Most notable, however, is that only in Minnesota is the President above 50%. In Michigan and Pennsylvania the President's lead is four points or more, but his number remains below 50%.
50% matters to an incumbent, because every election is first and foremost a referendum on the performance of the incumbent. It might have been Mitt Romney's strategy to make the race a referendum on Barack Obama, but it was not a new concept. And 50% matters to an incumbent because historical data suggests that voters who remain undecided voters through a campaign are not likely to support an incumbent in the end. This seems to be a fairly rational result when one thinks about it. If the incumbent had been able to close the sale, or if the challenger had made a persuasive case, a voter would not be undecided.
This historical tendency is important in assessing the state of the presidential race, and it suggests not only that the President's lead in swing states may be ephemeral, but that other states where the President is not at or over 50%, such as Pennsylvania or Michigan, may in fact be up in the air.
The unknowns as we head down to the wire are significant. If the ground game is not accurately taken into account, as Axelrod suggests, no doubt polling firms are struggling to account for the air war as well. The magnitude of media spending over the last weeks of the campaign is unprecedented, as we never seen the volume of spending and the magnitude of end game media that is underway in swing states.
A week ago, both campaigns turned up the heat on the abortion issue. Data from a recent Gallup survey illustrated why, as it indicated that women in swing states list abortion as their most important voting issue, at double the importance of jobs or the economy. Clearly the Obama and Romney campaigns are seeing similar results in their data, as unlike any campaign in memory, presidential campaign ads are making their case specifically on the abortion issue.
Abortion is an interesting and historically asymmetrical issue in political campaigns. Over the past two decades, support for a woman's right to choose has split the electorate, with slightly more than half of those polled in support, and slightly less than half opposed. Yet historically, while abortion has been a highly partisan issue, Pew Research data indicates that as much as 73% of pro-life voters rank abortion as a very important voting issue, compared to 22% of pro-choice voters. That is to say that anti-choice voters are more than three times as likely to have that issue drive their vote.
This is not news for Republican activists, who have build the modern Republican strategy as a coalition of single issue voting groups, including anti-abortion, pro-gun, anti-tax, pro-faith, anti-gay voters. The power of this coalition, conceived and driven by Grover Norquist, is demonstrated by the effective excommunication of the pro-choice wing of the Republican Party. Norquist has built effective control of the Republican Party over the past two decades by understanding that what people say they care about matters far less in the political world than understanding those things that will predictably and reliably move peoples vote.
For those undecided women voters, both campaigns are making their final pitch on the basis of the economy vs. women's reproductive rights. Notwithstanding the Gallup data, the Obama campaign is fighting two historical trends when undecided women enter the voting both. They are hoping that more than 22% of pro-choice, undecided voters will vote on that issue, and they are hoping that undecided voters will not trend away from the incumbent.
Somehow, through an opaque alchemy of Monte Carlo simulations, New York Times polling guru Nate Silver considers all of this. He considers the ground game and the air war, factors in the incidence of cellphone usage, and the voting trends of undecided voters, when he projects that Obama has a 73% chance of winning. Yet all of the data suggests that each of the battleground states are within the margin of error, and in a year with significant unknowns--with the normal uncertainties that we know about combined with new uncertainties that we have never measured before--it is easy to conclude that we might know far less than we think we do.