Debunking the lead crime hypothesis
In another article, I wrote about theories for why crime in the US was worse in 1990 than it is today.
Economist Rick Nevin thinks that it was a totally unrelated effect, the result of a generation being poisoned by leaded gasoline and experiencing increased aggression.
Nevin even disputes the common view that violence in the 1920’s was caused by alcohol prohibition — he thinks that was the result of using leaded paint.
It’s a neat theory, which he further supports by correlations in other countries. Leaded gasoline was phased out at different times in different countries, and Nevin points to several where the curves line up.
Crime peaks a little early in Finland, France, and Germany. Britain and Australia both have an extra crime spike after the lead spike. Still, things roughly line up.
We have data showing that childhood blood levels were higher while leaded gasoline was in use. There are experiments showing that lead exposure causes brain changes. And the crime rates seem to correlate, in many western countries.
It sounds possible. But there are a lot of reasons to be suspicious of the theory:
1. It didn’t happen in Japan.
Japan is one of the most densely populated countries in the world. If ever there would be a city where dense traffic would give the population lead poisoning, it would be Tokyo.
Japan used leaded gasoline from the 1920’s until 1980. Peak use was probably in the 60’s and 70’s. Japan saw no spike in murder rates, 20 years after that, in the 80’s and 90's:
You could argue that Japanese culture is different, and that murder is rare.
You could also argue that Japan has extremely strict gun control. Firearms are mostly illegal (you can still buy a shotgun but it takes a very lengthy process), and gun ownership is very rare.
Maybe Japanese society doesn’t tolerate violence and lead poisoning just made Japanese men into depraved consumers of hentai.
Also, there has been a decline in murder over time, even if there was never a spike, so maybe something did improve as lead was phased out.
Nevin does claim that the United Kingdom had a lead induced crime spike, and the UK doesn’t have a lot of guns. But the UK still has 10 times more guns than Japan does:
Hong Kong is about as well armed as England, and it looks like the murder rates also don’t fit the lead-crime hypothesis (there’s a peak in murder in the 70’s, not the 90's):
I don’t think finding one or two counter-examples disproves the whole theory, so we need to look at more countries.
2. Most of Latin America doesn’t fit the pattern.
Let’s start with Mexico.
Murder stayed low in the US, in most of the post lead era. Mexico saw a surge of murders starting in 2008.
Rick Nevin wrote a report on lead exposure and murder in Latin America. He points out that lead exposure in Mexico peaked in 1988.
It’s from a survey with only 3 data points, so we don’t know the exact peak year, only that things were in decline by 1990.
The crime wave started in 2008. In other countries, Nevin found crime peaked 19 years after the lead-crime peak. That should be in 2007. So we should have expected a long rise in crime leading up to a 2007 peak. Instead, we saw a long decline in crime leading up to an abrupt 2008 start of the crime wave.
Nevin predicted that Mexican crime would peak in 2011 and then decline.
Violence didn’t end after his prediction, it declined a little and then hit a new peak in 2017. That was 20 years after 1997, when leaded gasoline had already been phased out.
There’s a much better explanation for all these murders in Mexico: drug cartels were at war with each other.
The brief drop in murder, from 2011 to 2014 had something to do with the Sinaloa cartel gaining control. After the cartel’s leader, El Chapo, was captured and extradited to stand trial in the US, a new surge of violence started as others started fighting for territory and control.
We can tell that the murders in Mexico weren’t driven by lead because of where they happened. Here’s the year that Nevin said would be the peak of Mexican lead poisoned crime:
Mexico City has intense traffic, if this were lead poisoning, you’d expect the most violence there. The violence actually shows up near the US border, and in specific regions that were affected by cartel violence.
The highest violence was in Sinaloa cartel territory:
An event like a drug war can overwhelm any other subtle crime trends. Both of the 2 highest crime periods in US history also lined up with drug wars (prohibition in the 20’s and crack cocaine in the 80's).
If an author tries to explain a foreign drug war with lead poisoning, we should be suspicious when he tries to use lead to explain crime during a domestic drug war.
Here’s a longer term view for Mexico:
The US had a peaceful period in the 50’s before crime picked up from 1965 to 1994, so Nevin can call that the era prior to lead. Mexico had a violent past, before leaded gasoline was in use.
There is, perhaps, a 1970 to 1990 crime increase in Mexico that sort of resembles the US and might line up with increasing lead poisoning. But there’s also a crime decline from 1990 to 2008 which the lead theory doesn’t predict.
