Wolraich: Obama at the Gates of... Gates
Dr. C: In Praise of Writing Binges
Maiello: Gatsby Doesn't Grate
Before the devastation of Superstorm Sandy put the Presidential campaigning on hold (as of course it should, no matter how close in time the election is), it was clear that President Obama had moved up in the swing states, taking as the inflection point (the point for before/after comparison) October 23, the day after the third Presidential debate. One week after that date, it is easy to illustrate that the race has changed very slightly, but significantly, in the swing states that will decide the election. This blog does that, by taking the polling in each swing state, using October 23 as the break point to compare polling before and after, and also comparing the Obama/Romney matchup in "apples to apples" comparisons in which a given pollster (say, ARG) polled the same state at least once in the week before and after October 23.
Swing States, Ordered By Greatest Change To Least Over October 23 Breakpoint
Virginia (Obama +1.7 swing)
October 16-22 Polling, With Obama Margins: -3, -2, 2, -2, 1 [5 polls, average Obama -0.8]
October 23-29 Polling, With Obama Margins: 5, -2, -2, Tie, Tie, 4 (6 polls, average Obama +0.9]
Net change: Obama +1.7
Apples to apples: Rasmussen +1 (comparing a poll from the October 16-22 period to an ensuing poll from the October 23-29 period); PPP +3 [A-to-A average, Obama +2.0]
Virginia Analysis: Both the averages and the apples-to-apples comparisons make a good case: it is hard to argue other than that Obama has improved a bit in Virginia. I would not place undue weight on the PPP apples-to-apples, as PPP found that almost everywhere it looked, and the margin is high. The bottom line is that the average of the most recent polls is an Obama lead. This would, of course, almost certainly give Obama the election. Very hard to guess as to how Sandy will affect things in the Commonwealth. Obama would rather have his people making last minute voter contacts, and they can't. Romney would rather not cede the stage to a President trying to help in a disaster in a nonpartisan fashion, partnering notably with Governor Christie, just upshore.
North Carolina (Obama +1.68 -- but using longer prior period for comparison)
October 8-21 Polling, with Obama Margin: 3, -1, -1, -6, -4, -9, -3, -2, -6 (7/-4.14)
October 23-29 Polling, with Obama Margin: -1, 3, Tie, Tie, -5, -6, -8 (7/-2.42)
Apples to apples: Grove (Even), Ras (Even), Gravis (+1), PPP (+2) [average of Obama +0.75]
Net change: Obama +1.68
North Carolina Analysis: Here, I backed out the data in the set of earlier polls 8 days earlier, given that they were almost identical to the October 16-22 data, and give us more apples-to-apples comparisons. North Carolina remains the longest shot of these eight states to go for President Obama (Nate says 19% likely), but the three great polls for Obama in the last week -- Grove showing Obama +3, Elon and PPP showing ties, and Republican pollster Civitas showing a Romney +1, and then Romney +5 (SUSA), +6 (Rasmussen), and +8 (Gravis Marketing) -- make Obama's current week in North Carolina, his weakest swing state, roughly equivalent to the best week Romney has ever had in Ohio. Obama is not likely to win North Carolina, despite an avalanche of early voting, in which Democrats have a large advantage (though not quite a 2008 advantage). But it remains in play, with four good polls in one week and the early voting. Not a bad eighth option for the President. And it corroborates measurable improvement by President Obama in all eight swing states, in poll averages rich with Republican-leaning sources like Rasmussen, Gravis, and Civitas.
Florida (Obama +1.6 swing)
October 16-22 Polling, with Obama Margins: -1, 1, -1, -3, -5, -5 [6 polls, average Obama -2.2]
October 23-29 Polling: -1, -2, 1, Tie, -1 [5 polls, average -0.6 Obama]
Net change: Obama +1.6
Apples to apples: PPP +2, SUSA -1, Ras +3 [average +1.33]
Florida Analysis: It makes sense that Obama would rise more in Florida after winning the last two debates, because he had lost more ground to Romney in the preceding three weeks than he had lost in other states, as I noted two weeks ago, and as Nate Silver agreed thereafter. The goal, of course, is not to be improving, but to have more votes. The specific polling is promising. Of the last five, there is a tie from SurveyUSA, one of the best and most accurate pollsters, and a +1 from modestly blue-leaning PPP, with a -2 from more clearly red-leaning Rasmussen. It is hard to look at that and have a lot of confidence that either President Obama or Governor Romney will win Florida. It is clear that President Obama led in September, as he led in 10 consecutive polls at one point. It is just as clear that Governor Romney led from Denver until around October 23, as he won 14 of 16 polls over that stretch. The swing back to the President, in a state that surprised many with the massive number of new voter registrations, is poised to play a dramatic role in 2012, as a modestly surprising victory there would allow President Obama to lose every other swing state listed in this blog and still win the election. If Joe Scarborough really thinks that the election is a 50% proposition, as he stated when chiding Nate Silver for having read these polls, he probably also thinks that flipping a coin four times gives him a 50% chance of pulling four heads. Shorter analysis for Joe: Florida = coin; Virginia = coin; Ohio = coin that usually comes up tails; North Carolina = coin that usually comes up heads. Governor Romney needs four heads, and still might not win.
