Has the lead-crime hypothesis been debunked?
jabberwocking.com
jabberwocking.com
> The lead–crime hypothesis is the association between elevated blood lead levels in children and increased rates of crime, delinquency, and recidivism later in life. (Wikipedia)
(I was interpreting "lead" as related to leadership or being (mis)lead, and was thoroughly confused for the first few paragraphs until I decided to just look it up myself.)
The Lead-Crime Hypothesis may also refer to the idea that lead is the main reason for the global crime wave that peaked in 1990. The evidence for this is less convincing.
How can you control everything? These studies do not control what no one thought about that it could have an effect. Its impossible to control everything.
You should read it to see if you think they didn't control for socioeconomic factors well enough. Section II and Table 1 in particular. https://pubs.aeaweb.org/doi/pdfplus/10.1257/app.20160056
It's really not a long paper. Go read it.
I know it's fun to posture about how backwards the US healthcare system is but we actually do literally have the system in place that you're virtue signaling about us not having, as it applies to children
People act like the US has no welfare at all and it just isn't true. Children whose parents can't afford treatment can still get treatment, and have been able to for more than half a century.
My own niece was born in a hospital because of this. My adult sister received treatment for her collapsed lung at the same time, paid for by the taxpayer, because she couldn't afford it.
Medicare and Medicaid might not be perfect but acting like the US has no safety net is a lie
It sounds like the parent comment was saying exactly this occurred for children exposed to lead, and those in the treatment group went on to have better outcomes.
EDIT: Looking through the abstract, it appears the threshold was on blood / lead concentrations. This means those who differed only a tiny amount in lead concentrations received different treatments (some vs none), and had different outcomes despite being similar in all other metrics. Figure 4 was pretty interesting.
I heart English Language!
An Updated Lead-Crime Roundup (2018) - https://news.ycombinator.com/item?id=20101446 - June 2019 (11 comments)
New Evidence That Lead Exposure Increases Crime (2017) - https://news.ycombinator.com/item?id=17888291 - Aug 2018 (147 comments)
Lead: America's Real Criminal Element (2013) - https://news.ycombinator.com/item?id=17350912 - June 2018 (141 comments)
A Basic Cohort Test of the Lead-Crime Hypothesis - https://news.ycombinator.com/item?id=16349365 - Feb 2018 (45 comments)
New Zealand Study Provides More Support for Lead-Crime Hypothesis - https://news.ycombinator.com/item?id=16023218 - Dec 2017 (36 comments)
A Study That Bolsters the Lead-Crime Hypothesis - https://news.ycombinator.com/item?id=14488520 - June 2017 (121 comments)
Lead Water Pipes in 1900 Caused Higher Crime Rates in 1920? - https://news.ycombinator.com/item?id=11625158 - May 2016 (67 comments)
Lead and Crime: Another Look - https://news.ycombinator.com/item?id=10868980 - Jan 2016 (11 comments)
Did removing lead from petrol spark a decline in crime? - https://news.ycombinator.com/item?id=7618871 - April 2014 (270 comments)
Leaded gasoline caused violent crime? Critiques wanted. - https://news.ycombinator.com/item?id=5017804 - Jan 2013 (1 comment)
America's Real Criminal Element: Lead - https://news.ycombinator.com/item?id=5002806 - Jan 2013 (4 comments)
The Crime of Lead Exposure - https://news.ycombinator.com/item?id=2609158 - June 2011 (3 comments)
Has use of lead-free gas decreased the crime rate? - https://news.ycombinator.com/item?id=71878 - Oct 2007 (4 comments)
First of all, the lead-crime hypothesis is good news. It is great to realize that small entirely achievable environmental changes can cause such marked improvement in wellbeing in people's lives.
But for some reason it has not been accepted as good news by the mass media. And then you have papers like the one discussed by the article that try to debunk the theory on absolutely ridiculous grounds. (You see there are no studies disproving this theory therefore there must be studies disproving this theory, therefore this theory is wrong. What a bunch of BS!)
So why is the lead crime hypothesis treated so negatively by our elites? Perhaps to avoid another round of massive litigation. Perhaps to prevent people from finding other ways other pollution can affect people's behavior and thus prevent the banning of other substances and other rounds of massive litigation. Or perhaps to preserve an image key in international culture -- the violent city youth.
https://www.medrxiv.org/content/10.1101/2021.06.11.21258690v...