Let’s move on to Brazil, which has worse crime than either the US or Mexico:
Brazil banned leaded gas earlier than the rest of Latin America, because they wanted to support Ethanol.
Peak lead use in Sao Paulo was in 1980, and murder fell 20 years later:
The situation looks unique to that one city, however. Crime in Rio de Janeiro didn’t drop at the time, and crime in the rest of the country kept climbing.
It’s unclear when the rest of Brazil, outside of Sao Paulo hit peak lead emissions, but we do know that leaded gas was fully banned by 1989. Crime should have been falling nationwide by 2009 at the very latest. Instead, murder rates kept going up:
It seems that Nevin has picked the one city in Brazil that fits the lead-crime hypothesis. Other researchers say that Sao Paulo’s success was simply the result of more aggressive policing.
Rio’s crime rate possibly started to catch up in an effort to clean up the city’s slums prior to the 2016 olympics.
In a 2nd post about Central America, Nevin notes crime spikes in 3 other countries and also attributes them to lead:
El Salvador’s murder rate was among the worst in the world, in 2015.
Why was it so high? Was there a sudden surge in driving, 20 years earlier? Wikipedia explains:
“The national murder rate was the worst in 10 years, with 481 people murdered following the collapse of a truce between rival gangs”
Nevin is ready to attribute every crime spike and decline to his theory, even where it doesn’t really fit the curve, and even where there’s a good competing explanation (like cartel violence or a national gang rivalry). It should give you pause as to whether he’s been just as careless when thinking through the lead/violence connection in the United States.
As a general rule, Latin America had increasing crime past 1990, and continued to do so more than 20 years past the end of leaded gasoline use.
It’s true even in relatively stable places like Costa Rica:
It’s even true in the US territory of Puerto Rico:
Perhaps you could argue that Latin America has large factors that drive crime that obscure the impact of lead.
It’s an odd argument, to say that Japan is too peaceful to see the effect of lead while Latin America is too violent. It is still possible. Perhaps only Europe and North America have just the right balance to see the lead-crime effect. Crime is the result of many factors, maybe we need some perfect middle ground where things are neither too peaceful or too violent for the lead effect to show up, and only western countries provide this environment.
Kevin Drum writes about the lead-crime hypothesis, saying:
Nevin collected lead data and crime data for Australia and found a close match. Ditto for Canada. And Great Britain and Finland and France and Italy and New Zealand and West Germany. Every time, the two curves fit each other astonishingly well. When I spoke to Nevin about this, I asked him if he had ever found a country that didn’t fit the theory. “No,” he replied. “Not one.”
Okay, maybe Nevin just forgot to look at Japan or Hong Kong and looked really carelessly at Latin America?
There may be countries in Europe that don’t fit, as well, I haven’t checked in detail. At a glance, there are a few that look suspicious to me: maybe Spain, Portugal, Greece, Poland, or Sweden.
I don’t think any of this has disproven the lead-crime hypothesis, I’ve only shown that Rick Nevin is far too eager to apply it everywhere.
3. Crime rises at the wrong time for age cohorts in the US.
Leaded gasoline use rose in the US from 1940 to 1955. About 20 years later, violent crime surged:
Murders in the US went up in part because of demographics — there was a baby boom starting around 1945, birth rates went up 50%, peaking in 1960:
Young men commit most crimes, so when there are more young people relative to older ones, the per capita murder rate goes up. 20 years after the baby boom started, crime went up.
That doesn’t explain the entire rise in crime. The rate at which people committed murders also increased. Something else happened to make people more violent, starting around 1960–1965, and what caused it isn’t clear. Based on the timing, it could be lead poisoning.
If it were lead, you’d expect a few other things to be true.
First, you’d expect it to have some effect on all racial groups. If lead makes people more violent and aggressive, the black murder rate and the white murder rate should both increase at the same time and peak around 1990.
Second, you’d expect the increased murder rate to show up first in younger cohorts of murderers and then work it’s way into older groups, as each group was successively poisoned in childhood. When the lead pollution declines, starting around 1970, teens should start getting less violent starting around 1985, young adults around 1990, and older groups after that.
Can we test these hypotheses with the data?
The murder rate doubles for both white and black men, starting around 1965. This looks good for the theory, so far. Though this is a graph of murder victims, not of perpetrators, so we can’t say for sure.
