12 Days Left: Obama Trails in National Polls, But Likely To Win Electoral College

Today was a strange day of polling, like many others in 2012 -- it showed that Mitt Romney apparently leads in national polling, but is unlikely to win the Presidency.  All campaign long, the national and state polls have been subtly but clearly irreconcilable.  Today was more of the same, as the Obama momentum from the President's debate win Monday seemed to continue in state polls, but not in national polls.  The path to the White House is clarifying further, and it runs through Nevada, Ohio, and Wisconsin.  Here's what today's data set showed.

National Polls Holding Strong For Romney at Roughly +2...

There is no question that the national polling picture remains good for Governor Romney.  Gallup's seven day average shows him +3, though after cresting last week at +6, that suggests that recent days are even or almost even.  Rasmussen moved from Romney +4 to +3, still a large enough margin that, if accurate, would seemingly assure Romney of victory on November 6.  ABC/WaPo moved from +1 Romney to +3.  AP/GfK showed Romney +2, but that poll was from just before the third debate.  Finally, IBD/TIPP showed Obama +2, down from +3 the day before.

But Swing State Polls Generally Cannot Be Reconciled With National Numbers

Meanwhile, in the states comprising the nation in which Mitt Romney apparently led by about 2 points, President Obama remained poised to claim re-election.  In Nevada, Marist put President Obama +3, his eighth consecutive reading at +2 or better.  In Colorado, PPP put President Obama +4 and Marist called it tied.  In Virginia, in three polls taken after the third Presidential debate, both Rasmussen and Fox News put Romney +2, while PPP put Obama +5 (with the variability explained with respect to the white vote, which PPP sees as 41%, while FOX sees it at 34%).  In Wisconsin, PPP put President Obama +6, and in Iowa, PPP put him at +2.  In Pennsylvania, Rasmussen sees Obama as +5.  Even in North Carolina, PPP showed the race tied, where last week Romney led by 2.  Finally, the unreliable Gravis Marketing showed President Obama -1 in Florida -- the sixth poll in the last eight showing a 1 point lead either way (five with Romney +1, one with Obama +1).

Yet all of these numbers (not just the numbers from PPP, the Democratic pollster) fail to square with a Romney +2 national consensus.  Consider that President Obama won last time by 7.2%.  Then consider where these states were last time in relation to that national trend, and apply the same differentiating factor to a supposed Romney +2 national popular vote.  You would get very different results.  Colorado was +1.8 Obama in relation to the 2008 result -- so you would expect the average to put Romney +.2%, where today he was down 2%, which is 2.2% worse.  President Obama ran -6.9% in relation to his 2008 national trend in North Carolina -- and thus should trail by 8.9%, where today he was tied.  President Obama ran -4.4% in Florida in 2008 (winning it by 2.8%), and thus should trail there by 6.4%.  Instead, today's poll from a GOP-leaning robopoller puts him -1, which six recent polls of the last eight there or better.  Virginia was a 6.3% Obama win, thus .9% worse than his national trend, such that the President should trail today by 3, though he either trails by 2 or leads by 5 -- which averages to a .3% lead today, 3.3% above where he should be.  Pennsylvania was a +3.1% state in 2008 (Obama won it by 10.3%, and today it should be +1.1 for Obama if Romney is up 2 nationally -- but Rasmussen shows Obama +5.  President Obama's 6 point advantage in Wisconsin is 5.5% higher than it should be.  Though not the subject of polling today, Ohio, a state President Obama won by 4.6%, a -2.6% performance against trend in 2008, should be a state in which President Obama trails by 4.6% today, adding that margin to a supposed Romney national lead of 2 points -- where yesterday's polls averaged to a 2.7% Obama lead.

Of swing states polled today, only Nevada and Iowa actually square with the supposed national trend.  In Nevada, Obama's 12.5% win in 2008 was +5.3%, which would predict a 3.3% Obama lead if Romney leads by 2 -- and that is what we see.  First, even if two states fit the Romney +2 narrative, this does not undo the fact Colorado, North Carolina, Virginia, Florida, Ohio, and Pennsylvania deviate greatly.  Second, only Iowa really fits -- I am adamant that Obama will overperform the average of Nevada polls, as he did in 2008 and as Sen. Reid did in 2010.

It is also worth noting that two other states polled today fall in either category.  New York polled +26 Obama, which is the same amount Obama won New York by in 2008, meaning the President is over performing a Romney +2 national world by a whopping 9 points.  California, by contrast, polled +12 Obama, which is actually more consistent with a Romney +5 national result.

Thus, in review, today's polls show President Obama overperforming a Romney +2 world by 9 in New York, 9 in North Carolina, 5.5 in Wisconsin, 5.5 in Florida, 4 in Pennsylvania, 3.3 in Virginia, 2 in Colorado, and performing consistent with it in Iowa and Nevada, and underperforming it by 3 in California.

