Friday, October 26, 2012

They Should Spin. We Shouldn't Listen

Greg has a nice item out this afternoon about what the campaigns are claiming about their internal polls, which reminds me of something I forgot to include, although I've said it before, in my item the other day about how I'm following the polling: I'm absolutely, completely ignoring any claims about internal polls. This is really basic: anything they say (or anything that they leak) is going to be self-serving. And on top of that, there's no magic here; there's no reason for us to believe that the campaigns know more about the horse race from their internal polling than we know from public polls. 

What's more, most internal campaign polling isn't really about the head-to-head, anyway; it's about how to deploy messages, and what groups to pitch those messages to. Yes, presidential campaigns (and to a lesser extent statewide campaigns) do have to make decisions about where to move resources. But that's generally not the big decision they have to make. If that's all there was, I'm not even sure that presidential candidates would find it worthwhile to poll at all. 

The flip side of this is that the campaigns absolutely should be spinning all of it as much as they can. Losing campaigns need to keep their supporters fired up; winning ones want to prevent complacency. Neither side wants the other to dominate the information environment, on the polls or anything else. So, yes, you can expect the spin to continue, and even be upset with your side if they aren't doing it. Just don't pay any attention to what they're saying.


  1. Eh.

    Internal polling done by campaigns is generally *much better* than the public polling. Most public polls are horrible quality, and there doesn't seem to be much of an incentive for them to get better. Gallup was off in its 2010 congressional ballot results by 6 points, and yet news organizations are still paying for the Gallup poll in this cycle.

    You only get useful information from the public pollsters doing an RCP/Nate Silver-type average. The hope is that the problems with the polls are idiosyncratic to each poll, so an average will cancel those problems out.

    The issue, as you said, is that internal poll *leaks* are generally self-serving. No one is ever going to leak a result that looks bad for their candidate.

    1. Agreed...mostly. There seem to be public polls and pollsters with good and consistent track records. There's also every reason to think the internal stuff is on a different level of sophistication.

    2. Also, the idea that public polling sucks is sometimes people say a lot, but how often is the polling consensus way off? Is there any empirical and comprehensive reason to think it's bad and/or worse than it used to be?

  2. "..there's no reason for us to believe that the campaigns know more about the horse race from their internal polling than we know from public polls."

    You're the political scientist, and I'm just a political junkie - but that statement frankly shocks me, coming from a professional. In regard to the spinning and strategic leaking of internal polls I can see your points, but to say that billion-dollar *presidential campaigns are feeling around in the dark with the rest of us strains reason.

    You're telling me that the same Obama primary campaign that knew they had the nomination in mid-February because of "the math", the current Obama general campaign that is targeting waves of different groups in a number of different swing states - you're telling me that they are going by an in-house version of the same sporadic data you can get from Scott Rasmussen or the guy from Suffolk that stopped polling Virginia a week ago because it's "obviously a Romney state"?

    That it's just coincidence that both campaigns have (on background) ceded Nevada to Obama despite polling in the +2-4% range? That Romney isn't using some of his $111 million from the last three weeks to make even a halfway serious run at low-minority states like Pennsylvania, Michigan or Minnesota - because firms like Muhlenberg tell them 3-6% is entirely out of reach from a last-minute ad barrage? The campaigns are leaving these entirely-in-their-ability-to-measure decisions to the equivalent of sporadic 600-voter tests?

    In a word, I'm calling Balderdash.

    At this level, with the size of these campaigns? It would be political malpractice.

    1. I dunno. It seems to me that where the internal polling would be stronger is in, like Plain Blogger said, in targeted polling. Like figuring out how a message plays in Southwest Ohio versus Northeast Ohio. I bet they have some interesting polling of specific counties that shows even more detail than public polls do. And I bet they have way more data on things like the registered voter-likely voter gap than we do. But I also bet that the statewide and nationwide polling isn't that much different than what we see in the public polls.

    2. MSNBC story last week says that Americans will spend $2billion on Halloween stuff this year!

  3. OK, for the doubters: how would the internal polls be better?
    Yes, they can ask whatever questions they want. Thus, they would have more variables. That would mean they can figure out who is persuadable, what variables are associated with liking Romney/Obama, or whatever. But we were just talking about vote intention only.

    As to the question of vote choice, does anyone think there's a better one than "if the election were held today, who would you vote for?" I don't think I've seen any others, and if campaigns had come up with one, I would expect it to have bled out.

    Where the campaigns could have better data is on likely voter screens. Now, the polls taken right before the elections in the past are quite accurate (on average), so it wouldn't seem like ther's an improvement, but there could be. However, people are terrible at predicting their own behavior, and exaggerate their own past histories. I read a paper the other day showing that likely voter screens are terrible...a lot of those who think they'll vote don't, and a lot of those who think they won't, do.

    The only thing that they could improve on is methods: better protocols for callbacks, better trained callers, etc. That mostly takes money and time. Money they have; time they don't.

    Anyway, I just don't see whybthey would have better data.

    1. I'm just approaching this from an angle of "if I were a candidate, how much would I want to know these numbers in the days, weeks and months leading up to a huge election?"

      Me, I'd want to know a whole freaking lot.

      Why would my internal information be better than random public polling? I'm going to spend more money on it, and I'm going to insist on the highest standards for sampling, larger than the typical 500-600 voter samples and no-nonsense question formats.

      Could be I'm projecting a higher level of competence on the Obama campaign than you get from the standard national campaign. McCain really didn't have the money to keep up. Hillary had Mark "What do you mean there's a primary AND a caucus in Texas???" Penn as her chief strategist and pollster.

      But I don't see Axelrod and Plouffe as the sort of folks who would miss a trick, I say they want to be on top as humanly possible of any meaningful variable that can measured, and no-kidding accurate state polling has to be in your top three.

      The last thing I'd want on November 7th is to be asking woulda-coulda-shoulda questions about which state I might have better focused on in the waning days of an election that hinges on a single state.

      Money they have, volunteers they also have. (I saw a story last month that Obama was spending twice what Romney was for polling, in addition to the large disparity in field office/turnout investments.)


Note: Only a member of this blog may post a comment.

Who links to my website?