Did the violence in 1970 start with the youngest cohort? (those kids newly poisoned by lead?)
Again, the CDC data is hard to parse, because it tracks victims, not killers.
In 1972, black murder victims rose in every age group, relative to 1962:
This isn’t looking good for the lead-crime hypothesis, it looks like murders have gone up for all ages, not just young people. But it’s still possible that a new generation of violent young men was killing adults of all ages.
We really need to look at the age of the murderers, not the victims.
After a long search, I finally found the data in this paper on lead and crime:
This doesn’t look right at all, for the lead-crime hypothesis. If lead were the explanation, the murder rate should start rising first in teenagers. Then, 5 years later, it would rise in the 20–24 year old group. Then, after 10 years, the 25–29 year olds would become more violent, and so on, as the lead poisoned generations grow up.
That’s not what happens. Instead, every group becomes more violent in the 60’s, all at the same time.
This is one of the key points that changed my opinion on the lead-crime hypothesis. In the first sections of this article, I was just looking for countries that didn’t fit the pattern, finding ways the idea gets overused, but I still thought it could be a leading cause of crime in the US.
Here, we can clearly see that lead can’t be the reason that crime rose in the 60’s. Generations didn’t get poisoned, one by one. Society got more violent for every age group, all at the same time, starting sometime around 1965.
After 1976, we have clearer data and can see trends of homicide offenders by age group and by race:
There’s a huge spike of crime from 1984 to 2000 that’s limited to younger men and is especially large among young black men. This is also poorly explained by lead poisoning.
If lead were the explanation, the 14–17 year old group should hit peak violence around 1986 (16 years after peak lead) and the 18–24 year old group should hit it around 1990 (20 years after peak lead).
In reality, both 14–17 year olds age groups hit peak violence around 1993, and both trend up and down at the same time.
If this were a linear effect of lead poisoning, violence for the teenage groups should trend up and down over decades. Instead, it spikes up and down in only about 12 years.
If lead were the explanation, the 25+ age group shouldn’t hit peak violence until 1995 or even later (as more generations of men grow up poisoned). In reality, the 25+ year old murder rate peaks in 1980. That’s true for both white and black men.
The best explanation for the teenage crime wave is not lead but crack cocaine, which arrived to American cities in 1984. The reason why it’s limited to younger men is unclear — one author theorizes that mass incarceration and the war on drugs had already been going on for long enough that many of the older dealers had been incarcerated, leaving younger men to take over drug distribution.
For black women, the murder rate is highest in 1975, well before peak lead poisoning. It continues to decline after that.
White women were never particularly murderous, they never saw a crime spike or decline.
There’s a fairly simple explanation for women’s murder rates. In 1975, black women used to kill their husbands more often:
White women have been been killing their husbands less often, as well, but the starting rates were much lower.
I would speculate that women started killing their husbands less because divorce got easier to obtain, not because of lead poisoning. Also, many black women simply didn’t get married after the 70's.
If we expand the analysis to look at other races, we see that Hispanics had a murder spike during the crack epidemic but Native Americans did not:
Should we conclude that reservations were unaffected by leaded gasoline? It’s possible, maybe tribal living is much less dense than urban living. But it also seems possible that the crack distribution wars didn’t happen on tribal land.
This data also shows less of a spike for white men, perhaps because the prior BJS data combined white and hispanic men. Crack was an urban phenomenon that affected both black and hispanic men.
It seems clear that the murder spike from 1985–1997 was due to the crack epidemic and not due to lead poisoning.
It’s not as clear what caused the murder increase from 1965 to 1980 but it mostly wasn’t lead. That was a turbulent era. The war on drugs started. There were riots related to civil rights and protests against the vietnam war. There was a large black migration from the south to northern cities. All of these are plausible factors in the rise in crime.
4. Murder spikes in the US vary by weapon.
We already have enough data to say that lead wasn’t the main factor driving crime between 1965 and 1995. But we can test or refine the lead-crime hypothesis a bit more by breaking down murders by category.
If lead made people more impulsive and aggressive, you might expect that they would be both more likely to knife each other or shoot each other. In fact, the crack epidemic murders of the late 80’s and early 90’s were mostly all shootings with pistols, with no increase in murders with other weapons.
I don’t think this argument is conclusive. Maybe lead makes you impulsive and it’s just easier to shoot someone when you’re mad. But I think it’s still an interesting data point — it points more to these being premeditated shootings (like over drug turf) than purely impulsive crimes committed because of brain damage.
