« International Women's Day | Main | Purim again. Sigh. »

Fun* with numbers again!

So, David Wasserman writes that Purple America Has All But Disappeared, by which he means that at the county level, margins of victory in elections are getting bigger. He writes that “if you feel like you hardly know anyone who disagrees with you about Trump, you’re not alone: Chances are the election was a landslide in your backyard.”

Now, Mr. Wasserman is scarcely the only writer who exaggerates the broader impact of real changes, so I don’t really want to peck at him specifically. I’m just writing about it as a recent example of a thing that gets to me about political science and stats and the people who write about them. Essentially, they take deviations from the historical path that are huge by statistical standards, and then apply those as if they were huge by social standards. Let me explain what I mean, using this article as an example, though it is scarcely the worst or the most irritating such.

In 1992, according to Mr. Wasserman’s data, 39 percent of people lived in counties where one candidate got at least 60 percent of the major-party vote. Now, a 60/40 split in an evenly divided country is a big deal—Bill Clinton got 53% of the major-party vote nationwide. In 2012, it was up to 50% of people living in such landslide counties, despite the election being even closer; in 2016 it was up to 61%. That is in fact a big deal and a big change, and totally worth studying—by political scientists, I mean. It’s also probably worth people like me, who are interested in our democracy, knowing. It’s a Good Thing, then, for Mr. Wasserman to write about it.

But if, instead of looking at the rate of change, you are looking at it how it affects the lived experience of people… I’m not so sure. Remember the numbers we’re talking about here: from 39% to 61%. That is, two out of five people experienced such landslides in 1992 and now three out of five do. What has changed is about a fifth of the population—or, to put it more accurately, 78% of the population lives in a county that either had no landslide in either year or had landslides in both.

Well, that’s not actually super-accurate, because over twenty-four years, there’s just a shit-ton of change. The student worker at the counter who voted for Hillary Clinton in our landslide county hadn’t been born when Bill Clinton was elected in 1992. I was living in one landslide county in 1992 (the City and County of San Francisco, in fact) and a different one now. If we’re talking about the lived experience of people, it now generally involves moving from place to place, which in large part, I think, accounts for the shift Mr. Wasserman documents. But as important as it is to keep that in mind, let’s remember that we’re still talking about most of the counties unchanged.

And then: look at the claim about hardly knowing anyone who votes the other way. Remember that the landslide counties are counted as 60/40 splits, which is a big, big deal in terms of deviation from the national numbers, but is still fundamentally a three-to-two breakdown. Think about it this way: if there were twenty yard signs on your block, and eight of them were red, would you say there were hardly any red signs? You would not. I hope. Because that’s just not what hardly any means in people’s actual experience. And the change from 11 blue signs and 9 red ones to 12 and 8, or even to 14 and 6, does not constitute the formation of an impenetrable wall of homogeneity. 12 blue and 8 red (or vice versa) is actually a pretty good working definition of purple, isn’t it?

Aha! say the people who have actually read the article: Mr. Wasserman may exaggerate the shift in some counties, but he clearly states that “The electorate’s move toward single-party geographic enclaves has been particularly pronounced at the extremes.” And it’s true, the data show that the extreme landslide counties, by number or share, have increased far faster than the ordinary landslide counties. From 4% in 1992 to 21% in 2016, and that’s a huge rate of change—and I will again emphasize that this is a Big Deal and should be studied and noticed and talked about.

And yet… again, we’re talking about a change from one in twenty to four in twenty. 17 out of 20 people (or, if you prefer, 85%) did not experience that change. And that change was presumably a shift from a county going 65/35 to 75/25—a shift one out of ten neighbors. So emphasizing those counties, we might say that two out of ten people experienced a change that may have involved one out of twenty of their neighbors, and another one or two out of ten experienced a somewhat bigger change, involving one out of ten of their neighbors.

Most people experienced no such change at all.

Again: most people did not, according to the statistics presented in this article, live in counties that experienced any such monstrous shift in lived experience as Mr. Wasserman describes, that is, 5% or so fewer neighbors of the opposite Party. Some people did! Enough to make a big difference in who wins elections! Enough to make a huge difference in policy outcomes, in culture, in representation. More than enough to make it worth tracking and studying what’s going on. Way more than enough to change how campaigns should manage their resources. Just not most people, and at a guess, many of the ones who did experience that 5% change wouldn't even have noticed.

Tolerabimus quod tolerare debemus,

Post a comment

Please join in. Comments on older posts will be held for moderation. Don't be a jerk. Eat fruit.