Gallup has Obama functionally tied or ahead of Romney in their registered voter samples since mid-October, but Romney has never trailed in their likely voter samples. Today, Gallup has Romney ahead of Obama 51-46 percent amongst likely voters, but tied at 48 percent amongst registered voters.
Pew’s October 4-7 polling, had Romney ahead of Obama 49-46 percent amongst likely voters, but tied at 46 percent amongst registered voters. In Pew’s October 24-28 data, Romney and Obama were tied at 47 percent amongst likely voters, with Obama slightly ahead 47-45 percent amongst registered voters.
In the NBC/WSJ survey from September 26-30, Obama narrowly led Romney 49-46 percent amongst likely voters, but by a 51-44 percent margin amongst registered voters. In their October 17-20th polling, the NBC/WSJ polling showed the race tied at 47 percent amongst likely voters, but Obama ahead 49-44 percent amongst registered voters.
Historically, likely voter sampling has been accurate. But over the last decade, the likely voter methodology has been consistently inaccurate. Likely voter polling tends to undercount Black, Hispanic and single, particularly younger women, voters at the precise time those voting blocs are gaining a larger share of the national electorate.
The minority vote cast a 26 percent share in 2008 and it broke 81-18 percent for Obama over McCain. On October 26, Gallup revealed that their 2012 likely voter sample pegged the minority vote at only 20 percent. If Obama carries minority voters by over a 4-1 margin even if their share doesn’t grow from 2008, as it has over recent elections, Gallup’s likely voter results would be suspect.
Gallup’s 2012 likely voter sample also puts women at 52 percent of the total vote, but 2008 exit polls revealed that women were at a 53 percent share. If the gender gap bounces back toward double digits, that undercount could be significant. Especially since the female majority has grown over the last three presidential cycles (from 8 million to 11.9 million more women than men voting in presidential years).
Gender in conjunction with a racial gap, undermines the validity of likely voter surveys, especially since the female voters undercounted, tend to be younger single women, who are more likely to vote Democratic, than married women.
Since 2002, likely voter surveys have consistently blown the margins of victory in my State of New York. In the 2010 congressional elections, the likely voter samples dramatically undercounted the surging Hispanic vote for Reid in Nevada and Bennett in Colorado, falsely projecting their defeats. In addition, by underestimating the minority vote, likely voter polling in 2010, exaggerated the margin of Republican victories in closer than expected contests for Governor in Ohio and the U.S. Senate races in Illinois and Pennsylvania.
Obama loses if the likely voter surveys are correct in terms of the white vote: Obama has fallen from the 43 percent he received in 2008 to 37 percent in recent polls. But I think these likely voter surveys are undercounting both single women and under 30 voters, so key to Obama crossing his tipping point for victory - 40 percent amongst white voters.
I project a 51-47 percent Obama victory in the popular vote (with 2 percent going to minor party candidates from the libertarian right), provided the president gets high grades for handling the response to Hurricane Sandy and if October’s jobs numbers are in line with September’s better than expected gains.
But unlike the pollsters and their likely voter methodology, I am acknowledging that my prediction is an informed hunch. A hunch based upon overlaying the science of registered voter polling, over long term trendlines on turnout. The likely voter polls could be right and I could be wrong. But something tells me 47 percent becomes a doubly haunting number for Mitt Romney. Numerology anyone?
Gyory is a political consultant with Corning Place Communications in Albany, New York and an adjunct professor of Political Science at the University of Albany.