Iowa (Obama +1.0 swing)
October 16-22 Polling, with Obama Margins: 8, -1, 1, Tie [4 polls, average Obama +2.0]
October 23-29 Polling, with Obama Margins: 4, 2 [2 polls/ average Obama +3.0]
Net change: Obama +1.0
Apples to apples: Obama +2 (derived by comparing 10/24 PPP +2 to average of +1, -1 from PPP polls on October 19)
Iowa Analysis: If you take the PPP numbers out of recent Iowa polling, you see Obama with 8, Tie, and 4, for an average of +4.0. I think the actual number is lower, and closer, but it does seem clear that President Obama leads, and has banked a large advantage in early voting. The endorsement of the Des Moines Register is excellent news for Governor Romney, but this is a state that was slow to embrace him, as Rick Santorum won the Iowa caucus with a low 25% vote share. The Democratic strength and culture in Iowa is old Democrat -- Mondale, Dukakis, unions, and Tom Harkin-style liberalism -- and less suburban Bill Clinton-style New Democrats. The state has been moving slowly to the right, and still favors Democratic Presidential candidates by 1.5% to 2.0%. The Iowa Supreme Court's decision legalizing gay marriage led to the recall of three of the four Justices who penned it; the fourth is up on the same ballot this year, and his re-election is thought to be favored, albeit narrowly. Iowa seems to be with the President today, but also seems like one of the most logical Republican pickup targets on this list after Florida and North Carolina.
Colorado (Obama +0.5 swing)
October 16-22 Polling, with Obama Margins: 2, 5, -4 [3 polls, average Obama +1.0]
October 23-29 Polling, with Obama Margins: 3, 3, tie, 3, 1, -1 [6 polls, average Obama +1.5]
Net change: Obama +0.5
Apples to apples: PPP (-2) [Note, ARG showed a +3 from Romney's Denver bounce that predated the 10/16 window]
Colorado Analysis: President Obama led in almost every, but not every, poll of Colorado through the end of September. Governor Romney then went on a run from October 2 through October 21 in which he led 7 out of 12 polls of Colorado -- moving from a significant deficit (consistently 3-5 points in favor of Obama), into at least a tie. Heading toward November 6, the last six polls show five narrow Obama leads, a tie, and a narrow Romney lead, which (combined with President Obama winning the state by 9 in 2008) makes sensible Nate Silver's estimation that an Obama victory here is 55% likely. Colorado seems to mimic somewhat closely (though with numbers a bit more muted) the moves in the national polls indicated by Pew's three surveys, which showed Obama +8, Romney +4, and tied now. As I have mentioned in every piece, and will continue to mention in every piece, public polling in 2010 consistently understated the standing of Colorado's Democratic Senate candidate, Michael Bennet. Polling that shows the race +1.5 for President Obama does not read like a jump ball to me for that reason. I would add that among the last six polls of Colorado, the only one showing a Romney lead (and that of 1 point) is an ARG survey that compares favorably to a Romney +4 ARG survey of Colorado released on October 8. Thus, while the only apples-to-apples data among the two week samples is consistent with President Obama falling back 2, there is also a 3 point gain from the peak of Governor Romney's Denver bounce in comparing ARG. Given the narrowness (0.5%) of motion in Colorado, that the meager apples-to-apples data are not really corroborative is no big deal, though it stands in contrast to each other swing state, suggesting in turn the validity of this piece's overall thrust.
Ohio: (Obama +0.18 swing -- unless you draw the cut line after October 23, which makes it +0.55%)
October 16-22 Polling, with Obama Margins: Obama 1, 3, 5, Tie, 1, Tie, 3, Tie, Tie, 2, 5, Tie [12 polls, average Obama +1.67]
October 23-29 Polling, with Obama Margins: Tie, 2, 4, 2, 4, 1, -2, 3 [8, average Obama +1.85]
Net change Obama +0.18. (Note, however, that moving inflection point to draw line after Rasmussen October 23 poll changes the prior week sample to 13 polls averaging Obama +1.45, while the current week moves to 7 polls averaging Obama +2.0, from which you derive a net change toward Obama of +0.55%. This does not involve grouping older polls with the modestly better current week; it simply draws a line after a poll that would not have fully priced in the evolving public consensus that Obama won the third debate, which Gallup several days later put at 56/32.)
