As always happens at the end of a high-profile election cycle, the world is a-Twitter (heh) with polls these days. Those of us who concern ourselves with such things breathlessly watch for each new bloc of data, sifting the crosstabs and sampling characteristics in attempts to read which way the trends are going, both nationally and state-by-state.
This post isn't really about that. If you want good information on the state of the Presidential election, I commend you to Nate Silver at the New York Times and The Princeton Election Consortium as places for a well-reasoned look.
This is more about the limits of polling as an election predictor. Polls are useful tools, but there are people they can't reach. They have to make guesses about who will vote and who won't, and sometimes those guesses are wrong. Disparities between the results of different polls are usually a combination of simple margin error (the luck of the draw relating to the particular people they happened to talk with), plus the differences in those guesses made by the various polling organizations.
Elections can have surprising results when more of one group of people turns out than pollsters anticipated. This can happen because of a unique motivating factor (say, African-Americans turning out for Barack Obama in 2008, or women voting in this election out of concern over reproductive and health care rights), or simply because pollsters guessed incorrectly about the interest level of that voting bloc and the capacity of the campaigns to turn them out to the polls.
In this election, all of these factors devolve to the benefit of Barack Obama. Reasons why below the Frisson de Taupe.
Pollsters need to decide who they think will vote, and who won't. They do this by applying what is called a "likely voter" screen to their data, selecting a subset of all the responses they received that represents an accurate representative sample, as they see it, of the larger electorate. The problem with this is that there are people who do not meet the criteria pollsters set to define a "likely voter" based on their past history, but who may very well turn out to vote. And there are others the pollsters simply have a very hard time reaching.
This is a real thing. Barack Obama's victory in 2008 was characterized by a wave of turnout by people of whom the conventional political wisdom was that they simply don't vote in significant numbers. Likewise, Harry Reid's retention of his Senate seat in 2010—the wave year of the Tea Party, in a purple state—was unforeseen by pollsters like Silver, whose track record in that election was otherwise sterling, and reflected Latino turnout none of the pollsters had anticipated.
Now let's take a look at the Invisible Vote. Who are these people?
-->They're new registrants: pollsters don't have their contact information yet.
-->They're the young, who haven't been old enough to vote for long enough to be seen as "consistent" or "likely" voters.
-->They're those who have only a mobile phone, but no land line—which also tends to be younger voters.
-->They're minorities who may be under-responsive to polling...and whose turnout is consistently underestimated in this election cycle.
Note that all of these constituencies, if they vote, are much more likely to vote for Barack Obama than for Mitt Romney. Romney's voters tend to be older, white, male, social conservatives and/or the affluent. All those groups have a far higher history of turning out, and they easily meet the likely voter screen of pollsters. They aren't off the radar: often, they are treated as the "norm" against which other constituencies are compared.
What this means is that Romney doesn't have a pile of surprise supporters out there. His supporters are being counted in the polls...if anything, they are being overcounted, as the significance of these constituencies as a slice of the overall vote mix is often exaggerated.
Barack Obama, however, has thick veins of gold just waiting out there for him to mine. He did so in 2008 and his ground game is better this time than last. Yes, some enthusiasm has dropped off, especially among the young who thought he was going to deliver them their every dream. But thanks to Facebook and Twitter, everyone knows who Mitt Romney is now, and even if they've soured a bit on the President, they know he is bad news.
The Latino vote will be critical for the President's chances in Nevada and Colorado, and he is currently polling at 3 to 1 over Romney with Hispanic voters. Special effort is being made both by the Obama campaign and by Latino organizations to register Latino citizens to vote, and to ensure that they do. If, as I believe, Latino turnout is higher than pollsters have projected, it will certainly be to Obama's benefit.
Thus far, the Obama campaigns efforts to mobilize what pollsters may consider unlikely voters appears to be working. Early voting in places like Ohio and North Carolina show thus far that higher proportions of these constituencies are turning out than projected; higher numbers, in fact, than in 2008...while every pollster has assumed that minority and youth turnout will be lower this year than it was then.
Looking at the landscape today, Obama is still winning. He weathered Romney's bump, his numbers are rising in the horse-race and, more importantly, extending his margins in most of the battlefield states, while Romney needs to run the table of such states to win. Right-leaning Real Clear Politics (which refuses to acknowledge even Michigan and Pennsylvania as in the President's pile of safe states, when they certainly are) and the bizarre outlier results of Gallup recently notwithstanding, anyone not directly committed to the Romney campaign acknowledges that Obama is in the driver's seat now, particularly after Romney's unPresidential, amateur-hour showing in the foreign policy debate.
There are no grassroots pots of gold for Romney to find to help him win, whereas they are abundant for the President. The Invisible Vote may very well be the factor that tips some states his way, and ushers him to a second term.
Reposted from Green Dragon