Q (, comment,) and A

In response to my post on the Pew poll, Itchy writes:

So what is a swing like this supposed to mean? That Republican voters who weren’t going to vote suddenly decided to do so, and Democrats decided against voting — because of one debate?

How many people who will end up voting even watched the debate? I understand that the media reaction amplifies the consensus opinion about the results, but are there really that many people out there whose vote swings on something like this?

I don’t much believe in convention “bounces” or in these post-debate polls. I understand that, on the whole, polling is supposedly fairly accurate, and I don’t pretend to understand it in depth, but I have a hard time taking it seriously when polls swing this wildly.

Good questions, indeed.  First off, by all accounts a huge number of Americans watched the debates.  Kind of one of those makes you feel good about American democracy moments.  That said, whatever minds have been changed I strongly suspect are still more changed by the media echo chamber than by Romney’s actual performance.  (On a related note, Drum has a number of interesting posts about the fact that there’s a hack gap– had the situation been reversed Limbaugh and Hannity types would still be defending Romney, but as it was, pretty much all the liberal commenters lit into Obama).

Actually, was most intrigued by one of Drum’s commenters on a recent post about the debate.  I’m not quite an expert on likely voter models, but this strikes me as about right:

Likely voter models.  That’s the key here.

The reason we can see very large gains from what are really modest changes in the election outlook is that the answer to the likely voter question is subject to fluctuation based on very recent events.  Gallup, for instance, asks voters to say on a 1-10 scale how likely they are to vote this November.  Gallup rates those who say “10″ as “likely voters” (that’s about 75% of the registered voter respondents).  Anyone who says 8 or 9 is a “Probably.”  Anyone who says 7 or less is “Less likely to vote.”

So Romney has a great debate.  Some of the disguntled Republicans who were feeling so-so about him, and would have said “7″ before the debate, say “10!”  (Meaning: “Heck yeah!  I’m gonna vote!”)  This shifts the likely voter pool dramatically, and that determines whose poll responses get counted.  At the same time, you have to combine this effect with the sample-skewing effect of the fact that a few Republicans who would have hung up the phone on the pollster instead answer the call.  All this tilts the field toward the Republicans before the first undecided voter moves an inch.

Is it real?  Well, I would think it’s somewhat real.  A huge number of Americans are really not going to vote this election.  It matters a lot which 40% doesn’t vote and which 60% does.  How stable things will be between now and November, though, is trickier.  Hard to know.  My concern if I were the Obama campaign is primarily perception right now.  Polls can be self-fulfilling.  Early voting is happening, and there’s not a lot of time.

Now, to Itchy’s main point, my fallback position is that, ultimately, we really need to remember to think about polls as a snapshot in time.   If the election had been held in the past few days, I really think Romney would have done way better than if the election were held in the days before the debate.  That’s real and that’s reflected in the polls.  Of course, what I mentioned in the earlier post– there’s no way the PID numbers in the Pew poll will be true on election day reminds us that we need to keep this “snapshot” idea in mind.  It’s also why things like Nate Silver’s model is so handy, because it continues to use fundamental factors to help predict the actual outcome.

As I said before the debate, we were almost bound to see some tightening.  The debate seems to have made that tightening even more than history should have led us to expect.  That said, the fundamentals would still seem to, just barely, favor Obama.  But in a two-candidate election, just barely is enough.  For a while there, I was thinking 52-48.  Now we’re looking more like 50.5 – 49.  What the latest numbers are not cause for is the absurd Andrew Sullivan-esque panic.

 

About these ads