Childhood lead exposure: Does it lead to crime?

The “lead-crime hypothesis” argues that childhood exposure to lead is a primary driver of criminal behavior from childhood into adulthood. The reasoning behind the theory is that childhood lead exposure correlates with behavioral traits (e.g., aggression and impulsivity), which can increase motivations for criminal behavior. The idea has circulated among academics and the general public. Several careful well-designed studies have found that lead exposure affects crime rates, but debates continue regarding how strong the relationship is. As current levels of lead exposure are already dramatically lower than they were decades ago, it is unlikely that major improvements in nationwide crime rates would result from further eradication of lead pipes and service lines.

Lead is a neurotoxin that can penetrate the blood-brain barrier and interfere with brain development, particularly in children. For example, reductions in IQ and self-control are two documented effects of lead exposure. Thus, it is definitely plausible that childhood exposure to lead could increase criminal behavior insofar as it can change a person’s biology. However, the extent to which lead affects criminal behavior remains contentious – some claim it has been the primary driver of nationwide violent crime, while others argue that the relationship is more complex and moderated by other factors.

The lead-crime hypothesis began after researchers noticed strong aggregate correlations between lead exposure and rates of some reported crimes. The content of gasoline rose steadily from the 1940s through 1970 before falling quickly and reaching approximately zero by the end of the 1980s. In the peak of America’s lead problem (around 1970), double-digit blood-lead levels weren’t considered elevated unless they exceeded 60 micrograms per deciliter (µg/dL). By 1991, lead exposure reduction efforts had been fairly successful. At this time, levels weren’t considered elevated unless they exceeded 10 µg/dL. This threshold decreased even further in 2012 to 5 µg/dL.

During the 1970s and 1980s, there was a consistent rise in the number of violent crimes recorded in the Uniform Crime Report (UCR). This trend continued through the end of the 1980s before falling dramatically in the 1990s — a trajectory that was remarkably similar to the trajectory of lead exposure rates (except it was lagged by 22 years). Regarding homicides, though, trends remained relatively steady throughout the 1970s and 1980s with some intermittent increases and decreases before falling in the 1990s. The 1990s saw a major crime decline across all categories — a decline that coincided with the time that many people exposed to lead as children would have “aged out” of crime (i.e., 22 years later). Based on this correlation, people theorized that early-childhood lead exposure was a primary driver of crime. So the theory goes: following the removal of lead from petroleum gasoline and paint (coupled with new environmental laws), lead poisoning declined, which coincided with a crime decline about 22 years later.

A 2000 study by Rick Nevin argued that lead poisoning of children can increase their motivation for violent crime during adolescence, based on temporal correlations between shifts in children’s blood levels and subsequent changes in IQ. Results showed that long-term trends in population exposure to lead (mostly due to the lead found in gasoline and paint) were consistent with subsequent changes in violent crime, including murder rates. However, this link between lead exposure and crime is correlational and does not necessarily indicate causality. The data were reported in the aggregate as well, which has inherent limitations that I discuss in more detail below.

In 2007, Nevin extended the analysis further by examining lead exposure and crime correlations across a series of developed countries. The end of leaded gasoline occurred in different places at different times, allowing Nevin to examine whether shifting exposure patterns corresponded to changes in crime rates. This study showed a strong association between preschool blood lead, IQ, and subsequent violent crime trends over several decades in multiple countries – such as the United States, Britain, Canada, France, Australia, Finland, Italy, West Germany, and New Zealand. While this study was better than the 2000 study, it still had its limitations. Namely, it should be noted that temporal crime trends are affected by a variety of factors, and are notorious for being susceptible to outside influences, such as massive societal or technological changes that also develop over long periods of time. For example, the 1990s crime decline is often attributed to changes in incarceration, policing strategies, and changes in victim reporting rates, police crime data collection, and participation in the UCR program.