Most of the variation is driven by handgun shootings:
It does look like there has been some decline in knife murders since 1990.
If lead poisoning makes people impulsive enough to kill, is the effect limited to killing strangers? Does it also show up in killings of friends or family members?
We have some data here, spousal killings have been declining since 1975 (as mentioned before, the biggest change is from less black women killing their husbands, but spousal killings also went down for every race and gender):
Homicides of other family members are mostly flat for both black and white people (On a per capita basis, it’s declining since 1975).
Again, this doesn’t invalidate the theory, but it clarifies it in an odd way — lead would have to make people violent enough to kill strangers but not impulsive enough to kill family.
We can try to clarify the effect a bit more by breaking things down by murder of strangers or acquaintances:
It’s hard to tell if there’s something here. Maybe there’s a drop in killing friends around 1993. But the relationship between murderer and victim is unknown for most crimes, so we can’t really say.
5. Crime didn’t drop enough after lead was fully phased out
Preschool blood lead levels have been dropping since 1970. The average child has less lead poisoning than in a century. Anyone under 35 has grown up without much lead poisoning. The murder rate has not kept dropping:
Violent crime should be back at 1960’s levels, but it remains high:
And crime actually started rising again, in 2015 and in 2020. Today we’re back at a 1980 level of violent crime.
The murder rate in 2010 went down to about where it was in 1960, before leaded gasoline.
Strangely, though, aggravated assaults were still as high as they were in 1980.
Rapes are also just as frequent as 1980.
Why did murder drop back to 1960 levels while other crimes stayed high?
It’s possible the definition of assault has changed over time and it includes more minor crimes. It’s also possible that reporting of rape has gone up.
But another theory holds that murders have gone down because emergency medicine improved. Both car crashes and aggravated assaults became more survivable from 1960 to 1990, at about the same rate:
It’s possible that the number of shootings is actually as high today as it was in 1980, but that more victims survive.
Robbery rates are also included in violent crime totals, and those are down. I’m not sure how to explain that discrepancy with assault or rape. Maybe robbery is less lucrative now that less people carry cash?
Nevin made a plot of several crime rates in a single graph:
It’s an inconsistent picture. It could be explained either by reporting changes in assault and rape or by emergency medicine getting better and saving more lives of people who’ve been assaulted.
Crime rates for young people also didn’t fall as fast as the lead-crime hypothesis would suggest.
After the crack epidemic ended, homicide returned to normal for 14–17 year olds. As of 2005, crime rates were still elevated for men 18–24 years old. That’s 35 years past the peak of lead and 20 years after lead returned to its levels from before leaded gasoline.
Murder rates in some cities were also as high in 2020, as they were at the peak in 1990:
Crime has big differences from city to city. Nevin’s theory is that lead caused the synchronized peak in crime in 1990, but it doesn’t explain why some cities like New York and LA have remained safer while others like Chicago and Philadelphia have returned to their previous high murder rates without any new lead exposure.
6. Other countries didn’t have a 1920’s crime wave.
Nevin blames the 1960–1990 crime wave on leaded gasoline and the 1920–1930 crime wave on leaded paint.
Many other countries used leaded gasoline and many had crime waves around the same time. Other countries used leaded paint without seeing a crime spike.
England saw a crime drop around 1917 (men were away at war) and little else, from 1900 to 1930.
It’s hard to find 100 year old crime rates from around the world to test this theory elsewhere. Again, this doesn’t prove Nevin’s leaded-gasoline theory wrong, but it calls into question his lead paint theory for 1920’s crime.
7. IQ scores weren’t affected enough to explain the crime increase.
Nevin’s theory holds that childhood lead exposure reduces your adult IQ.
As proof this happened widely, he points to a drop in SAT scores:
The curves don’t quite line up — math SAT scores bottom out in 1981 when lead poisoning would predict 1987 — but it’s an interesting piece of evidence.
One problem here is that the SAT is not given to every student in high school but only about one third of the students that go on to college. If the pool of students taking the test changes, scores can go up or down.
