Wolraich: Obama at the Gates of... Gates
Dr. C: In Praise of Writing Binges
Maiello: Gatsby Doesn't Grate
For two years, the prehistory of this race was one of Bush-Kerry -- an incumbent during war with a middling economy, a firm base, a modest deficit with independents, a middling flip-flopping opponent with big hair from Massachusetts, and just enough popular desire to stay the course to win a narrow re-election. The race now looks a bit more like Bush-Gore: it is rational to think that the Democrat might win the Electoral College and lose the popular vote, the Republican electoral path is narrower, but the Republican candidate has surprising momentum through the debate cycle and heading toward the wire. I still see President Obama's re-election as likely, but I am downgrading the probability of his winning from 80% to roughly 65%, as a result of the week's events. This piece explains why I still predict the President's re-election, and the reasons my confidence is not higher.
A Big Week For Obama Yields No Gains
With seventeen days left in this Presidential election, there is ample cause for concern in both the Obama and Romney camps. I am still of the view that President Obama is likely to be re-elected, but the probability of his re-election has slipped modestly in the last week, as his clear debate victory over Mitt Romney (barely reported on as such), led to a week of stabilization of the race with a razor-thin advantage for Obama. When I wrote last week that I saw an 80% chance of an Obama win (a very bullish assertion), I rested that in part on my prediction of future events. Specifically, I foresaw President Obama coming out with a great debate performance, which I thought would lead to a 2-3 point move in the national polls, essentially erasing half or so of Mitt Romney's post-Denver gains.
I was half right. The President, derided by his haters as a weak debater, delivered a virtuoso performance, and won the debate handily. While CNN scored it a 7% victory, that poll was skewed in favor of Republican opinion, as CNN sampled debate watchers weighted according to who watched, and not according to likely voters, who are 4% more Democratic and 4% less Republican than their sample. Assuming the Democrats were 3/1 for Obama and Republicans 3/1 claiming Romney won (an assumption borne out, for example, in PPP's Ohio polling released today), CNN's 46-37 was more like a 49-34. Polls of independent voters tended to show the President winning by 15-20 points. Not Denver, but bad for Mitt Romney.
For whatever reason, and there are many interesting reasons you could postulate, the debate this week did not have much effect on the national and state polling. (While beyond the scope of today's piece, here are some hypotheses to chew on: (1) the media didn't really seize on the idea that Obama won despite supportive polling, after they spent ten days after Denver coronating Romney; (2) the primary value of the debate was to see if Romney could appear plausible as President to persuadable voters, and he did; (3) first impressions matter more; (4) Obama by silence in Denver has conceded the validity of some of Romney's more insupportable policy proposals, or has treated the 47% issue as unimportant; or (5) the impression of momentum in polling and media accounts has made Romney seem the winner in media narratives, and while Obama in September was "winning" the battle of seeming ahead, Romney has now overwritten that narrative with a triumphalist narrative of his momentum that is feeding enthusiasm for him.)
The national tracking polls are little affected. Obama leads by 3 instead of 1 in IBD, is down 1 instead of 2 in Rasmussen, while Gallup has gotten too silly to discuss seriously in this piece. The state polls are consistent with a picture in which Obama has a 2 point or so lead, but not more. They have been mixed this week, and it is hard for either camp to claim significant momentum in the handful of still-relevant states. President Obama seems to hold narrow leads in three of these states, Governor Romney leads in two, and two cannot be described as leaning either way.
The President's Base Is 243 EV, Governor Romney's Is 206.
To see where the race is right now, we need to rule out states that aren't really in play. There are 43 of them. In particular, Nevada is a state that has moved modestly further toward President Obama, where he consistently leads by 3 or so in polls. Because Romney has led in one poll all year (a Republican-funded poll), because the numbers are so rigid and consistent, and because Senator Reid in 2010 and candidate Obama in 2008 so far outperformed the polls, I class Nevada as safe for President Obama. In the past I have counted President Obama's base as the Kerry states plus New Mexico. While New Mexico remains solid, New Hampshire and Wisconsin do not, as discussed below. Additionally, while there is no dispute that Mitt Romney has made significant gains in Pennsylvania, he has also failed to compete or spend any money there, and the average of polls show him down there by 4-5. He will not win Pennsylvania, but he may greatly regret not competing there. (As an aside, I agree with Nate Silver that Pennsylvania was as good a place as Wisconsin for Romney to go on offense, given the Republican victory in the 2010 Senate cycle. Why Romney would pour money and organization into Nevada and Colorado, both of which chose Democratic Senators in 2010, but not Pennsylvania, which did not, escapes me.) Thus, President Obama's base consists of the Kerry states plus New Mexico and Nevada, less New Hampshire. This puts President Obama at 243. When one takes the seven states that remain in play -- Colorado, Iowa, Wisconsin, Ohio, Virginia, New Hampshire, and Florida -- the President's base is 243 EVs, while Governor Romney's base is 206 EVs. And yes, that means I do not think the President can win North Carolina.
