Wednesday, October 10, 2012

All I Do Sometimes Is Write "A Little Perspective" Posts

I might want to get back into the discussion of whether and how twitter changed debate coverage, but one thing in Richard Just's column on it caught my eye:
[I]f the new polls are accurate, a not insignificant percentage of Americans have switched their votes since last Wednesday—meaning that they, like many journalists, have apparently decided that a few too many “uhs” by Obama on national television should override all the substantive issues facing the country.
I don't know what "not insignificant percentage" means to him, but it's worth pointing out just how few people we're talking about.'s a little hard to do, because the nature of polling makes it really hard to  figure out what changes were a result of the debate and which were not; as of now the HuffPollster polling average has Obama's peak lead back on September 21 at 4.1 points and dropping to 2.3 points by October 1 before the October 3 debate. Perhaps that's an artifact of the way they handle the data; perhaps it shows a real pre-debate change (and part of it is the Gallup shift from registered to likely voters, which has nothing to do with anyone changing their mind). But even using the maximum shift (that is, September 21 to now) and ignoring everything else, what it shows is that Barack Obama's support has slipped 1.6 percentage points while Mitt Romney's has improved 2.6 points.

That's a relatively small group! And even there, it's unlikely that very many people "switched their votes." What we're probably seeing is one of two things. Some of it may be people moving from a candidate to undecided or vice-versa (the current Romney plus Obama total adds to just under 93%, with undecided presumably eating up almost all of the rest). The rest? People slipping in and out of likely voter screens.

(Yes, it's possible that there are much larger migrations between the candidates which cancel each other out. Everything we know about voters, and all the evidence from the polls, says that hardly any of that is happening).

I'm not saying that any of that is unimportant -- if it holds, a 4 point shift is big stuff, and if it's all from the debate, it certainly would be a large debate effect. Although note that as of now we have no idea whether it will hold, and very little idea as to how much of it is from the debate -- and even the portion which is from the "debate" may be not because of the debate per se but because of press coverage of the debate.

But hardly anyone is doing what we would think of as changing their mind. Just not happening. Almost everyone who had already decided is staying put. Some people who were on the margins between a candidate and undecided are jumping back and forth over that line depending on when you happen to ask them and what's the last piece of information they heard; many of them are low-interest voters who usually wait until the last few days to start thinking about the election. And some of them are just more or less enthusiastic about voting, which seems entirely reasonable as a reaction to either the debate or the news coverage of the debate.

Nobody, or at least practically nobody, has actually changed their mind because the candidates' demeanor overrides substantive policy positions. Just isn't happening.


  1. nobody has actually changed their mind because the candidates' demeanor overrides substantive policy positions.

    Any American - not already a multimillionaire or older than 40 (and thus likely unable to self-insure) - who took a look at Ryan's "substantive policy position" wrt Medicare would, I think, find it difficult to cast a ballot for a ticket with him on it. Even brief reflection on the voucherization/privatization of Medicare should make it clear that a two-tier system will emerge that will truly suck for the "sick old" and the "old old".

    And yet, millions upon millions of people who can reasonably expect to be in scope for serious personal hardship from Ryan's Medicare overhaul will vote for a ticket including him anyway.

    So...this idea that voters make their choice based on the alignment of "substantive policy positions" with their interests...

    ...who exactly do you have in mind here?

    1. Certainly no one in Kansas, as is your point...

      Unless, some people vote on issues that do not start and end with their own narrow pocket books. Shudder!

  2. Interesting point about LV screens. One theory of mine: It seems that the polls that are more volatile, like PEW and some of the network polls, have looser LV screens. These are the polls that swung hard to Obama last month and posted big gains for Romney this month. The more enthusiastic side makes big gains.

    Polls that are less volatile, like Ras, have stingier LV screens. These polls barely moved at all this past week...or in the month beforehand.

    1. The news on LV screens is actually worse. I saw a paper the other day (think it's a working paper, not yet published) that the screens, while having some predictive value, are actually not terribly good at their core function: predicting whether or not a person will vote. I think that a number of people who get screened out will vote (but, I remember there being HUGE variation in this number...20-something percent in one survey and 50-something percent in another!), and that a number of those screened in won't vote (I think this number was relatively consistent across the various samples they used).

      For what its worth, I think they were using the exact same screening question in all the studies. Rather, it was the fundamental problem of asking somebody to predict what they were going to do in a month.

      None of this says that polls using LVs or RVs or whatever are "wrong"....but it is a cause for skepticism. That, plus the cell phone difference that seems to have emerged in this cycle (but did NOT exist in previous cycles!) has made me very iffy on poll quality in the last year.

      It's not that I don't belive the polls. But, I'm increasingly skeptical of what we can get from them.

    2. I believe the case is that the LV screens do a terrible job of predicting individual cases...but that despite that polls with a LV screen still do a better job of predicting the final result than RV polls.

    3. Exactly right. Which then poses a puzzler: can there exist a set of circumstances where the LV screen does worse?

    4. Did the Republican primaries give any clues as to how accurate the polls are this year? Or are Republican primary voters, perhaps being older, less likely to depend on cell phones anyway?

    5. Scott: no idea, but an excellent question. I wonder if posting it to Monkey Cage would get Sides or Gelman to ask Silver to look into his repository of data to follow up on his earlier post. They both have connections to him, so posting the same question at Monkey Cage might get it answered.

  3. The other thing they constantly tell us is that these polls are within a few points; so who knows. The undecideds, which should be called under informed, are what it's all about. It's not as though a big pro Romney person changed their minds nor one who truly wants what Obama has done. Clinton has a good YouTube out there on it with his take on the debate and Romney's shift to being a moderate-- maybe.

  4. The race will most definitely be interesting to watch as we near November, that's for sure! I've been enjoying your blog, wanted to share another way to get a little politically involved that might strike your fancy:

  5. This comment has been removed by the author.


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

Who links to my website?