Apples to apples: PPP +3, Rasmussen -2 (or -2.5% if you move the October 23 poll into the sequence drawn on October 16-22), Gravis +1, SUSA no movement. These data are corroborative of the half-point or so move suggested by drawing the cut line after the Rasmussen poll. They reflect either a +0.5% or a 0.375% movement to President Obama, given that whether you move the Rasmussen poll into the earlier group affects in turn how much Rasmussen is claiming the race moved from the earlier to the later period.
Ohio Analysis: The tortured attempt by hacks to spin these data into a Romney surge meme is exasperating. Today's left bar on "RealClear" Politics claimed that "Latest Polls" showed Ohio even. To reach that conclusion, RCP cited an article from Politico that was written as if the last two *released* public polls were the most current, to make the argument that in the two most current polls, Ohio was tied (in The Ohio Poll, drawn among October 18-23, but released on October 28, and then in an October 29 Rasmussen release showing Romney up 2). The obvious problem with that is that The Ohio Poll is from *last week*, a week in which there were several polls showing a tie, and that in the next six polls, President Obama led. Among those six polls were an Obama +1 from Gravis Marketing -- which along with Rasmussen, is responsible for almost every Romney lead ever polled in Ohio -- and a +2 from Purple Marketing, which also found a rare Romney +2 in August in Ohio.
Romney is not behind by much in Ohio. He has run well, considering his Let Detroit Go Bankrupt baggage, and President Obama's Ohio-friendly auto bailout decision. But how a reasonable person can look at the towering blue column, scarcely interrupted by red in either RCP's or Nate Silver's listing of these data, and see a race that is tied, is beyond me. Ohio has been campaigned to death. President Obama's and Governor Romney's campaigns have spent over $100 million in ads there alone (an amount in excess of the traditional $83 million limit for one candidate's entire national campaign). It is thus probably unsurprising that the polls of Ohio show a remarkably narrow range of variation. (If you consider Iowa, where Obama +8 has been followed by Obama -1, or New Hampshire, where Obama +9 has been followed by Obama -1, you quickly see what I mean. North Carolina shows comparable variability. Colorado and Florida show less than these states, but more than Ohio.) The great consistency of the results mean that it is more likely that they are valid. While I am not going to calculate standard deviations and confidence intervals, others have. HuffPo's Pollster detailed last week that there was a 96% chance that President Obama was ahead in Ohio. This week, that chance has risen.
Finally, 32% of Ohio has voted (Rasmussen found this, and found a 62/36 Obama lead among those voters). The Romney vote has to overperform on Election Day. Maybe Sandy will somehow aid that. Maybe all the polls are wrong. However, the final Rasmussen 2008 poll before Obama won the state by 4.6% was a tie (49/49), so we are used to Rasmussen missing badly at the end. The RCP average of polls, which included a McCain +2, suggested an Obama lead of 2.5%. Again, the President overperformed all of those numbers. It is hard to look at those early voting numbers and think the President is really underperforming almost every public poll, which agree nearly unanimously that he leads by about 2.5%.
Taking one more cut at this, let's look at the last week of Ohio polling and leave Rasmussen's older poll in. That means Obama leads by 2 (ARG), 4 (CNN/Opinion Research), 3 (SurveyUSA), 4 (PPP), 1 (Gravis), 2 (Purple Strategies) -- with Rasmussen's tie and Romney +2 as the only polls not showing an Obama lead. As between everyone else, and Scott Rasmussen (who has been publishing a daily swing state tracking poll showing Romney margins of +4 and +6 when the population-weighted averages of the recent Rasmussen individual state polls of those very states -- and the average of everyone else's polls of these states -- ran significantly better for President Obama), I choose everyone else. The President is up 2 or a bit more among Ohio likely voters, and one-third of the votes have been cast. Period.
Wisconsin (Obama +0.15)
October 16-22 Polling, with Obama Margins: 2, 6, 2, 3, 5 [5 polls, Obama +3.6%]
October 23-29 Polling, with Obama Margins: 6, 5, tie [3 polls, Obama +3.75%]
Net change: Obama +0.15%
Apples to apples: Grove +2, Rasmussen -2
Wisconsin Analysis: We have heard a lot this week from Governor Romney's campaign and its allies in the media (and a guy who commented here about Minnesota) about how Pennsylvania, Minnesota, Michigan, and of course Wisconsin, are all in play. President Obama led in 25 straight polls of Wisconsin, and won the state in 2008 by 13.9%. Governor Romney got a nice bounce out of selecting a Wisconsinite as a running mate, but then there were those 25 straight polls, and still none since mid-August showing a Romney lead. As a final note, the Rasmussen poll suggesting a tie in Wisconsin does so in part because it found over one-quarter of Wisconsin's African-American voters improbably lined up behind Governor Romney. Reallocated plausibly, the poll would have shown a 1-2 point Obama lead, forming a wall of now 27 straight Obama leads. President Obama leads here, albeit fairly narrowly. And when seven of eight swing states shows at least a tiny trend toward Obama from week to week, we can see a small but discernible move toward him.
