Mickey Kaus thinks he sees bad news for President Obama in a Gallup poll about gay marriage. Whereas only 13 percent of those polled say Obama's support for gay marriage will make them more likely to vote for him, 26 percent say it will make them less likely to vote for him.
I have a different takeaway from this poll. I think it shows that these kinds of polls aren't worth much.
Do you notice anything odd about the numbers below?
To me what stands out is that 52 percent of Republicans said Obama's support for gay marriage would make them less likely to vote for him. Now, do you really think that 52 percent of Republicans had a greater than zero percent chance of voting for Obama in the first place? Me either. And if the chances of your voting for Obama are zero, how can his position on gay marriage reduce them? Apparently a non-trivial fraction of the people polled aren't really answering the question that was asked.
What question are they answering? My theory is that the question that, when it left the pollster's lips, was "Does this make you less likely to vote for Obama?" enters the respondent's head as "Does this make you like Obama even less than you liked him before?"
In other words: to some extent what this poll measures isn't "likelihood that Obama's new position will change your vote" but "intensity of your reaction to Obama's new position."
If that's true, then an important question is: What kind of people are most likely to have an intensely negative reaction to Obama's new position? I'd say they're by and large people who already have intensely negative feelings about Obama. In other words: the people most likely to say that something Obama did has made them less likely to vote for him will tend to be the people who had roughly no chance of voting for him to begin with. So it isn't just that the 52 percent includes some people who were certain to vote against Obama all along--the 52 percent may have a larger fraction of those people than the other 48 percent does!
I assume that, if Mickey Kaus accepted this (conjectural) argument, he'd point out that it can work in the opposite direction, too: Maybe of the 24 percent of Democrats who said they were now more likely to vote for Obama, most were assured of voting for him anyway, and what they were really saying was: "I feel like celebrating over Obama's new position on gay marriage!"
Yes, maybe so. And, maybe, in similar fashion, both the 11 percent of independents who say they're more likely to vote for Obama and the 23 percent who say they're less likely to vote for Obama consist largely of people whose votes were in fact preordained to begin with. So maybe (I imagine Mickey arguing) even if the poll overstates the positive and negative effects of Obama's position on gay marriage, it overstates them symmetrically; so this issue is still on balance hurting Obama at least somewhat.
Maybe. But would this effect really work symmetrically? In other words, are anti-Obama, anti-gay-marriage voters and pro-Obama, pro-gay-marriage voters equally likely to (unconsciously) "hijack" a poll question like this and use it to express the intensity of their reaction to gay marriage?
My intuition says no. My intuition says that in a case like this, negative emotion, in its search for an outlet, is less discriminating than positive emotion--i.e., is more likely to hijack an unsuspecting poll question that strolls by.
Again, this is just conjecture. But there's definitely something fishy about the results of this poll--that 52 percent number, for starters. And to take the poll as clearly bad news for Obama, as Mickey does, requires assuming that the fishiness is symmetrical. I doubt it is.