Wednesday, October 24, 2012

The Divergence, Explained

There's a story on Teagan Goddard's Political Wire today about the apparently bizarre disconnect between the current state of punditry on election 2012, and the current polling data.

As you know (you wouldn't be reading these words if you didn't), there is a droning mantra of paid political analysis out there right now, insisting in a way we haven't seen since 1980 that the challenger is closing hard and fast and has all the momentum on his side. The public has gotten to know him, gotten comfortable with him and, especially after the first debate, come to find him plausible as a potential President. Mitt Romney has passed the "living room test," and with the strong economic headwind facing the incumbent, all that is left for the challenger is to close the deal.

Meanwhile the state-level polling data has shown very little movement in Governor Romney's direction. Even the much-ballyhooed national tracking poll by Gallup (which shows Governor Romney with a substantial lead in the countrywide popular vote) suffers from the unfortunate credibility-inhibitor that it dramatically -- one might even say intentionally -- under-counts minorities. In other words, aside from a poll of angry southern white guys, none of the hard data about the race is consistent with the punditry. President Obama is up by small but structurally sound leads in Ohio, Nevada, Virginia, and Iowa; he is dead even in Colorado and Florida, and within the MOE or just outside of it in North Carolina and New Hampshire.

Even more perplexing is pundit-land's seeming lack of appreciation for just how daunting Mr. Romney's comeback would be from this starting point: Barack Obama, it happens, could lose *all* of Virginia, North Carolina, Iowa, New Hampshire, Colorado and Florida, and with just Ohio and Nevada, where his leads are outside of the MOE, he would win with 272 electoral votes. We keep hearing that it's all about Ohio -- and for *one* of the two candidates it surely is -- but Mr. Romney's urgency to win Ohio is completely one-sided: If he wins it he can still lose. Indeed if the election were held today, and Ohio went to him (somehow), he almost certainly *would* still lose, because he would fail to carry Nevada, Iowa, and Virginia, and the Ohio victory would be meaningless. If Romney loses Florida, it's over; if he loses Ohio, it's over; if he loses Virginia, it's pretty close to over. If he loses Colorado it's basically over and if he loses Iowa and New Hampshire put together it's more-or-less over, and at the moment he is trailing or at best tied in ALL of these places. And still the pundits clamor on about Romney's camp feeling good about where they stand and looking forward to culminating their late turnaround on election day.

So what gives? 

The most popular theory -- one you may be thinking in your head already -- is that the pundits are trying to artificially enforce a "race-is-tightening" narrative in the interest of ginning-up ratings. But like all conspiracy theories this notion suffers from two fatal flaws: The large number of extremely diverse people who would have to all agree to wink at each other at once, and the cripplingly high incentive to cheat. If I'm a paid pundit, and everyone else is saying something that we all know isn't true, then I can corner the market on smarty-farty just by telling it like it is and being the only one who turns out right, n'est-ce pas? Forgive my wet blanket economics background here, do, but the notion that a couple of thousand paid and semi-paid political journalists from all points on the political spectrum could come together to agree to actively misrepresent the only thing they get paid to care about is, to borrow my Grandfather's language, "pretty thin."

The other theory is that somebody -- either the pundits or the statisticians -- are just garden-variety wrong. But this idea fails at least one of the two tests described above, as well: In a community of this many people, half of whom get paid to be correct about the analysis and the other half of whom get paid to be correct about the data, the chances of a Dewey-Truman repeat get slimmer and slimmer every election cycle. There is too much incentive to be a little better at the game than the other people who share your side of it, whether your side is the stat-counting or the bloviating about it in print. (Anyone who hesitates to agree might remember -- albeit painfully -- just how sure we all were that John Kerry would win with comparative ease in 2004 because the pollsters were undercounting people who didn't have land-line telephones to call.)

If both of these popular theories seem unlikely to me, it's only because they also both seem un-necessary. We all know that polling data lags the true state of things on the ground by a few days, because it is during those few days that the poll in question has to be "in the field," then collated, then drafted, then edited, then published, and then press-released to the political sites. But what about the punditry? Surely there's no lag in the punditry whatsoever, right? I mean, after all, a pundit only has to write a new column to reflect any new state-of-the-race dynamics as they tumble into his or her individual consciousness. In theory, a pundit who was saying in writing on the morning after the first debate that Mr. Romney had all the momentum and was likely to win, could just as easily turn around on the following Monday and publish a new column in which he or she conceded that the battleground polls didn't seem nearly as impressed with Romney's debate victory, and maybe the whole election isn't as different from where it was before. In theory, the one-man or one-woman pundit operation could just say that things had changed, as they changed, and thereby update its reporting in real time: unlike the polling data, there would be no corresponding time-lag on the comment side of the house.

In practice, it seems to me, the psychology of political punditry doesn't quite work that way.

Suppose you get paid to write about politics. Suppose, for example, that you have a column for the New York Times. Now suppose that the first debate appears to be a decisive victory for Governor Romney. The first question is, do you publish that sense of things right away?
Be careful before you answer: If you crash this news, and then the post-debate numbers end up behaving more in the manner of a convention bounce then you will look like an idiot. But this isn't a two-sided dilemma: If you play it cool, hang back, and say qualified things about the situation, you still have time to be right, later, when your initial impressions are confirmed. What a liberal political writer in such a situation and a conservative one would have in common, then, is that they both have a strong bias to be conservative in their analysis in the original sense of the word: Take it slow, let it steep a bit, and be sure. Slow to convince is *way* better than wrong, in the eyes of a reading public. Especially in such a cut-throat business.

Okay, so now suppose it is two weeks later: You and all your colleagues have, with varying degrees of breathless abandon, come around to the view that Romney is closing hard and could just about pull this off. You've said so. You've said so in writing. Do you retract the whole thing the first time you look at a poll out of Ohio that shows the first debate made no difference whatsoever? Well, gosh: Now the inertia is even *worse* -- because, of course, you have already changed your mind. You have already publicly changed your mind. If you change it back, now, you run the risk of really, really, really looking like an idiot. And this is something you really do not want to do.

If election 2012 goes the way of the state-level polls (as I fully expect it to), and Barack Obama is reelected (as I fully expect him to be), the story won't be that the pundits are wrong. They won't have been wrong. What they will have been is a lagging indicator. And that's not the same thing by a country mile. ...It's just a shame, since a lagging indicator with less than two weeks to go before the election isn't nearly as helpful or informative as a slightly more guileless, and by that measure *leading* indicator would be, at this stage.

Dave O'Gorman
"The Key Grip"
Gainesville, Florida

2 comments:

Anonymous said...

I keep fivethirtyeight.com on "speed dial" because they make me feel better, despite the fact that I am terrified they are wrong. Nate seems to do an enormous amount of number crunching, but the one thing I wish he'd do is publish the electoral votes for each state. And that one with the "what ifs" remember that one from 4 years ago. Who is that? Anyway, great, great article and I wish you'd publish THIS stuff since you are so incredibly good at it.

Anonymous said...

Nate and Dave should team up: Dave writing, Nate confirming.

This essay would make a good diary on the Great Orange Satan blog.