New Hampshire (Romney +0.75) (Or Obama +1.26, With Cut at End of October 23)
October 16-22 Polling, with Obama Margins: -1, 9, 3, -2 [4 polls, Obama +2.25]
October 23-29 Polling, with Obama Margins: -2, 3, 3, 2 [4 polls, Obama +1.5]
Net change: Romney +0.75 (unless we change cut line by one day, in which case it's Obama +1.26)
Apples to apples: PPP +3
New Hampshire Analysis: There are fewer data points in New Hampshire, and it is thus easier to make arguments either way about them. New Hampshire is the only state where you see regression in Obama's numbers after the October 23 inflection point. That is, unless, like we did with Ohio, we take a poll on the day after the third Presidential debate and group it with the polls conducted right before it instead of the polls conducted right after it. If we take the Rasmussen October 23 poll and consider it with the earlier data (on the theory that President Obama's debate win was not yet fully priced into the polls), you see seven polls from October 16-23, with an average of Obama +1.4, and then three in the next week, with an average of Obama +2.66. Again, as in the other swing states, there's an obvious cut line after which things turned up for the President. Without getting into the why of it, the President has improved in all eight swing states, with New Hampshire being the most equivocal data set -- and it's a data set in which the President led all three polls this week. If it comes to holding both Iowa and New Hampshire (that's two different events whose probabilities must be compounded, Joe Scarborough), President Obama will have something like a 35% chance of winning both, which (joined with Nevada, Wisconsin, and Colorado), would allow him to lose all four of the biggest swing states listed in this piece -- Florida, Ohio, North Carolina, Virginia -- and still reclaim the Presidency.
Conclusion: Race Frozen for Several Days From Moment in Which Obama, Surging, Leads in Six of Eight Swing States
Yes, President Obama leads in 6 of 8 swing states. President Obama appears to lead modestly in Ohio and Wisconsin, with almost no polling suggesting otherwise. The races in Iowa and New Hampshire are lightly polled, but also reflect 2-3 point Obama leads. The race is very close in Colorado and Virginia, where data are equivocal but where the President leads in poll averages over the last week. The President fell further in Florida, but rebounded further, and has now drawn two leads and a tie in current polling, though it is clear that Romney leads the average of polls there, by the same less-than-a-point that marks Obama's advantage in Virginia. Finally, North Carolina finds the President coming back, but is probably for him what Ohio is for Romney -- a close loss that was never going to be a win.
Anyone whose analysis is driven by data rather than by a partisan desire to come out somewhere different (either for Romney to keep his voters motivated, or to see an irresolvable tie, to please readerships drawn from both sides of the partisan divide) will see this. I don't purport to know the percentage likelihood of President Obama's re-election. My guess is probably better than some people's. Well, at least better than Joe Scarborough's. But I do know to a high degree of confidence that the President leads in Ohio, and that one-third of all votes are in. And I know that if he wins Ohio, it is extremely likely that he wins the Presidency. I also know that with the President's modest surge back (predicted here immediately after the third Presidential debate very much as it then happened), there are a lot of other paths available for President Obama to re-election without Ohio. Virginia + Wisconsin + Iowa (all of which the President leads in). Virginia + Wisconsin + Colorado (all of which the President leads in). Florida (where the President has led in two recent polls and has closed to within one point -- the margin by which pre-election polls in 2008 underestimated President Obama's performance at the polls). Colorado + Wisconsin + Iowa + New Hampshire (all of which the President leads in).
Many in the media who are functionally allied with the Romney campaign (Krauthammer, Rubin, the RCP editorial crew) keep pushing the idea that the President is "failing" based upon their very partisan assessment of his campaign and its rhetoric, or that Governor Romney has a lock on "momentum" because of isolated polls here or there (such as the "latest polls" in Ohio that were really one current poll and a poll drawn from a sample 6-11 days ago). This reminds me a great deal of the 2005 baseball season, when my Chicago White Sox had the best record every day of the season, and won 17 of their last 18 (including sweeping the World Series) without having a single player in the top ten in the voting for the 2005 American League Most Valuable Player Award. The Yankees had four players in that top ten, which caused me to remark to my father and other Yankee fans that the White Sox may have won the World Series 4-0, but the Yankees apparently won the Most Valuable Team award. I think Barack Obama, like the White Sox he also roots for, will settle for mere victory in place of accolades from a biased or even-Stevening media establishment that has failed for four weeks to notice or acknowledge his superior performance as a candidate, measured thus far in the only ways that it can be.