The State of the Seven Relevant States: Obama's Tenuous Edge
In this section, I class states on a continuum from most likely to vote for President Obama to least. In order, they are: (1) Wisconsin (10 EVs); (2) Iowa (6 EVs); (3) Ohio (18 EVs); (4) Colorado (9 EVs); (5) New Hampshire (4 EVs); (6) Virginia (13 EVs); and (7) Florida (29 EVs). As you can see in looking at this ordering, and also at Obama's base of 243 EVs, the first obvious point is that it is hard for President Obama to win without Ohio, and harder still to win without Ohio or Virginia, though very conceivable. As such, the fact that President Obama leads in Ohio, though very narrowly, bulks large in this analysis and in any assessment that President Obama's re-election is very likely. (If President Obama led meaningfully in Virginia, the odds would be just as favorable, but as discussed below, he does not.)
1. Wisconsin (85% probability of Obama win).
Wisconsin is performing much like it did in 2000 and 2004. In 2000, it went very narrowly for Al Gore in an election Gore won nationally by .5% of the popular vote. Gore won Wisconsin by .2% (5,700 votes), but would have won it handily but for the 3.6% siphoned off by Ralph Nader. In 2004, Wisconsin went for John Kerry by .4% (11,400 votes) in an election Kerry lost nationally by 2.4%. In both cases, Wisconsin was essentially 2-3 points more Democratic than the national trend. It is thus unsurprising that in September, Obama led there by more than 5 but less than 10 points, slightly stronger than nationally, or that Obama has led since Denver in Wisconsin polls by 2, 3, 2, 1, 6, and 2 without trailing once. These polls are highly consistent and suggest a narrow Obama victory. If that holds, Obama will have 253 EVs. It gets trickier from there.
2. Iowa (70% probability of Obama win).
Like its demographically similar Big 10 neighbor, largely rural and even whiter Iowa is performing much like 2000 and 2004, when it was narrowly with Gore, and then narrowly with Bush over Kerry. In both cases, Iowa had a narrowly Democratic lean, almost exactly like Wisconsin's. In 2000, Al Gore won Iowa by .3%, but the closeness was fake, because Nader took 2.2% of the vote, or Gore would have won by roughly 2%. In 2004, Bush beat Kerry by .7%, 1.7% below Bush's national margin of victory. In both cases, one can see a lean of 1.5-2% toward the Democratic ticket. President Obama's 9.6% victory in 2008 was 2.4% above the national trend.
Recent Iowa polling has favored Obama, but has been slightly more inconsistent and equivocal than that in Wisconsin. Since Denver, Obama has been +2, tied, +3, +8, and -1 (PPP). While the average of these numbers is almost exactly the same as Obama's recent Wisconsin polling, one of the results is a Romney lead of one point, and from PPP, a Democratic pollster, at that. Of equal or greater concern to me is that the Des Moines Register poll from the heart of Obama's post convention bounce only showed the President up 49-45. I will place great weight on the next Des Moines Register poll, and expect it to show in a range between Obama +1 and +3.
A final tool at our disposal in analyzing Iowa is reporting on the return of early ballots. As of yesterday, Democrats had pulled early ballots by a margin of 15% more than Republicans (45-30), and had returned them by a margin of 18% more than Republicans (48-30). While the Romney campaign is bragging about matching the Obama ground game, after the 2008 cycle, and in light of the early voting data, it appears that the Obama team is obtaining a small but potentially critical turnout advantage.
Summing these parts, I am fairly concerned about the PPP poll's finding that in Iowa and New Hampshire, and among a majority-female sample, Romney's favorability now exceeds the President's. That is not a statistic to freak out about; it can mean that there are more Democrats who approve of Romney generally but won't vote for him than there are Republicans approving of an incumbent Democratic President. The best thing I can say about the PPP poll is that it averages with other Iowa polls showing an Obama lead (and that if you use PPP in all swing states, Obama wins Ohio, Virginia, Wisconsin, and Colorado narrowly, which is another way of cautioning against cherry-picking this Iowa result and fixating on it unduly).
Whistling through the graveyard here, one would expect Obama to win Iowa narrowly, as Gore did and Kerry almost did despite a 2.4% deficit nationally. This should give Obama 259 EVs.
