Ask how to solve problems, not why they happened. By Matthew Yglesias

Read time: 10 minutes

Ask how to solve problems, not why they happened

Identifying causation isn't the same as doing policy analysis

Little kids, journalists, and academics all love to ask “why?”

Why did this happen? Why are things this way? X happened because Y happened first, but why did Y happen?

These instincts can serve us well in finding interesting stories or research topics, but I think in a way they can also derail us when we’re trying to solve problems in public policy.

A few weeks ago, for example, I found myself in an argument with a think tanker about why school absenteeism exploded during the Covid-19 pandemic. My antagonist was saying that this was part of the downside of prolonged school closures. I argued that if you look at the numbers, absenteeism soared by almost as much in the places that kept schools open — a nine percent increase in the districts with the most in-person schooling versus a 12 percent increase in the districts with the least. That three percentage point gap isn’t nothing, but it’s smaller than the gap between the poorest and the richest districts and the whitest and the least-white districts. What’s more, we know that poorer and less white districts had less in-person schooling. So it seems to me that:

Absenteeism rose sharply even in districts that didn’t close for the pandemic.

There was a larger rise in the districts that closed longer, but this is accounted for by the differential demographics.

Therefore, closures per se did not have a significant causal impact on absenteeism.

Note that I say this as someone who was against school closures all along and would be happy to claim vindication on this point. I just truly don’t think they have a lot of explanatory power on the absenteeism point.

What I think is that the pandemic itself eroded the attendance norm. After all, even if you kept schools open, you had to shift norms around sick days. You had some parents who were afraid of their kids catching the virus. And you were in the midst of a national movement that was making everyone more skeptical of tough enforcement measures. The attendance norm, I think, was a casualty of the pandemic and, to an extent, post-Floyd concerns about enforcement, not of school closures.

At any rate, causal inference is fun to argue about. I bet that clever academics using more advanced math than me eyeballing charts will be able to shed more light on this topic over the years and maybe even resolve it conclusively.

But even though I enjoy this sort of thing, it’s also pretty plainly irrelevant to the question at hand, which is “what, if anything, can we or should we do about absenteeism right now?”

Problems need solutions, not explanations

The United States is a very rich country, yet our life expectancy is closer to that of Poland or Panama than Germany, Japan, or Italy. And that’s striking because we are actually a lot richer than Germany, Japan, or Italy at this point.

It’s common when people are writing about this to frame the question around “why is American life expectancy so short?”

And that is definitely one way into the topic. The United States, for example, has more homicides than those countries. And even though relatively few people are murdered even in the United States, homicides have a large influence on life expectancy statistics because murder victims tend to be young. You can go on like this and try to do the statistical decomposition and that will definitely tell you something. In particular, I think one big takeaway is that short US life expectancy has less to do with our outlier system of health insurance than one might think. Americans are more likely to die violently, to die in car accidents, and to die of drug overdoses than are Europeans. We’re also a lot fatter, which generates health problems.

That said, I don’t think the statistical decomposition really addresses what matters here.

Dying is bad, and avoiding death is worth investing material resources in.

The USA has more material resources than Italy, but much higher mortality rates.

It seems like we should invest some of those resources that Italy doesn’t have in trying to avoid dying.

What this calls for is tractable policy ideas that would have a large influence on American mortality rates. Just observing that it would be nice for Americans to be slimmer doesn’t get us there. One issue is that we are probably fat at least in part because we are rich, and we use our riches to buy lots of tasty snacks and this just happens to be bad for you. On the other hand, in today’s world, maybe we could use our riches to buy everyone semaglutide. But that gets us into other questions. Apparently it only costs Novo Nordisk $5 to manufacture a dose of Ozempic, which suggests that some form of price controls or bulk discounting might be useful. But of course, that raises questions about the impact on innovation. Which raises questions about the regulatory costs of innovation.

Throughout the whole Ozempic pricing discussion, I’m continually unclear about production bottlenecks. You keep hearing that despite the high price in the United States, there is limited availability. It’s one thing to say it costs $5 to make a dose, another to say what it would cost to create the capacity to generate many more doses.

My point here is not to solve the obesity crisis, but just to say that you start with the question “why do Americans die younger than Italians,” but you actually end up debating a totally different topic about pharmaceutical regulation. Similarly, I think it’s clear that one big reason why Americans die younger than Italians is that there are all these guns everywhere. But while “if all the guns in American vanished, the murder rate would drop” is true, it’s not a workable suggestion.

The root causes of crime

A related issue is that people sometimes like to talk about the root causes of crime — crimes happen much more in low-income areas, for example, so maybe poverty, or the concentration of poverty, causes crime.

