On Talking About Election Results

As I see smart bloggers and pundits puzzle over the results last night (samples: Matt Yglesias, Ezra Klein, Marc Ambinder), I see a pattern.  They seem to know that there's something wrong with the way pundits react to elections, but...well, listen to them.  Yglesias: "It’s always dangerous to try to sum up a bunch of disparate events into a single trend, but..."  Ambinder:

It's a funny habit we political pundits have. If, say, 3,000 votes separate a winner from a loser, we forget that a small shift in some part of a state could have swung those votes the other way, and we tend to massively over-interpret the meaning of the tiniest of margins. So let's say that the results in Arkansas were flipped -- that Lincoln won by a point, the interpretation ought to be nearly the same, logically. 

OK, I think I see what the problem is.  They're trying to do the wrong thing.

Hey, pundits: don't tell us what the election "means."  Don't try to interpret.  Don't try to figure out "what the electorate is saying."  We don't know those things.  Sometimes because we have insufficient information (we didn't even have exit polls, but exit polls or not you're not going to be able to do a real analysis of the electorate's preferences overnight).  Sometimes because the thing, like the "meaning" of an election, doesn't exist.  It's especially hard to extrapolate from a random group of primaries that happened to be on one day to general elections months from now (for example, an "anti-incumbent" vote is something very different when it's a dedicated Democrat voting against a moderate Democratic incumbent in a primary than it is for that same Democrat to vote for a Republican in November, and willingness to do one thing may not at all signal willingness to do the other).  It's hard enough to extrapolate from this set of primaries to all primaries. 

What does exist, and what we do have some information about, are the effects of elections, and I think it's perfectly reasonable for everyone to start figuring that out right away.  We can figure out, or at least have some informed speculation, about whether a result is good for the Republicans or good for the Democrats (see Jonathan Chait on Nevada).  We can talk about likely policy effects.  We can talk about likely interest group reactions (that's the good part of Ambinder's piece).  Again, all of this is even easier when it comes to general elections, when we can have a pretty good idea of the policy implications of election results. 

Getting back to Ambinder: he's correct that it's entirely goofy to talk about "meaning" as changing dramatically if a handful of votes went the other way, if by meaning pundits are talking about interpreting the mood or the wishes of the electorate.  On the other hand, if one thinks about close elections, it made a huge difference that Kennedy beat Nixon, that Nixon beat Humphrey, that Bush beat Gore.  So it makes no sense at all to ignore those effects just because a contest was close and could have gone either way.  We have actual results.  We can count the actual numbers of incumbents defeated, or the actual numbers of Republican women nominated.  My advice?  Stick closer to those things, and avoiding amorphous questions about mood and meaning.