Hypotheses To Explain the State/National Poll Discrepancies

There are several ways of explaining the discrepancies between the national polls suggesting a roughly 2 point Romney lead, and the today's state polls.  Let's consider each in turn.  First, if you throw out the PPP polls, the picture is significantly different.  For example, Colorado is now about where it should be (even).  Virginia is now one point off where it should be, which makes sense.  This means Colorado, Virginia, Iowa, and Nevada make sense, and comport with a Romney +2 world.

But tossing out PPP's data doesn't make much sense, however, for several reasons.  For one thing, PPP has a better track record of accuracy at the state level than does Rasmussen.  There is no reason to simply dump its polls.  For another, PPP is the only pollster to show a Romney lead in Iowa in the last month (+1 last week), and showed a Romney lead in New Hampshire, below a great +8 for Obama from a good pollster (UNH).  Finally, dumping PPP doesn't cure the discrepancies in Pennsylvania, Ohio, Florida, Wisconsin, or North Carolina.

A second explanation is that of a systematic methodological flaw in the national polls.  While beyond the scope of this piece, it is strongly suggested by the consistency among the incorrectness they represent in relation to the state polls.  I have often noted of Rasmussen that it seems to show better numbers for President Obama in individual states than it does nationally, which doesn't make much sense to me.  Gallup's +7 in its seven day roller was so far away from all state polling then in the field as to discredit Gallup's methodology for this cycle.  But the steadiness of the ABC/WaPo numbers, and their being closer to the state numbers, mean that its shift to 50/47 Romney is better evidence that Romney may lead nationally than (to me) are Rasmussen or Gallup.

This in turn suggests the third explanation -- that some of the swing states have swung more Democratic in relation to the national trend at the Presidential level (an Electoral College advantage hypothesis).  It is hard to argue other than that Ohio has -- a state that was +3.5% GOP in 2000, +.4% Democratic in 2004, and +2.6% GOP in 2008, it is now polling +4.0% or more above that Romney +2 national poll consensus.  Florida too is now polling as a Democratic-leaning state, unless you think President Obama leads nationally.

A fourth explanation we can call the California/Illlinois hypothesis.  California in the last two polls has shown a move toward Romney more consistent with a Romney lead of larger than 2 points.  For that reason, given California's vast population, President Obama's national numbers could be explained by reference to his 12 point decline there in relation to his 2008 performance.  Illinois has likewise moved further against the 2008 trend away from Obama, while remaining safely in his column.  Together they comprise 16% of all Americans.

What Is Actually Happening

There is some truth in each of these explanations.  For one thing, PPP does have a modest Democratic lean.  I do not believe that President Obama is up by 5 in Virginia, for example, and doubt that he is tied in North Carolina, though either result seems possible from a sample derived in the glow of Monday's debate.  But PPP has a strong enough track record that its results deserve to be averaged with the more Romney-friendly results which also belie the supposed national Romney lead.  And it is likely true that California could be exerting a modest drag on Obama's national numbers without threatening to give its electoral votes to Romney.

But the core explanation seems to be a mixture of flawed national polling, and changing dynamics in a handful of swing states.  Ohio, after the auto bailout and two years of organizing on Democratic issues such as the repeal of Ohio's antiunion law, is now more like Michigan or Pennsylvania.  It would not be surprising to see Ohio go for Obama by 3, while its northern and eastern neighbors did so by 5 and 6.  Nevada is now a blue state, with a 5-7 point lean above the national trend. Increasing African-American and Latino registration in Florida has put that state closer to the national trend, perhaps sitting right about at even.  An alliance between affluent NoVa and African-Americans there and throughout Virginia has made that state bluer, and it sits roughly at the national trend.  Wisconsin appears to be performing modestly better for President Obama in relation to the supposed national trend, and thus appears close to safe.  Iowa and New Hampshire are no more blue than they once were, and appear to sit on the national trend -- they appear to be jump balls.

It is hard to reconcile where that group of states is with a nation in which Mitt Romney leads by 2 points.  By feel, I believe that the national popular vote is between Romney +1 and Obama +1, and that Obama leads as follows:  Pennsylvania by 6, Nevada by 6, Michigan by 5, Wisconsin by 3, Ohio by 2.  I believe Colorado and Iowa are ties leaning to Obama.  I believe New Hampshire and Virginia are ties leaning to Romney.  I believe Romney leads by 2 in Florida and 3 in North Carolina.

Today's data are fully consistent with Obama holding Ohio, Nevada, and Wisconsin, which would mean that he would win.  From the lay of the polls yesterday and today, I would guesstimate a popular vote tie or that Romney leads by 1 point or less, which means that today's data seem to accord with Obama winning the electoral college with 272-287 EVs, and Romney narrowly winning the popular vote, thanks in part to irrelevant changes in the voting habits of Californians, and perhaps also Illinoisians, as the President has slipped by 12 points to a projected 13 point win there.  Lacking a further Obama surge, and lacking a Romney move in Wisconsin, Nevada, or Ohio (we have seen no such breakthrough thus far), such an outcome is increasingly likely.  The more likely alternative is not a Romney win in defiance of dozens of Nevada or Wisconsin polls; it is more likely an Obama popular vote win that humiliates the national pollsters.  Twelve days left to see which of these odd results comes about.  Interesting stuff.