I can’t agree with Scott Alexander’s latest entry – on the likelihood that ten independent variables collectively account for the sharp drop in crime over the past 10 years. It’s called How Likely are Multifactorial Trends?, and it reasons from the a priori unlikelihood of 10 otherwise-independent factors randomly trending in the same direction to the conclusion that there has to be some more focused cause for the observed drop in crime.

It’s not that I think multifactorial trends are less likely than monofactorial ones; that much I’ll buy. I’m just not sure that I buy that this one isn’t underlyingingly monofactorial.

The mistake that I think is being made here is not counting "general frustration with the crime level" as an underlying – if admittedly unmeasurable – variable. If enough people get fed up enough with something, they will each worry about it in their own way, and this has a way of focusing attention on it. Even in the original Vox list that Alexander is responding to, this alone would tie together about half of the listed hypotheses – namely: (1) More criminals are getting put in prison, (2) More police are on the streets, (3) Broken-windows policing, (4) Police have improved policing in other ways, (5) More permissive gun laws, (7) People are using less cash, (9) Gentrification is taking over crime-ridden neighborhoods, (12) Crack consumption has declined, and (13) America’s gangs have gotten less violent.

Some of these are less obvious than others. For example, the ones about guns, gangs, and crack – so let me explain my thinking for those three. With guns, it’s that once the crime rate pushes past a certain level of psychological tolerability, people start demanding more ability to fight back. If the police can’t help, they have to help themselves. If this sentiment grows to the point where the usual gatekeepers start privately secondguessing themselves, you might, indeed, see a relaxing of gun laws. Crack and gang violence are probably correlated to some degree, and the point in both cases is that once there’s a moral panic about something, public attention gets fixed on it. That means police resources are devoted to it, parents spend more time talking to their kids about it, etc. This results in a lot of pressure that gets responded to even by the criminal element. Remember – the criminal element is a parasitic class. They don’t actually want to bring down the establishment because they feed off of it. If the establishment decides that something crosses a clear line, I think they generally take steps to keep that thing under wraps.

Indeed, I think if you wanted to you could throw (7) People are using less cash into the mix too. True, that’s mostly under the "tecnology has improved" hidden variable, but some of people’s willingness to switch to cash might be motivated by fear of being mugged – this being particularly true of the shopkeepers who make the cashless payment methods available to their customers.

Speaking of the "technology has improved" hidden variable, that one, if extant, drives probably 4-5 of these factors. There’s some overlap, so let’s call it "2 additional," and if you buy that then you have two underlying, hypothetical hidden variables underlying 63% of what’s there.

So, where does this leave Scott Alexander? Well, on the one hand, I’m just arguing in favor of his point in a bit more detail. After all, his main point is that if you’ve got a model where ten independent things have to randomly all trend the same way for a while to make it work, your model is weak. And what I’ve done is go through and argue that a lot of what seem like independent factors might actually stem from the a common source, in more or less roundabout ways. Which is nothing if not conceding the point that models with fewer factors are more explanatory in general than models with many. On the other hand, the factors I’m hypothesizing are of dubious scientific value in that they can’t be observed in any way, not really even indirectly (not without a certain amount of circular reasoning anyway). They’re purely speculative.

When you’re in a situation where the underlying cause/causes is/are scientifically nebulous, you end up with what looks like a wild coincidence of unidirectional trends. And it’s worse for the fact that the underlying cause won’t have caused them all to the same degree, or in the same way. It’s just, there’s a "force," for lack of a better term, nudging everything in a particular way, and that comes out looking like a hugely coincidental jumble.