As for whether this is a matter of degree rather than kind, there frankly have not been enough studies done of pre-post illness, and there should be. I know that depression is common in cases of severe pneumonia, and maybe there is a loss of gray matter in those cases too.
It's a bit difficult to commit violent crime at that age.
Yet it's easy to sit on the internet posting and believing in outright insanity and extremism.
It will take until 2040+ for many of these people to leave the political system.
If it turns out that lead exposure the culprit the throws a big monkey wrench in that.
Random thought as well, what if it turns out that the rise in obesity and type II diabetes is due to something similar?
I’m not suggesting pollution doesn’t have any correlation to crime (it probably does) but it doesn’t seem to be one of the more significant factors if you wanted to try to explain why some neighbours / populations commit more crime.
That is an oversimplification, and I think it's in the original article. Let me try to give an alternative explanation.
In an ideal world where all the studies have the same amount of subjects and all the countries have the same conditions, when you make an histogram of the result of the test you expect to see a Gaussian distribution. Some graphics in https://en.wikipedia.org/wiki/Normal_distribution
If half of the studies have 100 participants and the other half have 10000 participants, you expect to see the sum of two Gaussian distributions with the same center, it looks like a low wide mountain with a bump in the center. In a more realistic universe, each study has a different number of participants, and you get something in between. A rounded symmetrical bump.
Also, in a magical world where there are two type of countries, you will see the sum of two bumps with different centers. If they were far enough, you would see two humps. If they were too close, you will see only one hump but wider than the expected only form noise. And in between you can get weird shapes.
In a realistic word where each country is unique, and they are not so different, you get a bump. With enough variations, some luck, and crossing your fingers, you expect to see a bump that is similar to a Gaussian. It's not exactly a Gaussian, but somewhat close enough. More technical details in https://en.wikipedia.org/wiki/Central_limit_theorem
So if you have a very big number of studies, you expect to see in the histogram something like a Gaussian. In the first graph, instead of the histogram, the article uses another representation. In an ideal word, you expect to see an inverted ʃ. See again https://en.wikipedia.org/wiki/Normal_distribution
In a real word you expect to see something somewhat similar to an inverted ʃ. But the graphic shows only a L, Where is the top horizontal part of the inverted ʃ???
It's a very difficult question, and it's difficult to understand without a deep analysis. (That I can't do.) Perhaps there are good reasons, but it's strange.
In particular, the vertical part of the inverted ʃ in the graphic is close to 0, too close to 0. You can see that the lower tic of the inverted ʃ goes quite a bit to the right, so the "missing" upper tic should go approximately the same distance to the left. (Or there must be an explanation. Not all distributions are symmetrical, but it's strange that it's so asymmetrical.)
Now, the "missing" experiments in the "missing" upper tic of the inverted ʃ are the ones that say that the hypothesis is wrong (or even that lead is good for you). So it would be nice to have an explanation of thee weird distribution or the disappearance of these experiments.
Oversimplifying, there are no studies disproving this theory, but from the distribution there should be studies disproving this theory, therefore there is something weird happening here.
(Note: Sometimes the vertical part of the inverted ʃ is so far away from 0 that you don't expect any result in the negative part. Just the two tics at the top and at the bottom of the inverted ʃ, both in the positive part.)
OTOH, this meta-analysis was published, showing a negative result ...
Based on the blog post, I'm most concerned about the focus on murder rather than violence more generally.
# of violent incidents >>>> # of violent incidents that are also murder
So there really might not be enough murders to do the statistical inferences you want to do. Also, if there is a qualitative difference between murder and other violence, not merely a quantitative difference in degree, then it might get at the wrong things. Like, sometimes people go into a violent rage, and occasionally those end up killing some one. OTOH, sometimes people, say, rob a store, get scared, and shoot, and that's murder too. And then there are the people who carefully plan a murder to get something they want. Those all seem like they involve different kinds of brain states...
Maybe affect is so strong that to get opposite result you'd have to botch so many things that your chances of getting through peer review are so low that not a single wrongly successful research debunking this made it through peer review. Probably not many people are trying to debunk it and those that failed to prove it don't publish because people don't like publishing negative results.
If the effect is so strong that hardly ever happens, then it should also be so strong that there aren't many papers showing only a weak effect, but there's a huge amount of them. Maybe they're all botched but got published because their results agreed with the accepted idea whereas a contrarian botched paper would get more scrutiny and be rejected?
https://en.wikipedia.org/wiki/Publication_bias
In less interesting problems, the authors may decide to just not publish it.
