New evidence that lead exposure increases crime

Running tap water is seen in Flint, a city struggling with the effects of lead-poisoned drinking water.

A recent investigation by Reuters found that lead exposure affects kids in communities across the country — not just in high-profile cities like Flint, Michigan. This is worrisome, because elevated blood lead levels in kids have been linked to an array of developmental delays and behavioral problems. More ominously, this could also increase crime. Kevin Drum and others have argued that lead exposure caused the high crime rates during the 1980s and early 1990s. There has been suggestive evidence of such a link for decades, though it hasn’t gained much traction in research or policy circles. But the case that lead exposure causes crime recently became much stronger.

The “lead-crime hypothesis” is that (1) lead exposure at young ages leaves children with problems like learning disabilities, ADHD, and impulse control problems; and (2) those problems cause them to commit crime as adults — particularly violent crime. For many years, the major source of lead in the environment was leaded gasoline: car exhaust left lead behind to settle into dust on the roads and nearby land. When lead was removed from gasoline, lead levels in the environment fell, and kids avoided the lead exposure that caused these developmental problems. About 20 years later, when those kids became young adults, crime rates fell. This, proponents say, is what explains the mysterious and persistent decline in crime beginning in the early 1990s.

It’s an intriguing idea — particularly since we don’t have a better explanation for the big changes in crime rates during this period. Several studies have found correlations between lead exposure and crime, at varying levels of geography (from neighborhoods to nations). But correlation, as we all know by now, does not imply causation.

The main challenge in measuring the effect of lead on crime is that lead exposure is highly correlated with a variety of indicators related to poverty: poor schools, poor nutrition, poor health care, exposure to other environmental toxins, and so on. Those other factors could independently affect crime. The challenge for economists has been to separate the effect of lead exposure from the effects of all those other things that are correlated with lead exposure. A true experiment — where some kids are randomized to grow up with high lead exposure and others not — is out of the question. So economists have gone hunting for natural experiments — events or policies that divide otherwise-similar kids into comparable treatment and control groups.

And they’ve found them. Three recent papers consider the effects of lead exposure on juvenile delinquency and crime rates, using three very different empirical approaches and social contexts. All have plausible (but very different) control groups, and all point to the same conclusion: lead exposure leads to big increases in criminal behavior.

One of these papers considers the aggregate effects of lead exposure on city-level crime, using U.S. data from the early twentieth century. The authors, James Feigenbaum and Christopher Muller, noted that one of the primary ways individuals were exposed to lead during this period was by drinking water pumped through lead pipes. But not all cities had lead pipes. If a city was far from the nearest lead refinery, it would likely have pipes made from another material. Comparing places with and without lead pipes might allow us to estimate the effect of lead exposure on crime, but we’d worry that places near lead refineries are systematically different in some way that could confound our estimates: perhaps they’re subject to more pollution, or are wealthier. To address this, the authors exploit another interesting fact: lead only seeps into water when the water is acidic. This sets up a nice natural experiment that sorts otherwise-similar cities into the treatment and control groups we need. Those with lead pipes and acidic water are the treatment group (their populations were exposed to lead in the drinking water). Cities with lead pipes but non-acidic water, and cities with acidic water but non-lead pipes, are the control groups. These control groups account for the independent effects of lead pipes or acidic water — and whatever characteristics those features are correlated with. Using this experiment, the authors measure the effect of lead exposure on homicide rates lagged by 20 years (to give the kids exposed to lead time to grow up). They find that exposing populations to lead in their drinking water causes much higher homicide rates 20 years later, relative to similar places where kids avoided such exposure.

They find that exposing populations to lead in their drinking water causes much higher homicide rates 20 years later, relative to similar places where kids avoided such exposure.

This evidence on city-level violent crime is more compelling than previous correlational studies, but perhaps it would be even better to compare similar kids who live in the same community. This would allow us to control for more factors that might independently drive criminal behavior.

The next paper does just this, using data from more recent years. Anna Aizer and Janet Currie link data on preschool blood lead levels with data on school suspensions and incarceration, for children born in Rhode Island between 1990 and 2004. They note that kids who happened to live closer to busy roads within a neighborhood are more likely to have high blood lead levels, because the soil near those roads was still contaminated due to the use of leaded gasoline decades ago. This was especially true for kids born in the early 1990s, as environmental lead levels have fallen over time. They use those kids as the treatment group (high lead exposure) and similar kids who lived on other roads, as well as kids who lived on the same roads in later years, as the control group. These kids look similar in most other ways — they attend the same schools, their parents have similar incomes, and so on — so we would expect them to have similar outcomes. But Aizer and Currie find that being exposed to higher levels of lead increases kids’ likelihood of suspension from school as well as (for boys) the probability of being incarcerated as juveniles. The magnitude of their estimates suggest that the reduction in lead exposure due to the switch to unleaded gasoline may indeed explain a substantial portion of the decline in crime in the 1990s and 2000s.

The third paper comes at the lead-crime hypothesis from a different direction, and asks whether government programs that aim to reduce lead exposure can protect kids from lead’s negative effects. Stephen Billings and Kevin Schnepel measure the effect of CDC-recommended interventions for kids with elevated blood lead levels. Kids who test above a certain lead level twice in a row are provided intensive services  — including lead abatement in their home and nutritional counseling to mitigate the effects of lead exposure. The reason two tests are required is that blood lead tests are extremely imprecise. There are therefore a lot of kids who test over the threshold once but not the second time, for reasons other than their actual lead exposure. Billings and Schnepel use the noise in these test results as random variation that divides kids into treatment and control groups: kids who tested over the threshold twice get these services, while kids who tested over the threshold once and then just below the threshold the second time do not. The intuition is that these kids have similar blood lead levels, but due to random noise in the test, some are treated and others are not. By comparing what happens to those two groups of kids, Billings and Schnepel are able to measure the effects of CDC-recommended interventions on kids’ outcomes.

They use data on kids born between 1990 and 1997 in Charlotte-Mecklenburg County, North Carolina. The data include results of blood lead level tests as well as school records and adult arrests. They find that, relative to the control group, kids who receive the intervention exhibit substantially less antisocial behavior, including suspensions, absences, school crimes, and violent crime arrests. These results are striking for two reasons: (1) The kids’ blood lead levels — low by historical standards — are high enough to affect their behavior and put them at risk for suspensions and arrests. (2) The CDC-recommended interventions have a big impact and can substantially mitigate those risks. The authors conclude: “It is likely that increasing the frequency and intensity of intervention for lead-exposed children will yield a profound return considering the potential long-term effects of lead on health and human capital.”

These new studies give us compelling evidence that ongoing lead exposure in communities across the country will have long-term costs to society. But they also provide evidence that we can do something to help kids who have been exposed to high lead levels, and that the benefits of such interventions far outweigh the costs. (Billings and Schnepel estimate that for every $1 invested in the intervention they studied, society yields a return of nearly $1.80.) If President Trump is serious about reducing crime rates, this research suggests he should dramatically expand these programs. It would be a smart investment in public safety.