It’s not really even all that difficult to come up with other examples. What "caused" German Reuninfication? Nothing specific, obviously. Which means people can, will, and do argue endlessly about it. Before you can even have a cogent discussion about it, you have to define the parameters of the debate, because the field is just too wide. Was it that the Soviet Union weakened? Was is that the Germans just really wanted it? Was it these Monday Demonstrations? And, if you buy that it was the Monday Demonstrations, do you take it back a step and buy the theory that the regime getting caught red-handed falsifying election results the year before was finally a step too far for everyone? Was it the Solidarity movement in Poland? Was it that Margaret Thatcher got elected and the security status quo ended? I think anyone sensible is going to agree that if there’s a "THE cause," it’s just that the Germans wanted it. The national split was ultimately unsustainable. But what form it took and when it actually happened were subject to a number of random noise-type variables. For example, the Soviet Union weakening. For example, the SED falsifying elections results and getting caught at it by the West German media. For example, a long-simmering trade union dispute in neighboring Poland. For example … whatever, really. The point is just that large numbers of people in Germany feel like there should be one Germany, and they have a vague idea of where the borders should be and what counts as "German" and so on. It doesn’t matter that no one has the slightest clue how to begin to study that particular underlying variable – the point is that I don’t think anyone seriously doubts that (a) it exists and (b) it "caused," in some important sense, German Reunification. As Willy Brandt may or may not have put it: "Jetzt wächst zusammen, was zusammengehört." ("Now that which belongs together is growing together.") That’s all there actually is to it: a vague agreement among pretty much everyone in the world, especially in Germany, that the border was wrong. But good luck trying to spell out what that actually means in measurable terms. You’ll have more scientific luck just calling it a giant coincidence that a bunch of things trended in the same direction in the 1980s.

Why is religious belief in sharp decline? Again, if you’re looking for a clear cause, you won’t find one. Any attept to study it is going to come up with a lot of seemingly independent factors. Technology has improved, leading to an explosion the availability of explanations for mysterious-seeming phenomena. Communication and mobility have improved, so people are more exposed to other people’s religions. Social controls have weakened. And so on. You can go on all day like that. If there’s a "THE cause" for it, though, it’s that no one ever really believed in religion to begin with. It was always a kind of murky proposition that had to be propped up with rituals and sophistry and propaganda – appeals to "faith" and so on. If religion seems to be eroding all at once, then just because some other things conspired to make it more socially acceptable to express doubts about religion, and this process feeds on itself. The run of history is long, and so while it might seem like a giant coincidence that a number of variables happened to align just at the moment that made religion less appealing to people, the truth is that it was an effect in search of a cause. If we keep spinning that particular slot machine every couple of decades, we’re eventually going to spin three lemons. And so we did.

Why did humanity develop language? Chomksy’s a big "single cause" guy here. He loves to speculate openly about a single congitive quantum leap that made the whole project possible. But this is more or less the way a creationist uses the anthropic principle. It seems hugely unlikely that a coincidence of evolutionary causes could have given us language, so we ascribe it instead to one crucial but ultimately magic "evoluationary quantum leap," such as the development of mental recursion (Chomsky’s current favorite "underlying cause"). But the truth is that a coincidence of evolutionary causes gave us language, in part driven by the immense usefulness of same. Language comes out of a nexus of independently-evolved extremely useful things, and the usefulness of these things acting in concert assures that it survives. The underlying cause is "language is hugely useful." But you can’t study the causal force of something like that. So, you study instead all the many proximate causes that came together to make it possible.

So, Scott Alexander is right and he’s wrong. He’s right that models that attribute trends to every available cause are unappealing. He’s wrong that there’s always an alternative explanation. Just because a trend exists and has a cause, it does’t follow that the cause is studyable or even articulable. In a great many cases, we’re pretty sure we know what the underlying cause is, but there’s just no way to talk about it meaningfully – certainly not at the present level of scientific knowledge. So, we content ourselves with complex causes. Because we have to.

I bet a lot of money the decline in the crime rate turns out to be like that. The underlying cause is that crime is extremely unpleasant, and there is a threshold above which it becomes intolerable, and we passed that threshold in the early 70s, and so a bunch of independent forces were brought to bear on the problem, to greater and lesser degrees, and more or less directly. At the level at which you can study it, it looks like a huge coincidence. At the level of actual causality, there’s a single cause that we can’t quantify or really even talk about meaningfully – but we nevertheless all know what it is: enough was enough.