In political areas, it's a risk that the far-right or far-left decides to support your preprint and later when you find the error claim that Big Something paid you to suppress the true. And the other side to claim that you are a far-other-side supporter.
Also, "my" model is an informal version of the model used in the research paper. You can read there an explanation with more technical details and more accurate.
The point they are making is that selection bias in which studies get published and which get suppressed can fool us (and may have done so) into believing an effect is real that isn't, or into thinking an effect is much larger than it is. Forget about which study this is and look at the point they are making on its own. Is there bad science going on here?
> but the mass media reaction to it just shows how much is wrong with our society
There is a whole other branch of mass media that desperately wants the lead-crime hypothesis to be not just true but the only reason for the crime rate falling in the 90s. Why? because it's much preferred over the alternative answer: Roe vs Wade. That side of the mass media is very aggressive in defending the lead-crime hypothesis.
Good discussion on the subject here: https://freakonomics.com/podcast/abortion/
And they suspiciously focused on murder only, they may have peeked at the data before deciding.
Theoretically the connection is sound (so I think it warrants a pretty solid prior probability), after all unwanted children lead to decrease in socioeconomic status, which is a very surefire way to have a lot of problems in life, and without a supportive family and state, the number of options quickly dwindles to illegal and not so great ones. (Basically the "not so great" option is that in the Soviet dictatorships around here - Eastern Europe - at least the full employment mandate meant that there was enough make believe work available to not starve. )
It's not just treated negatively by elites, but I think that's very straightforward. Because a biological or chemical explanation for behaviour, especially such a simple one, throws a pretty big wrench into the moralistic and individualistic worldview that underpins our Western societies, in particular in regards to crime or intellect.
If you accept the, from a scientific point, very obvious observation that a bunch of chemicals in your drinking water can turn you into an idiot or a criminal you just about undermine most stories of self-earned merit or responsibility we are being fed in our mother's milk and that justify bringing the hammer down on said inner city youth.
Lead is a lot harder to do it with because it was in water or paint.
The worse thing I think comes from some really old chemistry books from the early 1900's I inherited. The thing from those books is educated people back when they started putting lead in paint and gasoline absolutely knew lead caused developmental brain damage.
It's a good example actually. We've done incredible damage by moralizing drug use rather than treating it as medical issue and instead promote quackery like AA which has horrible outcomes.
Could you explain what is true? You are talking about a societal wide chemical change in behaviour in all aspects of life.
To start, what % or amount of the reduced crime is due to the lead-crime hypothesis?
Since the lead has brain damaged somewhat equally everyone's impulse control can we see this in TV viewing habits and other everyday behaviour? I'd expect if it can cause us to murder more, it'd also have many everyday effects.
Do you have any conversations around it's effects on other animals also in these environments. How would our pets changed or city animals like birds have changed their behaviour?
Societal wide chemical changes are pretty out there, but they happen, IQs rise as countries become richer through diverse food security but this a limit function, 90% of people get enough iodine so the IQ changes will on be the 10%. Lead-crime hypothesis is ~100% changed.
The Middle East was late to reducing lead. Do the think the we will see more stability there. Do you think lead-crime hypothesis is that important? What is true?
You'd expect them to be clustered around that value symmetrically, right? That's clearly not the case in the diagram that the blog post author quoted. There's a cluster around very weak positive effects, and then a long tail of strong positive effects but no matching long tail of weak negative ones. This suggests either that the negative results were truncated out, negative results have been hacked to positive ones, or there is some confounding factor at play. And I think the modeling in the paper is just an attempt at finding the confounding factor, and not finding one.
That said, the author's arguments around negative results being publishable in this field + studies not disappearing seem pretty strong.
(Or, at least this is my reading of the situation as a total amateur.)
Here's a potential confounding factor: maybe the cluster of very weak positive effects includes some studies that initially showed very weak negative effects but that had their methodology tweaked to avoid showing an "impossible" result. (I don't think anyone argues that atmospheric lead _decreases_ crime rates.)
If so, that would mean that there are some "missing" negative studies, but that there are also "too many" zero/very low result studies. If corrected for that, I suspect that the meta-analysis would show a positive effect size, though lower than 0.33.
