconomic factors like income and education. In other words, if they comprise 26.6% of the remaining vote, it will mean that they have voted at levels far higher than their income or education levels would have predicted.
Could they exceed that level? Could Obama capture 28% or more of the white vote? Yes, on both counts. But from where I sit, those circumstances fall within the realm of the possible, not the nearly certain. I know that polls are underestimating black voters, I hope that black turnout on election day and white support will exceed the polls as well. And that leaves this race to close to call.
November 3, 2008 3:26 PM | Reply | Permalink
The remaining voters are, by definition, a less engaged population.
"By definition"--you sure about that? I'm reading anecdotals about a lot of black folks waiting to vote tomorrow because it feels more like being part of the historic moment. That seems more committed rather than less.
/quibble
November 3, 2008 9:54 PM | Reply | Permalink
I was just going to make a similar point. I think many people want the symbol of that date. I've talked to lots of people who decided to vote on the day. I would have done the same, except I'm volunteering tomorrow.
November 3, 2008 10:13 PM | Reply | Permalink
DrBB, TheraP, Ben, A, Storm, et. al.:
Thanks for your responses.
I'm sure that voters who purposefully wait until election day to cast their ballots are a non-zero set. Some of them are going to be just as committed as any early voter. Then there are those who waited in the hopes of avoiding the early voting lines; or whose work or family obligations prevented them from voting until now. But they're not the only folks who haven't yet cast their ballots. Turnout in the African-American community is not going to hit 100%, in Georgia or anywhere else. To the extent there are non-voters, they must be in the pool of those who haven't yet cast their ballots. To the extent that there are marginal voters, voters who can be deterred by lines or intimidation, voters who intend to vote but don't get around to it - they, too, are in this latter pool.
Here's one way to look at it. There are 6.5 million adults eligible to vote in Georgia. Of them, 5.76 million are registered, and 2 million have already cast their ballots. Now back in 2004, fully 77% of registered voters cast their ballots. Let's be bold, and forecast a remarkable and unprecedented 85% turnout in Georgia, largely driven by record black turnout. That would require another 2.87 million votes to be cast today. And let's say that black turnout is just off the charts, hitting 90% of registered voters. In that optimal scenario, black voters would account for 29% of those who vote today, and closer to 32% of the overall vote. And yes, they'd carry the state for Obama.
(These numbers also underline where I part ways with Articleman. When he projects 78% turnout, with African Americans comprising 31.8% of voters, he's implicitly suggesting that everyone else will turn out at the rate of 72% of registrants. But four years ago, the rate among non-black voters was 79%! I have trouble buying lower voting rates this year than in the past. The reasonable assumption is that voting rates among non-black voters will rival those in 2004 - around 80% - and be slightly higher in the black community.)
The problem is that that's an ideal scenario. Perhaps the stars will be in perfect alignment today. But there are a great many variables in flux, and the margins are very small. That's why I think Georgia hangs in the balance.
November 4, 2008 8:31 AM | Reply | Permalink
Well, in my mostly-white precinct, voting rates were through the roof, mostly due to young people (it's a university town). This might be due to a fear of long lines causing long lines, but it's definitely not only that. I wouldn't be surprised if Charlottesville has a voting rate of 150% of its population per the last census (due to a large number of students declaring they lived in Virginia rather than with their parents). In other words, I think your 80% might still be a bit low.
November 4, 2008 8:49 AM | Reply | Permalink
Ben,
In the specific case of Charlottesville, there were 27,567 active registered voters on the rolls as of November 2. That's a sharp increase from the 21,341 registered four years ago. (As it happens, though, it's still lower than the census estimate of 38,214 residents 18 and over, although it's probably closer to the number of those residents who are eligible to register and vote.)
Here's what's most interesting to me. In 2004, there were 15,450 votes cast in Charlottesville, for a turnout of around 72% - about the same as the turnout (71%) statewide. It wouldn't shock me if Charlottesville again mirrored statewide turnout - which might exceed 80%. But with greatly increased registration, simply following statewide turnout patterns would yield a vastly larger impact.
Nationwide, young voters are registering at much higher rates (witness Charlottesville) and may be voting at marginally higher rates, two phenomena that combine to increase their impact. But the real story with young voters is the shift in preferences. The fact that Democrats went from a 10-point margin in this demographic to a 30-point margin is going to have a much, much bigger impact than the registration gains or turnout increases.
But with black voters, it's a different story. They've always been solidly Democratic, although the margin may increase by 10-20 points this year. But the astonishing thing to witness is that both registration rates and turnout rates have skyrocketed. Normally, you hope for one or the other. (Since expanding the pool means adding marginal voters, many field operations expect that vastly increasing registration will actually decrease turnout percentages, even though they'll register a net gain in votes.) That's why they're such a big deal this year.
I hope that the media postmortems are sophisticated enough to pick up on this distinction. If Obama carries the under-30 vote by 20-30 points, and it comprises as large a percentage of the total in this year of increased turnout across the board as it has in the past, that will be an astonishing and historic triumph, and set the stage for continued Democratic success. And if he can drive increased turnout on top of increased registration in the black community, that will be a second triumph, even more astonishing than the first. I fear, though, that the youth vote will be measured by the same standards as the black vote, and that what would in any other year be an astonishing feat will be adjudged a disappointment.
November 4, 2008 9:39 AM | Reply | Permalink
Thanks for putting some cold, hard facts to my over-hyped impressions. Still, reality ain't so bad, either, as you point out. :)
I'm always amazed at your ability to distill the facts so cogently, and I'm also always appreciative. Keep up the good work!
