Normie Distribution

 

I always used to be amused at the Twitter users paranoid about political poll results because they had themselves had never gotten the call from Emerson or Siena, and suspected that the polling outfits must be making the data up. As a New York City resident, a registered Democrat, and a certified Old Fool who kept a land line for a ridiculous amount of time and still today always picks up on a number I don't know on the possibility that the caller might be a long-lost relative or notice of a windfall from a class action I didn't know I was a party to (it's hard to scam me, I want to insist, but it's easy to get me to answer), I could assure them that it does happen, especially on questions of local politics. If the pollster never calls you, that's just a sign of how ordinary you are; there are so many millions of you that they're just never going to get around to your number. 

Sucks to be you! except when I'm trying to come up with an honest answer to whether allegation A makes me much less likely, somewhat less likely, somewhat more likely, or much more likely to vote for candidate C, or whether I don't know or don't care, in which group I increasingly fit, not because I'm any less partisan but either because I already know about the allegation and had already made my decision around it or I'm hearing it for the first time and have no idea whether it's even true or not (identifying it, in my understanding, as a "push poll", not meant to gather information but just to slime the accused candidate in some way that escapes overt publicity, and is probably false anyhow). Both of which happen all the time.

And then I'd also answer out of some residual sense of loyalty, as a one-time social science person, though I can't say I ever worked in one of the poll-happy disciplines. But then I always ended up screaming at the poll takers over the stupidity of the questions, often right from the beginning, because they love to start with the right-direction wrong-direction question, one of the absolutely stupidest in my opinion—a country or a municipality is not a goddamned boat on a random voyage to Badland or Goodland depending on the captain's expertise or virtue or anything else, it's a community of subcommunities of individuals in cross-cutting family and not-family groupings, all headed in different directions and more than one direction at a time, so that your answer to that question can't mean much of anything more than what kind of mood you're in at the moment crossed with what you imagine is the "correct" answer for your own political tendency. 

But that's not something the human at the other end of the line needs to hear about, especially if I get so mad I hang up, and wreck five or ten minutes worth of their work; nowadays, I'm mostly just firm but polite, as with a Medicare Advantage pitch: "Sorry, can't talk," and hanging up. I did take one the other day, on the New York mayoral race—it was fully automated so I could use my phone keypad and they said it would take under five minutes, which was of course a lie. I said I didn't know if the city was going in the right or wrong direction or who I would vote for if Cuomo, Adams, and Sliwa were the only candidates and that knowing the Anti-Defamation League had accused Mamdani of antisemitism made me "slightly more likely" to vote for him, don't know if that was the most effective answer or not.

Meanwhile, the defects of political polling have become a topic in the Democratic political science discourse, and its preoccupation with the concept of "popularism", the idea that (I'll just caricature it a little) Democrats need to be designing their campaign manifestos on the basis of what's "popular" as measured in polling, because, after all, how else would you know what's popular, if you couldn't determine it scientifically?

I feel that I was involved in this popularism debate in an earlier phase for a long time, before the 2016 presidential election, when it was more an issue from the "left", and I used to spend time pointing out on the basis of poll results that folks really liked the idea of taxing the rich, or providing healthcare with no charge at the point of service, or things like that, which Democratic consultants were pretty negative about, because they were certain that folks don't like "socialism". Nowadays, the consultants have developed their own popularism from the "center", maybe arising from the trauma of the 2016 election, which was indeed a pretty shocking thing, seeming to have exposed terrible holes in the way we understood what the American people did and didn't want—the suggestion that "the American people" really wanted discrimination against immigrants of color and Black citizens, more police brutality, and tax cuts for the rich, the things the Donald Trump campaign had generally emphasized, and especially the pseudo-issues surrounding fear of transgender women in high school and college, who took advantage of the generous open-mindedness of liberals by demanding the use of girls' bathrooms and using their superior male muscles to defeat girl athletes in track and field sports and whatnot, though the numbers of such women seemed pretty close to zero and publicly limited to one young woman who was temporarily barred from sharing her fifth-place title with a trans swimmer, not that she herself was even particularly upset, in her own telling, just irritated with the liberal NCAA:

Immediately after the meet, Gaines said in an interview with The Daily Wire: "I am in full support of her and full support of her transition and her swimming career...because there's no doubt that she works hard too, but she's just abiding by the rules that the NCAA put in place, and that's the issue."

So we're advised to say in all these cases that the Trumpies "make some good points"—defending human rights against them is "playing into their hands" if we do it overzealously.

Anyhow, the genial John Ganz has been developing a kind of polemic against popularism on the angle of the polling itself, that it doesn't actually provide an accurate objective picture of what the population is thinking, that it's in fact distortive and biased, "vulgar positivism" with far too much confidence in itself. For example:

Look at immigration. The numbers supposedly favored Trump so much that it was a mistake to attack him about the manifestly cruel process of deportations under his regime. Polling and punditry conspire to create one of those situations where “it would play into their hands.” On paper, you can see why: A majority of Americans once said they wanted less immigration and even to deport all illegal aliens. Now, Trump is underwater on immigration. What happened? Well, people saw what actually existing Trumpism looks like, and it’s ugly. It looks like a police state, because it is. And this brings us to the problem with issue polling. When you ask someone in a survey, "Should we deport illegal immigrants?" they are going to hear the word “illegal” and reason, “Well, yes, they did something wrong, it’s not fair.” On a very abstract level, it’s almost definitionally correct, tautological in its rational soundness: a person who is not supposed to be here should not be. Then come the actual images of what it looks like to realize that abstract equation, and people go, “Oh, no, this is awful.” Now the public likes immigrants again and wants them to have a path to citizenship rather than repression.