Although in principle I don't see why it couldn't be the case, as leas poisoning also has symptoms such as depression and fatigue, which could marginally sway people towards inaction.
That could be counteracted by stimulant use, for example.
If it's true that lead exposure motivates crime, but at the same time causes people to feel too depressed and fatigued to act on their motivation, then my point is that if lead-exposed people then take stimulants (like meth, coke, or crack, for example) then those stimulants could counteract the depression and fatigue effects of the lead enough to allow people to act on their motivation to commit crime.
In case some of the downvoters are skeptical that stimulants could counteract depression or fatigue -- they certainly can. Not saying this is a smart thing to do, due to the addictive potential of these substances or their other deleterious effects. But they certainly can stimulate people in to action if they're otherwise too depressed or tired to do anything.
This is just a hypothesis. But it would be interesting to see if there's a correspondence between lead exposure, stimulant use, and crime.
The nexus of lead exposure, alcohol, and crime would also be interesting to investigate, as alcohol is already well known to be associated with violence, impulsive behavior, and poor decision making -- adding lead exposure in to the mix might potentiate this in to criminal activity.
I was responding to someone who said "lead poisoning also has symptoms such as depression and fatigue, which could marginally sway people towards inaction."
What I just quoted is an argument against lead being implicated in an increase in crime. I was just pointing out why lead might be implicated in crime after all, despite the above quoted argument against it.
So what exactly am I doing wrong here?
They already know it's likely wrong. You pointing out it could be wrong isn't needed. They already know there are lots of anti-lead arguments, so you making another one isn't needed.
I was just throwing an idea out there. I'm not convinced of anything.
But people seem more sedentary and depressed than ever before. Animal studies show the same - the controls are less active than the exposed group.
Maybe you only made people too weak to be capable of any violent crime.
Well, you need to post this reply to the person I was replying to instead of to me, as it was their contention that "lead poisoning also has symptoms such as depression and fatigue, which could marginally sway people towards inaction."
Scientists wouldn't have to argue this. They could just argue the null hypothesis as the results were really close to 0.
I'm not sure. This is a safe assumption if you don't know anything about anything, but I bet it's possible to construct or simulate scenarios where the hypothesis is "true" and where you also see the distribution of effect sizes shown here.
I haven’t read the linked paper yet, but its authors’ argument is that they have orthogonal evidence that p in this case should be considerably larger than 1/24, so given 24 total studies on the lead crime hypothesis, at least one should have returned a negative result purely by chance.
The validity of this claim depends entirely on whether their argument that p>>1/24 is sound.
They're saying you'd expect the results to follow a normal distribution centered at some value. Instead, we see a large number of studies clustered just above zero and a right tail, but no left tail. I think they're arguing that this is because the left side of the distribution is unpublished, and without knowing what it looks like, you can't come to a conclusion.
> The authors present a model that says there should be more papers showing negative effects just by chance.
To the layperson it sure looks like the lead-crime hypothesis is confirmed. Really curious to see a more in-depth follow-up on this meta-analysis.
Since this is a meta-study, drawing conclusions from the absence of null-results and the absence of clustering is prone to a lot of confounders. The article mentions publication-bias, where non-null-results are published with a higher probability. Other such confounders will exist, e.g. data-availability-bias in certain places and populations, location-bias because researchers study their usually first-world, urban, university-town surroundings more often. Community-bias, because researchers tend to cluster and get inspiration for new studies from their community and peer group on the conferences they visit. Data-similarity-bias, because they can only compare studies where data can be normalized to a common base. Quality-bias, because they necessarily include well-done and badly-done studies, in different amounts, usually without weighing them properly, because that is very hard to do.
Most of those confounders cannot be corrected for. So one could draw a conclusion like "we do see null-results in the expected amount and clustering in the expected amount, therefore the result of this meta-study is a confirmation of the lead-crime-hypothesis". But the only correct opposite conclusion is "we do not see null-results and clustering in the expected amount, so our result is 'we do not know'".
So the imho correct interpretation of the meta-study is "don't know, further research or better meta-analysis needed".
They're small, leach surprisingly little, and what they do leach is in a form that isn't readily bio-available.
Which isn't to say they're completely harmless, just not so bad as you might expect.
The caveat is that when it is removed, it may not be so bound.
Advice seems to be leave it alone unless it's flaking.
You may well be right.
> I’m amazed it has taken this long to get a single approved 100LL replacement, considering high octane unleaded fuels have existed for decades.