3. Ohio (65% probability of Obama victory)
If Ohio doesn't make you nervous, you're not paying attention to the data, or not interested in either Obama or Romney winning. There are good reasons here for both to fret. President Obama is in the lead, but that lead is small and appears to be narrowing, without much more room to narrow. Here, Ohio is performing more like it did in 2004 (as a neutral or Democratic-leaning state) and not 2000 or 2008. In 2004, John Kerry lost Ohio by 2.1%, representing a mild (0.3%) overperformance for him in relation to his national loss of 2.4%. In 2000 and 2008, by contrast, Democrats Gore and Obama underperformed their national trends by 4.0% (Gore losing by 3.5%), and 2.6% (Obama winning by 4.6%).
Romney should be very worried, because it is hard to interpret the polls other than showing a narrow Obama lead. In the last week, Obama has led by 3 in SurveyUSA, up 2 from the week previous, has led by 3 from FOX, has led by 1 in Rasmussen (for the fourth poll in a row!) and led in PPP by 1, down from 5 in the week previous, and is tied in Gravis (up 1 from 10 days ago). Ohio is very heavily polled, and the weight of the polling is a narrow Obama lead. Romney should also be worried because PPP and SurveyUSA are both finding that one in five likely voters have voted, and that Obama is rolling up a large plurality (ranging from 58-38 to 76-24 depending on the sample. In a close race, the ground game is likely to make the difference. Obama has also outspent Romney there.
Other trends should worry Obama. His lead is arguably slipping a bit. Romney has narrow edges among independents there, consistent with his national leadership among independents since Denver. Ohio has mirrored the swing among largely white independent voters toward Romney in Iowa and New Hampshire, but with what seems to be a bit of a harder floor for Obama in that white vote. From those trends, one could see the data to this point consistent with Romney eking out a victory in a virtual tie, or Obama holding on by up to 3. Barring a further late shift against him nationally, Obama is the narrow favorite in Ohio, and thus to win the Presidency again. This would give Obama 278 EVs, and would allow him (in combination with winning any other state below) to lose Wisconsin to Romney.
4. Colorado (60% probability of Obama victory)
I am slightly more bullish on Colorado than is Nate Silver. Nate views it as essentially a jump ball, while I think Obama has the edge. Obama leads among independents there per PPP, and has led among PPP, Gravis, and Rasmussen since the Denver debate. Other than ARG, Romney has not pulled better than a +1 in three weeks. But my stronger feeling about Colorado rests upon the 2010 swing and miss by all pollsters in Nevada and Colorado (overpredicting Republican vote share). Obama led handily in most of August and September, the Romney surge has brought Romney to rough parity, and the twin forces of the Obama ground game and poll undersampling of Latino voters makes me think Obama will beat his poll numbers by a point or two. Put another way, if Michael Bennet won Colorado in the 2010 electorate, Barack Obama should win in 17 days.
Colorado, in combination with Iowa and Wisconsin, would give Obama 268 EVs, leaving Obama one state (Ohio, New Hampshire, or Virginia) from winning.
5. New Hampshire (50% probability of Obama victory)
New Hampshire is the key to an Obama victory without Ohio or Virginia. It narrowly tilted Democratic in 2000 and 2004, opting for Bush over Gore but would not have if one reallocated one-third of the Nader vote to Gore, and opting for Kerry when Kerry lost nationally. Combined with running 2 points above margin for Obama in 2008, New Hampshire has been consistently about a +2% state for Democrats, which should make Obama a narrow favorite.
Despite that fact, the last six polls (excluding high and low) amount to a tie -- Rasmussen and Suffolk both showed the race tied, and Rasmussen and PPP have since shown it +1 for Obama and Romney respectively.
Several considerations seem to have caught up to President Obama in the Granite State after he led there most of the year. For one, New Hampshire, unlike Ohio and Wisconsin, is not predominantly white -- it is essentially entirely white. Without a margin of black voters who poll at or above 90-10 for Obama, New Hampshire's polls are swingier. New Hampshire is also legendarily independent in its politics, although you wouldn't know it from the strong continuity among its choices in the last three Presidential elections. President Obama's post-Denver slide with white independent voters has hurt him more in New Hampshire and Iowa, both of which show +1 for Romney in PPP polls.
The bottom line with New Hampshire is that Obama could win without Ohio by stitching together his base (243 EVs) + Wisconsin (10 EVs) + Iowa (6 EVs), Colorado (9 EVs), and New Hampshire (4 EVs). The problem with this kind of thinking is that if Obama is performing well enough with white independents to win New Hampshire, where he appears tied today, he should have hung on in Ohio to begin with. For this reason, the primary path back to the White House for Obama is base + Ohio + almost anything else OR base + Virginia + almost anything else. Which leads us to...