John Paul Wright and colleagues also found support for the lead-crime hypothesis in a longitudinal study with results published at multiple points in time, including 2008 and 2021. The study used a cohort of 250 people ages 19-24 in Cincinnati who were born between 1979 and 1984 and were followed for many years. The study found that prenatal and childhood blood lead concentrations were correlated with subsequent total arrests and violent crime arrests. However, critics have argued that the study only examined one community that likely had many sources of antisocial behavior, which might possibly confound the relationship between lead exposure and antisocial behavior. In addition, the study examined results at the aggregate-level which may mask a considerable amount of variability between smaller geographic units (e.g. counties, cities, urban hot spots), making it difficult to geographically target interventions to appropriate areas.

At first, aggregate evidence appears compelling. As described above, Nevin found correlations between lead concentration and crime cross-nationally and nationally while Wright and colleagues also found correlations nationally within the United States. As a result, several researchers have interpreted aggregate-level results as evidence for a causal link between lead exposure and crime. However, many of the studies finding support for the lead-crime hypothesis are based on aggregated data at the county-level, city-level, and census tract-level, which may mask variability between individuals. Aggregate-level data is often easier to obtain, though it can oversimplify a relationship between two variables. Namely, there are many individual-level factors that can interact with lead exposure to produce criminogenic behavior, but it is not possible to examine these differences when results are reported in the aggregate.

In 2008, Alfred Blumstein and Richard Rosenfeld also agreed that there was a clear correlation with time trends for environmental lead levels and violent crime rates, lagged by approximately 23 years. However, they contended that crime trends are impacted by many potential factors, and while lead exposure might be one of them, it is probably not the strongest predictor of violent crime. In short, there are various individual-level mechanisms that may moderate the lead and crime relationship. For this reason, criminologists Robert Sampson and Alix Winter suggested in a 2018 publication that the lead-crime hypothesis might be decent for explaining individual-level differences in criminal behavior but not necessarily for explaining aggregate-level trends. A 2006 publication by Bernard Weiss and David Bellinger explained another key problem with some of the past research, even among the most rigorous designs: exposure to environmental pollutants is not randomly distributed and often related to spatial characteristics and other risks to normal development (e.g. family dynamics, self-control) that are difficult to flesh out.

Generally speaking, as researchers began looking at smaller units of analysis (e.g. counties, census tracts), the lead-crime relationship appeared more complex. Researchers Paul Stretesky and Michael Lynch examined cross-sectional associations in the U.S. at the county-level, publishing results in 2001 and 2004. They found associations between air lead levels and homicide rates and violent and property crime rates, but they also found evidence that resource deprivation (i.e., fewer healthcare options, lack of access to nutritional options) moderated the relationship. The association between air lead levels and property/violent crime was highest in areas that had higher levels of resource deprivation; conversely, the weakest association was found in counties that have low levels of resource deprivation.

To build on past research, Brian Boutwell and colleagues published a study in 2014 that examined census tracts, a smaller geographic unit than used in previous studies. Their statistical models accounted for geospatial correlations among a large sample of over 100,000 children in St. Louis, Missouri; results were then aggregated at the census tract-level. Results showed that greater lead exposure was associated with increased violent, non-violent, and total crime. However, researchers note that the nature of the sample in this study (e.g. one city) does not provide insight as to whether children with the highest lead exposure levels are also the most violent or crime-prone, particularly later in the life-course. This has important implications for policy, as lead concentration levels may be low enough already such that a significant drop in crime would likely not result from further lead reduction efforts.

Research by Kim Lersch and Timothy Hart in 2014 also examined census tracts and used modeling approaches that were able to account for spatial correlations with other factors (e.g., socioeconomic status). They examined the spatial distribution of toxin exposure for over 100 census tracts in Hillsborough County, Florida, finding that toxin exposure was not randomly distributed, but clustered in certain locations. Thus, location clusters were accounted for in a geographically-weighted model. The results showed that the model’s predictive power changed when including geospatial correlations in the analysis — it improved for property crime but not for violent crime. However, the researchers comment that because the study aggregated results at the census-tract level, the results are limited and methodologically complex.