If you look at it, though, far and away the biggest demographic correlate of crime is gender. Men commit murder at astronomically higher rates than women, and this is so well know that people tend to forget to even mention it. It would be like doing a big article about how the sky is blue. And murder is the least of it — how often do you hear of a girls’ night out going awry and turning into a bar fight? Men (and boys) commit way more violence. And lots of other crime-related things are downstream of that. It’s not like women don’t have the physical strength to hold a pistol and rob people at gunpoint. And it’s also not like it’s never happened. But women don’t rob people at gunpoint with nearly the frequency that men do, because if you’re going to threaten to shoot people, you sometimes need to pull the trigger and women are just much more reluctant to do that.

Everybody knows this and just about everybody knows it has something to do with testosterone, to the point where few people bother looking into exactly what the mechanism is.

But if what you want is a causal explanation of why crime happens, or a statistical account of the predictors of crime then this “something to do with testosterone” is going to do way more work than looking into concentrated poverty, street lights, gang takedowns, or anything else. Of course, if you look into the gang takedowns, you’ll find that you are mostly taking down men (I don’t even need to waste any time looking into whether that’s true because it so obviously is).

The reason for the focus on these other things isn’t that “lack of streetlights” is a bigger explanation for crime than testosterone. It’s that lack of streetlights seems like a problem we can address.

But note that the issue here isn’t that darkness biochemically activates the masculine urge to kill. It’s just that one of the things we can manipulate relatively straightforwardly through policy is the odds that a crime will be detected. Streetlights, hotspot policing, DNA databases, and surveillance cameras all work, and they all work the same way: When people believe they are likely to get caught, they are less likely to commit crimes. And when people believe potential perpetrators are less likely to victimize them, they are less likely to unleash preemptive violence for their own security.

This is not the only possible solution. Most men aren’t murderers. Most poor people aren’t murderers. Most poor men aren’t murderers. All of the people committing serious violent crimes are outliers relative to their context. One family of solutions tries to identify those outliers and provide them with services to address mental health, impulse control, or substance abuse needs. Another solution is to lock them up and throw away the key. But changing people is hard and incarcerating them is expensive (and, in the US, often inhumane), and I think that in most cases, raising the odds of detection is the most cost-effective solution around. And yet it pretty flagrantly ignores the baseline question of why, exactly, an idiosyncratic minority of the human population behaves differently in low-odds-of-detection scenarios.

Egalitarian growth

People don’t talk quite as much about income inequality as they did 10 years ago, but the subject was often approached as a kind of murder mystery — who killed the egalitarian economy of the 1950s and 1960s?

If it turned out that the answer was “skill-biased technological change” had driven inequality, then rich people were innocent of the charges, so the good progressive position was to say this wasn’t true. But if trade and globalization did it? Well, then we were looking at global class war and the rich were guilty. Except what if the increase in trade was driven less by policy than by technical improvements in container shipping — then maybe the rich were innocent after all? But if tax cuts drove inequality, then definitely guilty again.

I think the right thing to say about all this is that trying to understand economic history is clearly a valid undertaking that can shed some light on policy problems.

But we will understand the history better if we don’t approach it as if these questions have a simple relationship to contemporary policy debates.

And we will understand contemporary policy issues better if we ask questions that are forward-looking rather than backward-looking. One classic idea in economics is that we face a “big tradeoff” between efficiency and equality, so anything we do to reduce inequality is going to reduce the rate of economic growth. But is that really the structure of the choices facing us? Suppose we increased the supply of medical doctors. That seems like pro-growth policy that would tend to reduce inequality. So would skilled immigration in general. If you substantially increased the number of legal work visas available to people with six figure job offers, that would drive a lot of economic growth and also tend to compress the wage distribution.

Regulatory barriers on housebuilding are a major drag on the growth rate of the American economy and they place the largest burdens on relatively low-income renters.

You can go on and on, down the list. It’s not that Mancur Olson’s big tradeoff point is wrong, exactly. Most redistributive policies are, as he calls them, “leaky buckets” that lead to a less efficient overall economy. But lots of redistributive policies operate to redistribute resources upward. The dentists’ cartel is impeding efficiency not in the name of equality (dentists are, on average, richer than their patients), but in the self-interest of dentists. You can have a pro-growth egalitarian policy that simply tries to focus on finding the leaky buckets of upward redistribution.

In all these cases, the key to solving problems is to try to focus on solving the problem rather than asking what caused the problem.

Like many people, I am nearsighted. This has something to do with the shape of my eyeballs, and I guess (without looking it up) the origins of misshapen eyeballs have something to do with genetics or early childhood experiences since I can’t imagine what the other candidates would be. And there is a technology that involves reshaping eyeballs with lasers, which some people use. But for me, and for most nearsighted people, the convenient solution is corrective lenses. And that’s not really addressing the “root causes.” If you see a group of people who can’t see, then the solution is they either need information about glasses or maybe they need money to buy the glasses. Not because that’s the most sophisticated analysis of the problem, but because it’s the most tractable fix.