How Dirty Air Steals a Fifth of a School Year — And Only From Some Kids

Breathing moderately dirty air — air the U.S. government long called "safe" — measurably slows how much students learn. Not through dramatic smog events. Through ordinary haze on ordinary days. Researchers tracked the PSAT scores of nearly 10 million American high schoolers to prove it, and the way they proved it is as worth learning as the result itself.

The problem they had to beat

Dirty air and low test scores travel together. On its own, that correlation is nearly worthless, because polluted places are also poorer places — less school funding, more crowded classrooms, more stressed families. If kids in smoggy areas score lower, you cannot tell whether the air did it or whether poverty did it and the air merely rode along. To blame the air, you need pollution to move for reasons that have nothing to do with a student's life. The study used two separate tools to get there.

Tool one: compare each student to himself. The PSAT is taken more than once, so you can line up the same student's score in a cleaner-air year against his score in a dirtier-air year. This strips out everything permanent about that student — his family, his neighborhood, his baseline ability. Whatever's left can't be explained by "poor kids are just different." This technique is called a fixed effect: you hold a fixed feature (the individual) constant so it can't contaminate the comparison.

Tool two: let distant wildfires do the polluting. Same-student comparison isn't quite enough, because air quality in a given year might still track something that changes year to year — a local factory opening, a recession. So the researchers isolated one specific source of pollution swings: smoke drifting in from wildfires hundreds of miles away.

This is an instrumental variable — a factor you use to move the thing you care about (local air) without touching the outcome you're measuring (scores) through any other channel. Smoke from a fire 400 miles away doesn't know whether a school is rich or poor. It just arrives, dirties the air, and leaves.

But note carefully what makes this work — because this is the reusable lesson. An instrumental variable is only valid if it affects the outcome through nothing but the channel you claim. Economists call this the exclusion restriction, and it is an assumption you must defend, not a property you get for free. If distant wildfire smoke also, say, kept parents home from work in ways that changed how kids studied, the method would break. The reason this instrument is credible is precisely that a fire hundreds of miles off plausibly has no other route into a teenager's test score except the air he breathes. When you next see an "instrumental variable" study, that is the question to interrogate: is there really no other path?

What they found

A year of typical pollution exposure cuts learning by 0.04 to 0.06 standard deviations. To translate that: a full year of normal schooling adds roughly 0.3 standard deviations of measured skill, so a loss of ~0.05 is 14 to 19 percent of a year's academic growth — about a fifth — erased by air.

Two details turn this from a statistic into an explanation.

The ordinary days do the damage. The harm comes almost entirely from moderate pollution — 8 to 12 micrograms of fine particulate matter per cubic meter, which is below the EPA's historical safety line. Fine particulate matter, or PM2.5, is soot and dust small enough to slip deep into the lungs and cross into the bloodstream. The rare "dangerous" spikes are not the main culprit. The everyday grey is.

Timing reveals the mechanism. Pollution during the school year hurts scores. Pollution during summer doesn't. If dirty air were slowly damaging children's brains in some general way, summer exposure would hurt too. It doesn't. That specific pattern points not at diffuse neurological harm but at instructional disruption: bad air makes students and teachers foggy, tired, absent, and distracted — on the days school is actually running. Learning is a fragile daily process, and a polluted day is a partly lost day.

The unequal part

The same air does not harm everyone equally. The effect was three times larger in disadvantaged schools, and larger for Black and Hispanic students, at identical exposure.

Read that precisely. It is not merely "poor kids breathe dirtier air," though they do. It is that the same dose does more damage to them — most likely because their schools have thinner buffers: worse ventilation, no air conditioning, fewer resources to absorb a bad day. So pollution doesn't just lower achievement across the board. It widens the gap. It is an invisible tax that lands hardest on the students least able to pay it.

The portable idea

Carry this model away: a stressor's damage depends on the resilience of the system absorbing it. The shock can be identical while the outcome is not, because the buffers differ. Slack absorbs; a system stretched thin transmits.

You'll see it everywhere. A rate rise barely dents a cash-rich firm and bankrupts an over-leveraged one. A heatwave is an inconvenience with air conditioning and a killer without it. A missed paycheck is a nuisance with savings and a catastrophe living week to week. The shock is constant; vulnerability converts it into harm.

This reframes how to fight inequality. You don't only attack the shocks — cleaner air, lower rates, cooler cities. You also build buffers into the systems that absorb them. Better ventilation in a poor school does the same work as cleaner air, because it changes how the same air lands.

And it sharpens the question to ask whenever one group is harmed more by something: Is the dose different, or is the resilience different? Usually both — and the resilience half is the one almost nobody bothers to measure.

Distilled from NBER Working Papers