IQ researcher James Flynn wrote about this decline in SAT scores:
When the College Entrance Examination Board established an advisory panel to analyze the score decline, the latter estimated that about half of the score drop had to do with the broadening of the candidate sample, the other half reflecting a downward trend in the general population itself. Although the panel did not disclose its evidence, they said their estimate rests on a “relatively firm statistical basis” and they do cite two compelling facts: from 1963 to 1970 the score decline was accompanied by a broadening candidate sample, but from 1970 to 1977 the score decline was even worse despite the fact that changes in the candidate sample were insignificant. They also note that since 1970, the score decline has shown up within all categories of SAT takers, within students with good high school grades as well as bad, within offspring of high-income families as well as low-income, and within whites as well as blacks.
Flynn says the SAT math test had a standard deviation of 130. Nevin thinks that math scores went down by 30 points during peak lead poisoning. That’s equal to 3.5 IQ points lost.
Using the college board’s adjustment for different test takers, we might conclude that only 1.75 IQ points were lost.
We can compare these numbers to what’s predicted by Nevin’s lead poisoning theory. He proposes a dose-dependent relationship between lead and IQ:
His theory says that lead exposure was 5 mcg/dL in 1940 and 20 mcg/dL in 1964. That should cause a difference of maybe 6 or 7 IQ points based on Nevin’s theory, higher than either number we’d get from the SAT results.
The data doesn’t quite line up. But, maybe the kids who became murderers received worse than average lead poisoning and the kids taking the SAT received less.
Another dataset we can check are NAEP test scores, which are given out to a large number of US students in 4th, 8th, and 12th grade. 12th grade NAEP scores dipped in 1982, around the same time as SAT math scores declined:
It’s a brief decline, but let’s assume that it’s caused by lead pollution.
I think the standard deviation of the NAEP test is 39 points. The NAEP score drops by 8 in 1982. So that equates to about 3 IQ points.
The data on younger kids possibly confirms Nevin’s theory: 13 year olds and 9 year olds scored worse in 1973. Those differences are bigger, showing 4 and 8 IQ points lower than today’s numbers. And it looks like scores maybe stayed lower for the youngest groups.
I can imagine other theories to explain the data. Maybe we emphasized standardized testing more as time went on, so 9 year olds got better at taking math tests? Maybe nutrition improved and kid’s brains started growing faster, so the younger age test scores improve but the Age 17 scores stay flat. The age of puberty has also gone down, over time. Maybe kids are growing up faster, in general.
Either way, let’s continue to assume this is all the effect of lead, and there was an IQ drop of somewhere between 2 and 4 points for young adults.
How much should the crime rate rise if IQ drops by that much?
I went looking for studies that correlate IQ with crime. The best study I found was from Sweden, with almost 1 million participants and a high level of certainty:
If IQ goes down by 15 points, crime doubles.
It looks like the results are the same in US data, a difference of 15 IQ points is about enough to double the odds of incarceration:
So, for a drop of 2 IQ points, crime should only go up 13%.
For 4 IQ points, crime should go up 26%.
Crime rates actually went up more like 100% from the 60’s to 70's. The IQ loss is not enough to explain this difference.
In contrast to SAT and NAEP scores, James Flynn found an entirely different effect when looking at IQ scores. Flynn thinks IQ is rising over time, and every generation gets a little bit smarter. He thinks there was an IQ increase of 7 points from 1948 to 1972, during a period of time where crime went up.
He also found an increase of 3 points from 1932 to 1948.
Flynn’s results don’t line up with SAT scores, and they go in the wrong direction to explain changes in crime. If IQ has been increasing since 1940, crime should have gone down through the 60’s and 70's.
Either the Flynn effect isn’t real, or IQ changes were not the primary factor driving crime up in the 60’s and 70's.
8. Cohort studies are inconclusive.
The best way to actually test the lead/crime hypothesis would be a randomized experiment where we took a bunch of kids, randomly exposed half to lead for years and then followed them through life.
Since we can’t ethically do that experiment, researchers have instead tried to track children through life, measure their blood lead levels and see what the outcomes were in adulthood.
The best study done comes from New Zealand. Researchers followed 579 children from childhood to adulthood. Blood lead levels were measured once, at 11 years old, and then researchers tracked the kids until they were 38.
It seems like a cleaner experiment than you’d see in the United States, because the kids were racially homogenous and because the lead distribution was similar for rich and poor families. Some kids in each socioeconomic group had higher blood lead levels and some had lower:
Following the kids through life, the study found that:
Criminal conviction was more prevalent and more frequent at higher BLLs: 8 of 33 participants (24.2%) with a BLL of 5 μg/dL or less had a criminal conviction compared with 24 of 82 participants (29.3%) with a BLL above 15 μg/dL
So, the lead poisoned kids are more criminal. We’re comparing rates of 24% or 29% criminality among fairly small numbers of kids. This could be statistical noise or it could be a meaningful effect.