To summarize, the relationship between lead and crime is far from direct; this becomes apparent when one focuses on the specifics of the pathway of how lead exposure leads to crime. Admittedly, the fact that units of analysis (e.g., geographical regions) used in research have narrowed over time from the country-level to the census tract-level is certainly an improvement from past research. In addition, it is evident that subsequent studies strengthened the approach by looking beyond broad nationwide lead and crime trends by examining whether this relationship was consistent across different countries. However, most of these studies had low causal validity (for reasons ranging from small sample sizes to inability to address confounds). The research on this topic has continued to evolve over time to unpack the complicated pathway between lead exposure and crime.

In a 2020 publication by John Laub and Robert Sampson, the researchers posit that the impact of lead exposure is age-graded in nature and might fit into the larger theory regarding crime throughout the life-course. Age-graded theories rely on many individual-level factors that interact with each other, making it a somewhat complex theory, though it might be one of the best theories to explain differences in individual-level criminal behavior. In a 2009 publication, Jesenia Pizarro and colleagues reviewed studies that have examined the individual-level impacts of lead exposure on biosocial factors (e.g., low IQ, attentiveness, hyperactivity, and externalizing behavior problems). They postulate how this could plausibly lead to risky (or criminogenic) behavior. Further, Pizarro and colleagues explain how potential pathways between lead and crime can vary. For example, if lead can impact the brain’s neurotransmitter systems, it might increase the likelihood of children responding to challenges in their social contexts in ways that provoke rejection by parents and larger society. This could then contribute to a child’s externalizing behavior problems and association with the delinquent subculture. Other theories are that lead exposure is related to factors such as family environment and socioeconomic status, which potentially confound the association between lead and crime.

This is consistent with suggestions by researchers Herbert Needleman and David Bellinger in 2001 and 2004 — that the more important outcome from lead exposure is that it interferes with social adjustment and impulse control. Similarly, a study published in 2012 by Howard Mielke and Sammy Zahran examined 1950-1985 fluctuations in lead concentration in different cities (i.e., Chicago, Indianapolis, Minneapolis, and San Diego), and they did find an impact of lead exposure on adult violence. However, they also found that variation in childhood lead exposure was indeed linked with other cognitive and behavioral mechanisms, such as impulsivity, aggression, and lower IQ.

A recent cross-sectional study published in 2021 by Nonhlanhla Tlotleng and colleagues also examined the impact of lead and parental education on aggressive behavior using a sample of 100 children in South Africa. They found that lower parental education and bone-lead concentration were associated with aggressive behavior in children, which could pose adverse effects later in life such as violent behavior. Interestingly though, the primary source of lead exposure was environmental pollution from houses previously painted with lead, yet the pathway between type of housing and bone-lead levels was not statistically significant. Other research by Eric Dubow and colleagues in 2009 showed that parental education can affect subsequent occupational success, mediating the lead-crime relationship. Their study followed over 850 people for 40 years in a rural county in New York. The research found that lower levels of parental education negatively impacted children’s occupational success, but was moderated by family interactions, child aggression, and teenage aspirations. These variables were also correlated with areas of concentrated disadvantage and higher risk environmental lead exposures, making this relationship even more complex.

Routine activity and opportunity for criminal behavior are also important when considering individual-level differences (particularly age-graded impacts) that are not accounted for in many studies. As discussed in a 2010 publication by Graham Farrell and colleagues, childhood lead exposure seems limited to explaining United States’ violent crime trends because there are major factors that could interfere with the relationship, such as routine activity, criminal opportunity, age, urbanization, and population density. Alternatively, it is possible that age, urbanization, and population density can impact one’s routine activities and access to opportunities for criminal behavior, which in turn can change one’s motivation for criminal behavior.