Positivism being the philosophical position that all empirical truths are effably knowable (in opposition to Popper's critical rationalist view that the only empirical certainties are negative ones—you can definitively prove when such-and-such a proposition is not true, but the alternative hypothesis that seems true is never more than unproven so far), and "vulgar positivism" the corollary assumption that things are pretty generally always what they seem, which should be obviously false on its face but persists in the guise of "common sense".

The positivism problem with political polling, I think, is to know what the respondents mean: What reality is the survey capturing? With Ganz's immigration example, for instance, did they understand the question the same way before and after Trump's deportations began? Had the meaning of the question changed in the light of current events, or had the events made them start thinking about it in a different way? Was it the minds that changed, in response to the situation (I used to think unauthorized immigration was bad, but now I'm starting to think it's OK)? Or were the minds reacting, consistently, to a changed situation (I used to think the law regulating immigration was OK, but not if it's like this)?

What's pernicious is when the phenomenon we think we're seeing in the polling data isn't really a phenomenon at all, but a kind of metaphor. For example, one of the most apparently straightforward kinds of polling is the standard horserace question: "If the election were being held today, which of the following candidates would you vote for?" The answer (Mamdani, Cuomo, Sliwa, don't know, or whatever) isn't in any particular way ambiguous, and there's little reason for the respondent to lie (that may be changing if our country is becoming a police state); it would seem to mean exactly what it says. And yet when you aggregate the responses from a series of polls, they seem alive: that's why we call it a horserace, with one of the candidates is pulling ahead in support from the polled population and another falling back. But of course the horses aren't actually moving at all, it's the population. Or the odds. Or statistical noise (studying it over a long enough period enables you to decide if it's noise). Maybe it is prompted by something a candidate is doing, or failing to do, and of course the consultants want to analyze that. The thing I particularly hate about the metaphor in practical terms—and the undemocratic thing, come to think of it—is that it keeps the focus on the show and away from the voters and turnout operations, which should really be the main thing in some perspective (identifying the "will of the people" as opposed to merely the guy who will purport to represent it, to say nothing of trying to analyze whether they do represent it or not).

Another is the metaphor of overall political orientation as the left-right spectrum, which I've pictured as a very long train (no doubt headed toward Wrong Directionville) with a single file of passengers standing shoulder to shoulder, occasionally swapping positions, waiting for the Overton Window to pass by on the windows they're facing. That's from the point of view of the ideological commentators, though; for the pollsters it's merely a normal distribution in the familiar form of the bell curve, two tails of the extreme arranged left and right around the overpopulated hump of the centrists or moderates or whatever you want to call them, as in the illustration at top.

Please note that that picture isn't the result of any empirical observations, but rather a mathematical construct used by political scientists in the discussion of the median voter theorem devised by Duncan Black (not our Atrios, it was published in 1948), which holds that in any voting system in which the winner of the majority always wins,  the candidate whose views most resemble those of the median voter, I guess the one who is most mid, will be the winner. In order to prove the theorem, it was necessary to define the median voter, and the bell curve, positing the median voter at the top, was used for that purpose. I don't think there's any formally adequate way of classifying all the voters and lining them up as in my train analogy Real numbers like these July figures from Gallup are often used to justify the bell curve

with a left-right breakdown from Democrat to Democrat leaner to non-leaner to Republican leaner to Republican, but that is plainly problematic from the start (this year, for instance, Democrat leaners (the group that has gained the most over the past year) are likely to be left of Democrats (that's why they don't identify as Democrats; they're mad at the party with the sense that it "refuses to fight"). The total number of independents is a lot larger than either pure partisan group, but it's certainly not midway between left and right, or coherent at all. As I keep complaining, independents, especially the swing voters who profess not to lean either way, are not midway between left and right as much as they are people who don't think much about it at all, and at least at the moment they're a pretty small group, hardly even the tip of the bell,

Indeed, it's more than reasonable to prefer a completely opposite picture in the current polarization, with the center entirely hollowed out, like this from Dartmouth's Feng Fu, Matthew Jones, and Antonio Sorianni

where there's clearly no such thing as a median voter (it much resembles the bipolar model I'm always trying to sell) and no way to apply the median voter theorem unless you try to apply it twice, median left for one party and median right for the other. In this way, there's no reason to take the advice of Tom Suozzi or Josh Gottheimer as a formal principle. I imagine Suozzi and Gottheimer know their districts well enough to know what they're doing instinctively, but it's really not science, and they won't win that way in most districts (as Gottheimer learned, perhaps, in the New Jersey gubernatorial primary he dismally lost this year).

The usefulness of issue polls in particular tends to be seriously compromised by fetishization of the bell curve, the idea that opinion on every issue ought to break down that way. The consultants trying to dominate the Democratic party should find some more insightful way of doing issue polling, or they should sit down and shut up.