In the diary, I will examine the implications of the figures reported in the articles quoted below for interpreting polling internals from Gallup's October 9-15 tracking poll (showing Romney up by 52-48). I apologize if someone has already done this, but with today's Gallup poll holding steady at 51-45, people continue to be concerned, despite the preponderance of polls pointing the other way. I figure a little reassurance can't hurt.
From Ron Brownstein [National Journal, 8/27/2012, Obama Needs 80% of Minority Vote to Win 2012 Presidential Election]:
Obama’s strategic equation defines Mitt Romney’s formula: 61/74. Romney’s camp is focused intently on capturing at least 61 percent of white voters. That would provide him a slim national majority—so long as whites constitute at least 74 percent of the vote, as they did last time, and Obama doesn’t improve on his 80 percent showing with minorities.
From Jonathan Chait [New York Magazine, 8/28/2012, Why Is Obama So Confident?]:
The white share of the electorate has been dropping steadily for more than twenty years — from 87 percent in 1992 to 83 percent in 1996 to 81 percent in 2000, 77 percent in 2004, and 74 percent four years ago. Ron Brownstein’s recent reporting suggests that both campaigns expect an electorate that’s about 74 percent white. As pollster Mark Blumenthal has exhaustively shown, Gallup has systematically underweighted the number of minorities in its polls, due to technical issues related to the difficulty of finding and weighting poll respondents.
To the best of my knowledge the most recent internals provided by Gallup are from the October 9-15 tracking poll. You can find them here.
Gallup's figures show Whites voting 61% for Romney and Non-Whites (including Hispanics) voting 16% for Romney.
For simplicity’s sake, imagine the electorate as 10,000 people, and assume that Mr. Brownstein and Mr. Chait are correct about the likely racial breakdown of the 2012 electorate (74% White, 26% Non-White (including Hispanic)).
7400 will be White; Romney will get 61% of them, or 4514 votes.
2600 will be Non-White (including Hispanics); Romney will get 16% of them, or 416 votes.
In total, Romney will receive 4930 votes (49.3% of 10,000).
To reiterate, if the electorate breaks down racially as it did last time around AND Gallup's recently-instituted Likely Voter screen correctly identifies who will actually vote, Romney should have received about 49.3% of the votes in this poll, not 52%. The cross-tabs do not indicate the racial breakdown of voters passing Gallup’s screen, but the group is clearly more than 74% White. I can't imagine that Non-Whites will be less motivated to vote this time around. Are Whites more motivated? If so, why? Am I missing something important here?
Andy Sullivan [Reuters, 10/19/2012, As other polls show tight race, Gallup stands apart] quotes Gallup chief Frank Newport: "We try to keep our eyes on the boat and do the best job possible. We're going over some additional tweaks with our methodologists to make sure we're on top of it."
Which begs the question of the meaning of changing Gallup poll numbers in general? If they are constantly tweaking, do changes in topline numbers have any meaning at all?
As a final note, my daughter is newly registered in Massachusetts and voting in that state for the first time. She is also 100% certain to cast her ballot for Obama on election day. She would very likely not have passed Gallup’s LV screen if she’d been polled before 7 p.m. yesterday, when she opened a letter from the election commission telling her where her polling place is located.