A 2021 meta-analysis by Anthony Higney and colleagues compiled the results across many studies and their analyses accounted for the strength of the study designs. When the authors compiled results from just the strongest studies, they statistically estimated how much of the 1990s decline in homicides could be explained by lead exposure. With the stronger studies, they found that lead exposure could explain 0%–5% of decreases in homicides. In contrast, when the weaker studies were also included in the results, they found that lead exposure could explain a far wider range of 3%–36%, which essentially means the margin of error is larger in the latter. The results do not necessarily eliminate a link between lead and crime, though they do suggest the effect to be smaller than typically believed.

As I have shown, there has been a growing amount of literature challenging the lead-crime hypothesis. Some of the major criticisms are that rigorous studies are in the minority, with most of the lead-crime research relying on less compelling research designs. Most of the early studies exploring this relationship examined correlations between aggregate crime trends and lead exposure over time, which are not adequate in explaining causality. Further, they do not always adequately address other variables that occur at the neighborhood-level (e.g., concentrated disadvantage) and individual-level (e.g., parental education).

Also worth considering are the limitations inherent in officially recorded crime data (i.e., official police reports) that is utilized by many past studies. There are numerous limitations with UCR data; namely, the numbers reflect crimes that are reported to the police, and do not include crimes that are unreported or not documented properly. Further, UCR data is typically reported at the aggregate-level, making it more difficult to track changes between individuals. Moreover, the aggregate-level relationship found in UCR data has been challenged by Lauritsen and colleagues, who found that the relationship between lead exposure and lagged crime rates are not found in victimization data (as opposed to UCR data). This finding points to the importance of including both official and unofficial measures of crime when examining consequences of lead exposure. Similarly, a 2016 publication by Richard Rosenfeld and David Weisburd discusses how the timing of the crime drop has differed according to the data source used to document it. In contrast to UCR rates (which declined in the 1990s), the NCVS rate of serious violence began to decline well before 1990. In addition, authors found that several factors such as unemployment and inflation were related to the NCVS trend in serious nonlethal violence but unrelated to the UCR trend. This result challenges the prior research on the lead-crime hypothesis that is based on UCR data. Another problem with measures that is common among these studies is that outcomes pertaining to delinquency are not necessarily indicative of criminal acts (and do not always correlate with arrests), and are typically measured using parent- and teacher-ratings of child behavior.

Intriguing as some past findings are, there is still skepticism over whether lead exposure really contributes to delinquency, and to what extent. While lead exposure is clearly dangerous (especially in high levels), it does not seem to be a primary driver of the crime drop in the 1990s. It is fairly clear that lead can impact behavioral and cognitive mechanisms, but whether these problems translate into crime is another matter. Researchers have yet to pin down a direct mechanism that might contribute to delinquent behavior. To summarize, lead exposure in childhood may have played a small role in rising and falling crime rates in the United States but it is unlikely to account for a very high percentage as suggested in past research. At present, research needs to explore the varying methods used to establish the relationship between lead, delinquency, and crime.

Considering that the regulations on lead exposure are already much better than they used to be, it seems overly optimistic that reducing lead levels further would lead to a dramatic decrease in crime. Put differently, at low thresholds of lead exposure, further reductions may no longer affect the outcome of interest (e.g., neurocognitive problems, aggression). Rather than attempting to replace all of the lead pipes nationwide, it might make more sense to target efforts toward at-risk neighborhoods, or toward the small percentage of people who still have high blood-lead levels. A 2018 natural experiment found that children exposed to lead are amenable to efforts that have been done to reverse damage caused by earlier lead exposure, which could be an option targeted toward certain individuals.

Overall, efforts to reduce lead exposure in the United States have been fairly successful. Given how much exposure to lead has already decreased over time, policymakers should not expect further reductions in lead levels to correlate with massive drops in crime similar to those seen in the 1990s.  Other studies have not been able to establish the amount of lead exposure that is considered significantly impactful on a child’s well being, so it is hard to know whether reducing lead levels even further below the threshold of 5 ng/mL would have a dramatic impact on crime rates.