Suppose it is the effect of lead. Could we say that going from 1945’s lead levels in the US to 1960’s made people 20% more criminal?
It doesn’t seem like enough to explain the 100% increase in crime rates.
The study also tries to make a linear fit of the data:
Each 5-μg/dL higher BLL was significantly associated with a 1.29 increase in the odds of criminal conviction (95% CI, 1.06–1.56; P = .01). After controlling for sex, the association between BLL and conviction was slightly attenuated and not statistically significant (adjusted OR, 1.23; 95% CI, 1.00–1.51; P = .05).
There’s a wide confidence interval, the study says 5 ug/dL of lead poisoning causes a 23% increase in crime. But it might cause no increase or that it could cause a 51% increase.
If you pick the middle of that range, lead explains about half of the increase in US crime in the 60’s and 70's.
If you pick the higher end of that range, lead could explain the entire crime increase. And if you pick the lower end, lead doesn’t explain any of it.
Basically, the study wasn’t large enough to prove what’s true here. Unlike the Swedish study of IQ and crime, which used 800,000 kids, this study only used 500, so the error ranges are much larger.
Another study done on the New Zealand cohort looked at changes in IQ:
After adjusting for maternal IQ, childhood IQ, and childhood socioeconomic status, each 5-µg/dL higher level of blood lead in childhood was associated with a 1.61-point lower score (95% CI, −2.48 to −0.74) in adult IQ
That’s means the US should have lost somewhere between 1.5 and 5 IQ points from 1960 to 1980. Looking at SAT scores, it was between 1.75 and 3 points. The theories line up nicely.
When we looked at IQ drops, they didn’t seem high enough to explain a 100% increase in crime.
Maybe lead affects the brain in other ways besides IQ.
A third survey on the New Zealand cohort looked at personality traits. Children with high blood lead are slightly less agreeable, more neurotic, and less conscientious. They’re slightly more likely to have psychopathologies, like internalizing thought disorders. Lead had no changes on extroversion or openness to experience.
Perhaps if you combine lower agreeableness and lower IQ, that makes a higher crime spike than low IQ alone.
Do we have any evidence, anecdotally or otherwise, that people were less agreeable in the 70’s and 80’s than they are today?
Doing a study like this requires monitoring kids for decades, taking samples when they were young. I’m not sure we’ll ever get better data here, but you could do a meta-analysis of studies like this. This one looks like lead poisoning reduces IQ and increases criminality a bit. The exact rate of increase is hard to pin down, with the size of the study.
It’s probably not strong enough to explain the rise in US crime, but it could explain some fraction of it.
9. Nevin’s data on blood lead levels might not even be correct.
Most of the childhood blood lead curves that Nevin uses are estimates, not measurements. For instance, see this graph from the US:
The actual measured data points are in red or yellow. All the points drawn in black are just Nevin’s estimates. Everything prior to the 60’s is estimated, not measured. Nevin’s graphs give the impression of a very clean fit between rising blood lead and crime, but the reality is that we don’t know what blood lead levels were like in the 50’s. This clean fit is between real crime data and an estimate of lead levels. The estimate might be wrong, perhaps because it ignores other ways that children get exposed to lead.
In a paper questioning the lead-crime hypothesis, McCall and Land note that lead poisoning can come from other sources, besides gasoline:
There are a variety of sources of lead exposure. Among these are consumption of lead in paint, dietary lead intake from lead in soldered cans used for food preservation and cooking, and lead in gasoline (Bolger et al., 1991; Pirkle et al., 1994). Although many old buildings expose some individuals to leaded paint, exposure to lead used in food preservation and cooking have been greatly reduced in recent decades. During the 1970s, with the introduction of unleaded gasoline, lead exposure due to gasoline also declined dramatically.
McCall and Land try to reconstruct a curve of blood lead levels from some earlier measured data and compare that to violent crime. The relationship they come up with looks nothing like Nevin’s data:
From 1960, onwards, the two data sources match. But the earlier measured blood lead levels that McCall and Land found are much higher than Nevin’s estimates.
If their data is accurate, then there must have been some other large source of lead exposure prior to leaded gasoline. That is most likely exposure to lead paint dust, from kids living in houses that were previously painted with lead paint. But some might also be dietary lead from lead soldered cans or from lead pipes used for drinking water.
