Today's DailyKos/SEIU/PPP poll has an interesting pair of questions that directly measures what effect Obama's position on marriage equality has had on the election. A whopping 12% say that President Obama's announcement on same-sex marriage changed how they plan to vote for president in November.
Approximately 8% now say they are voting for Romney, 3% for Obama, and 2% Unsure. (Yeah, that adds up to 13% - rounding error.)
Holy Mackerel! Does that mean this is a net -5 pts for Obama???
No. Why? Because there's a follow-up question in the poll, for those who said they changed their vote: How did President Obama's announcement change how you plan to vote?
About 5% said they were planning to vote for Obama and now will vote against him. About 2% said they were planning to against Obama and now will vote for him. The rest said they changed their mind in some other way.
I Do Not Think They Meant What They Said They Meant
But now, when you delve into the murky depths of the crosstabs, things get tricky. (I'm going to switch to talking about numbers of respondents here instead of percents.) In brief: lots of people are fibbing.
50 people said they were voting for Obama and now are voting for Romney. But only 35 of them picked Romney in the first question! Likewise, of the 14 who said the opposite, only 9 picked Obama in Question 1. So basically a third of those who claimed to have switched their vote from Obama to Romney or vice verse were lying or screwing with the poll.
Of the 35 Obama-to-Romney switchers, 20 were Tea Partiers or Conservative Republicans (or both). Fat bloody chance they were going to vote for Obama in the first place. But they may not have been so motivated to vote Romney before.
Of the 35 Obama-to-Romney switchers, 11 do not have a favorable view of Romney. These seem most likely to be the ones who actually changed their opinion based on gay marriage. 7 of them are moderates, 8 are Dems or Indys. On the opposite side, of the 9 who say Romney-to-Obama, 3 have a favorable view of Romney, and seem most likely to have actually changed their vote.
Of the 35 Obama-to-Romney switchers, 2 say they are African American. That's 2 out of 101 total African-American respondents. That's a very small number - yes, large enough to make a difference in a tight election in Virginia or Ohio, but not the-sky-is-falling sort of numbers reporters were suggesting right after Obama made his statement.
The third answer, Switch Some Other Way, has 12 Obama voters (perhaps changing from Not Sure to Obama?), 34 Romney voters, of which 30 have a favorable opinion of Romney (probably didn't really change their mind about anything?), and 8 Not Sure voters, 7 of whom are Conservative or Republican (perhaps changing from Obama to Not Sure?)
Meanwhile, the the overall Obama v Romney number is about the same again as the last two weeks running, and again I don't see any significant changes among any of the demographic groups in the crosstabs. I don't see any significant changes for the question about excitement levels, either.
Adding It All Up - About 1% Net Change
Pulling all the numbers together, I would judge (please note! my judgement call here!) that we have 11 respondents who possibly switched from Obama to Not Sure matching the 12 who possibly switched from Not Sure to Obama, and 4 likely Not Sure to Romney. We also have 11 respondents likely to have switched from Obama to Romney, and 3 likely to have switched from Romney to Obama. That's a net deficit of about dozen for Obama, or about 1% of voters. I think it is reasonable to assume Obama really did lose more than he gained, an amount essentially unmeasurable in individual polls but apparent in the aggregate, and consistent with TPM's polltracker's recent change. Based on pre-announcement polls, I originally said"So there's no indication that the political effect of yesterday's announcement favors one side or the other in the November elections." It appears I was about one percentage point too optimistic.
This poll has some interesting lessons in the behavior of polling respondents. It appears people like to emphasize their opinion of gay marriage by exaggerating or flat out lying about voting behavior in polls. This probably would extend to other 'hot topics' as well, which would be well worth remembering whenever we see a question phrased "Would such-and-such make you more or less likely to vote for Candidate A?"