6. Virginia (40% probability of Obama victory)
Polls are wonderful. With their aid, you can see very clearly that President Obama is either ahead or behind in Virginia. If you believe PPP and Quinnipiac, Obama is ahead by 3, 2, or 5 since Denver. If you believe NBC/WSJ/Marist, Rasmussen, or ARG, Romney is up 1, 2, 3, or 1. (For some reason, RealClearPolitics does not list the PPP survey released on 10/20 showing Obama up 2 in its average of polls. If it did, Obama would lead. C'est la vie.) Nate Silver shows the tiniest of advantages to Romney in Virginia. The aforementioned Suffolk declared it would no longer poll Virginia, since Obama could not win it, which is hilariously and illogically triumphalist. At any rate, there is no dispute that Obama led in Virginia all year, and that Romney's Denver debate-driven surge has made the race virtually even.
I tend to think Romney is ahead by a point or two, because the polls showing him ahead are very consistent with each other, Quinnipiac has run high for Obama throughout the fall, and Rasmussen (which shows Romney up here) has shown decent numbers for Obama in Iowa, Ohio, and New Hampshire, not ahead of or behind trends there.
The thing to keep in mind about Virginia is that it gives more reason for hope than New Hampshire. As I explained above, New Hampshire is a state Obama can easily win, but he will probably win it if he has won Ohio and doesn't really need it. Put another way, given its electorate and the last several polls, New Hampshire correlates positively with Obama already winning, while Virginia is likely determined somewhat more independently. Why? For one thing, Obama won it by 6 points in 2008, a slightly stronger performance than in Ohio. For another, Obama has maintained a strong ground game in the state, and the state Democratic Party is turning out voters well for Tim Kaine. Finally, the African-American population and the suburbs of NoVa have treated Obama well. This is a state with a different demography and argument from the bailout-driven, class narrative of Ohio. Obama may well win Virginia and lose Ohio.
If Obama wins Virginia, he has (with his base) 256 EVs, and could win with any of (1) Ohio and nothing else; (2) Wisconsin and either Iowa, Colorado, or New Hampshire; or (3) Colorado and Iowa. Given all of these different combinations, the fact that Virginia runs on a modestly different algorithm from Ohio is good for Obama, creating a second, plausible path to victory, albeit one that requires him to overperform (or shift) polls by 1-2 points.
7. Florida (25% probability of Obama victory)
Good news and bad news. The good news is that with his base of 243 EVs, Obama winning Florida's 29 EVs ends the election right there. The bad news is that Obama has fallen further behind in Florida than in Virginia, making that outcome (though determined somewhat independently of Ohio, as evidenced by Obama's differential slide in the polls there) pretty unlikely.
It is true that Obama pulled a +1 in SurveyUSA in Florida earlier today. Despite that, Romney looks to be ahead by 2 in Florida while Obama leads by 2 in Iowa, Wisconsin, and Ohio. Florida thus almost certainly cannot be the tipping point state (Nate has it ninth). It is nice to see Nate in his October 20 piece accepting the thought he rejected a week ago but which I stated here then, that Romney has gained more in Florida than other states for some reason. The race is close enough that if Florida polls are undersampling Latino or black voters, Obama could still win the state, but the reality is that if Obama wins Florida, he will have won a combination of the foregoing states rendering it irrelevant. Florida is the state that would allow Obama to claim a mandate by getting around 330 EVs on a night toward the higher end of his possible performances.
Let us not speak of it again.
Conclusion: Obama Still Has More Paths Than Does Romney, and a 1-1.5% Margin of Popular Vote to Burn and Still Win
There are three paths to an Obama victory: Ohio and almost anything; Virginia and almost anything; and the base + Iowa, Colorado, Wisconsin, and New Hampshire. Of these three paths, the first seems more likely than not (60% likely); the second is plausible and modestly differently determined, providing a further chance for the President, and the third path is highly improbable. The probability of Obama successfully traversing any of these three paths (and of an Obama win in Florida) will vary based upon trends in the overall numbers. The reason Obama is the favorite is that the numbers very modestly favor him in Iowa and Wisconsin, and he has the better case for Ohio, and a jump ball in Virginia. Romney cannot split these; he almost certainly needs them both.
Accordingly, Romney probably needs another 1-2 point move in the national polls before he can be classed as a favorite. And for that reason as well, I believe that the over-under in the national popular vote representing the vote-share at which Obama wins is actually 1-1.5% below Romney's. That is, based on the polling in these seven states and where they stand in relation to the national trend, Romney has to win the popular vote by 1-1.5%, or he will lose anyway. Mark it down.