This could throw the entire lead-crime hypothesis into question, if blood lead was never as low as Nevin thinks, in the 1940’s and 1950’s. In that case, it could still be possible that the end of lead poisoning helped reduce crime in the 90’s, but that it did not cause the rise in the 60’s.
The same problem shows up in every one of Nevin’s international data sets. All the points in red here are measured, all the black lines are just estimates:
Conclusions:
The lead-crime hypothesis fails to be a grand unifying theory of crime.
Lead does not explain peak US crime in the late 80’s and early 90’s, crack cocaine does.
Lead poisoning does not explain the way that crime rises in the 60’s — crime rises among all age groups, not starting with the lead poisoned youth.
Crime has not fallen as much as you’d predict from the lead poisoning theory.
A number of countries in Asia and Latin America have murder rates that don’t match the lead exposure curves.
That said, it’s hard to entirely rule out lead as a contributor to crime. Take a weaker statement of the hypothesis, like: lead poisoning increased crime rates by 10% or 20%. That’s more difficult to disprove. It’s possible, based on the New Zealand cohort study or the US test score changes. It’s unclear if leaded gasoline was the only source of lead poisoning in the past — if it wasn’t, then lead would only explain some of the fall in crime since 1990, but it wouldn’t explain the rise in crime in the 1960's.
It’s not obvious what caused US crime to increase from 1964 to 1980. Some of it was just the demographics of the baby boom and a larger cohort of young people. If you want to credit a single thing for causing crime to peak in 1980, credit the legalization of birth control pills in 1960, which ended the baby boom.
Some other factor, besides demographics or lead, also increased the crime rate from 1965 to 1980.
Many western countries have crime curves that fit the lead/crime hypothesis, making it seem at a glance like a universal theory of crime.
This might just be because many had demographic waves at the same time — an increase in births after world war 2 and a decrease around 1960, with birth control:
Researchers should at least correct for these demographic changes before matching up crime and lead curves.
Nevin’s blog posts have some of the hallmarks of pseudoscience — they remind me a bit of ivermectin proponents, who would scour the world for one country where covid was in decline, and then attribute that decline to ivermectin, every single time:
Nevin’s desire to point to every cartel or gang warfare induced crime spike in Latin America and blame it on lead kind of reminds me of this.
He overuses the theory enough that it casts doubt on its own validity — maybe it’s a good strategy to drive blog traffic, but it makes me more skeptical, as a critical reader.
I think these kinds of theories fulfill a demand for answers about the world.
In the case of ivermectin, people wanted to know how to protect themselves from covid. They wanted an understanding of why covid came and went in waves. Many wanted a reason to not take the pandemic seriously, and a simple cure offered that. The ivermectin theory made you think that covid is treatable, that you could protect yourself, even that the world could eradicate the virus, if only the government would allow it.
In the case of the lead-crime hypothesis, many people notice that crime rates were worse in the 1980’s and 90’s. They want a simple answer for why that happened. People are complex and crime is driven by many factors. I’ve noticed a lot of fascinating things while writing this — like, I hadn’t realized how much spousal homicide has gone down since the 70's.
Some of the actual explanations for crime are kind of ugly to think about. Black people in the US live in more violent communities. Crack made its way to these communities in the 80’s and a lot of young black men died. Mass incarceration helped with the reduction in crime, but it’s also been destructive to those communities, it’s not a popular thing to support.
Having a simple theory for a complex phenomenon makes for good conversation. Learning about the lead-crime hypothesis gives some of the same kinds of emotions that learning a conspiracy theory does — it makes you feel smart because you know some hidden truth. I told many people about the lead-crime hypothesis after I first learned about it. Some were fascinated, others were incredulous. It was an interesting and polite conversation, either way. If I went to a dinner party and instead started talking about black kids smoking crack or the benefits of mass incarceration, few of my friends would be interested in that conversation.
In any case, no one is proposing we bring back leaded gasoline.
2020 saw a 30% increase in murder, in the US, and some big cities saw murder rates as high as they were at the 1990 peak.
If we dismiss the last crime wave as entirely lead induced, we risk missing out on other factors that mattered, like policing changes, or riots of the 60's, or the war on drugs. We risk letting crime rates grow back to 1990’s levels everywhere if we don’t look honestly at what drove them up last time and what’s driving them back up today.