November 4, 2008 10:35 AM | Reply | Permalink
This white guy is doing the same thing. I'm looking forward to the camaraderie of the polls tomorrow—my (majority white) precinct will probably go 90-95% to Obama.
November 3, 2008 10:19 PM | Reply | Permalink
First thing tomorrow I'm biking over to the polls, even though I've already voted! I just want to feel the excitement. I don't have to volunteer till 11, or I might have waited.
Boy was I emotional when they gave me that ballot!
November 3, 2008 10:27 PM | Reply | Permalink
from my limited data i think 98% is too high. Black right wing evangelicals are, inexplicably maybe, a non-empty set that won't vote blue.
November 3, 2008 3:43 PM | Reply | Permalink
There are many internals for Georgia, North Carolina, and Mississippi that, collectively, show the black vote at 98-100.
November 3, 2008 5:11 PM | Reply | Permalink
Don't forget that choices are not just made on one dimension. Those folks need to weigh their religious leanings against the historic nature of this election as well as other factors, like health care and so on. A few people are one-issue voters, but for many, many black people this election is so crucial to their identity... to a sense of making history and a sense of community that it may trump what they might have done otherwise.
November 3, 2008 10:29 PM | Reply | Permalink
Sure, I am happy to disregard them. We will have real election returns shortly enough. Here's looking forward to a Martin upset.
November 3, 2008 3:35 PM | Reply | Permalink
For what it's worth, Greg, Missouri is one of those rare states in which black voters both register and vote at higher rates than other ethnic groups. In 2004, they comprised 11.5% of voters. I expect that to be up this year, and if the race is as close as the polls presently show, it could be the critical factor.
November 3, 2008 3:42 PM | Reply | Permalink
If this DailyKos diarist, Devilstower, is correct, St. Louis voters always face major obstacles on election day, and tomorrow should be no different.
All this compounded with no early voting although I have read of record absentee voting this year and also of the Obama campaign's army of lawyers to be there tomorrow.
November 3, 2008 4:38 PM | Reply | Permalink
The part I don't like and don't trust is the policy of no early voting. Seems almost rigged to go bad, if you know what I mean. The TPMMuck site had WI as another state where the repubs are also getting in gear for similar obstructionist behaviors.
November 3, 2008 6:43 PM | Reply | Permalink
That's not happening now. They threatened it. But a judge ruled they couldn't do it. And they've now backed off - in WI.
November 3, 2008 10:18 PM | Reply | Permalink
All of what both you and articleman have said makes sense to me. Recall that Obama almost always exceeded polling in southern state primaries -- often by substantial amounts. I think we'll see the same tomorrow night. I would be no more surprised by an Obama victory in Georgia than I would an Obama victory in Ohio.
November 3, 2008 4:15 PM | Reply | Permalink
Yes, indeed, GA is quite Peachy. I'm starting to worry about VA. I cannot believe it but McLame is running this advert tying Obama with Wright, today, the last day before voting. Asshat!!!! In the last hour, I've seen that advert 5 times. Shit.
November 3, 2008 9:40 PM | Reply | Permalink
Fly, maybe you'd like to weigh in on the exit-polling and how nonsensical it is to do it at all this year, when there's so much early voting.
Indeed, how has all the polling "during voting" been affecting the outcome itself? But especially, how can people ever really draw good conclusions based on skewed samples from self-selected Nov 4 voters? I'm not referring to the final outcome, because we have the vote for that. I'm referring to all the parsing of demographics and reasons for choosing a candidate etc. that are generally discussed using exit polls.
I know Nate at 538 is going to be discussing this, but I'm concerned that lots of folks may misrepresent the exit poll data. Particularly the repubs who have higher percentages voting on the 4th itself... and thus their numbers will factor into the exit polling to a greater degree than the total vote would warrant. (I can just see them wanting to use this for nefarious purposes....)
November 3, 2008 10:25 PM | Reply | Permalink
Speaking of early voting, I just read an article in the Charlotte Observer that said they have about 2.5 million early voters so far, which is something like 41% of registered voters in that state...
Obama will take about a 10-12% lead going into tomorrow's election. That's a hefty amount for McCain to overcome.
I can't believe I'm saying that about North Carolina. What a year.
http://www.charlotteobserver.com/politics/story/296324.html
November 3, 2008 10:49 PM | Reply | Permalink
The final number was closer to 2.6 million - in any case, astonishing. But the size of Obama's lead among this pool of voters is a great unknown. PPP, for example, puts him up just 5 points - 52-47% among early voters. And that's in a poll with an almost-certainly-too-high estimate of 30% of voters being black. (Bear in mind, they're just 22% of registrants and 26% of early voters.) SurveyUSA, on the other hand, pegs black voters at 20% (almost certainly too low), but gives Obama a 56-41% edge among early voters. Those results are, to say the least,
confounding. And they underline an essential point - we're dealing with small sample sizes and volatile populations, so precision is not a simple task.
South of the Mason-Dixon line, party registration is a funny thing. Millions of voters remain on the books as Democrats, and may even vote that way in state and local elections, but haven't cast a presidential ballot for a Democrat in decades. So Obama's 20-point party registration edge in early balloting is an unknown quantity.
The bottom line? Thanks to strong Democratic and black turnout in early voting, North Carolina is too close to call. For anything more definite than that, we'll have to wait for the returns.
November 4, 2008 8:57 AM | Reply | Permalink
Does non-empty mean full?
November 4, 2008 2:21 AM | Reply | Permalink
does non-full mean empty?
November 4, 2008 10:10 AM | Reply | Permalink