Outside of some specialized racing fuels, no. And nothing fulfilling the other requirements (distillation curve, vapour pressure, etc etc) of 100 octane aviation gasoline (this is measured with the MON procedure, as opposed to RON or AKI you'll find at your local gas station) at somewhat reasonable cost has previously been introduced.
I think the reasons why it has taken so long are 1) it's a genuinely hard problem 2) it's not a very large market.
I think the problem would have been solved a lot sooner if there was any urgency to switch, either by the people buying it, or by regulators.
The switch with automotive fuels was easy. Cars are disposable by comparison and regulators banned lead-burning ones out of existence.
Homicide rates are used for good reasons: they are least subject to reporting variations, they are hardest to fudge, and it’s the most serious crime.
Some evidence against the lead-crime hypothesis is the recent (2020-2021) jump in homicides in the U.S.
With a fairly major confounding factor.
The hypothesis was never "lead is the only reason there can ever be a homicide". The loss of 20 million jobs in a blink of an eye might well cause a spike. NYC, for example, is dropping back down. https://www.newsweek.com/shootings-down-june-new-york-police...
You know despite year over year shattering records of gun ownership that homicides with firearms are almost historic lows right?
https://ucr.fbi.gov/crime-in-the-u.s/2019/crime-in-the-u.s.-...
Edit: “Hmm, he sourced direct evidence… but I don’t like this truth”
To prove that gun culture and ownership doesn’t lead to crime, you’d need to show more than the data linked, such as a comparison between countries with strong vs. weak “gun culture” or high vs low gun ownership
So instead of the impossible, let’s look at the only the USA’s historic gun violence in the 90s. We add more guns than there has ever been, 40% of all the guns in the world in the USA, crime goes down - we can’t say adding guns was the reason for a crime drop - but we can say it appears adding more guns did not cause more crime.
Unless you have reason to believe crime would have dropped even more in an alternate universe where a magic wand was waved to remove the hundreds of millions of guns in the USA. But, a statement like that should come with evidence.
2. Despite this decline, the US has a higher rate of gun ownership and availability than other high-income countries, and this correlates with one of the highest rates of homicide amongst those countries. So my naive read of the data would be that an even lower availability of firearms would lead to an even lower homicide rate.
You just pointed to bullshit. Not your fault, someone wanted you to think that. But I can promise no one even remotely associated to guns believes this. No. The number of firearm owning households has not decreased. That’s propaganda and you just don’t know any better. I’ve seen it as a competitor and a trainer, new gun owners are making up these records - not as Michael Bloomberg (the single biggest source of gun control money in the world) would have you think. The number of concealed carry permits are at record numbers. The number of training outfits has never been higher. The numbers of protestors for 2A civil rights has never been higher. The only place you’ll see otherwise are antigun outfits with extremely questionable methodology.
The question you should be asking is why do you have your sources lying to you and what else have they lied about?
So right off, you use a lie that is sold to you as comfort food. “Dont worry, it’s just the same few horders trading guns to each other, they aren’t a majority, you are”. Then you use that to make a completely evidencless assumption that less guns equals less crime - when we only have correlation of the opposite.
So… while I really don’t like to apples and oranges… In 1997, Australia banned most all good defensive firearms. They also nearly doubled police per capita. Crime fell! But… in the USA crime fell by a greater amount over the same time period with no ban and only more guns added.
So again; there is absolutely zero proof that less guns equals less crime, especially when faced that more guns hasn’t made more crime.
> So my naive read
Indeed. But the worse part is you don’t want the truth. You are looking for the lie. And going far to find it, why is that?
Clearly something else than gun ownership / distribution has a massive influence.
I mention ownership / distribution separately above since several of these countries have had fully equipped actual assault rifles like G3 etc (not crippled AR-15s) and ammunition stored in private homes for decades kn addition to huge amounts of hunting rifles, pistols, shotguns and unlicensed guns.
A few suicides a year that's mostly it and removing the guns doesn't seem to work: I haven't heard about a massive drop on suicides since the Norwegians decided to store vital parts of the rifles in local depots.
It could be the case that below a certain threshold, further reducing firearm availability does little to affect crime, but the US is probably far from that level
So it is not about gun ownership but about gun culture abd culture generally.
Here is data showing that homicides did not go up during the pandemic in other countries: https://www.unodc.org/documents/data-and-analysis/covid/Prop...
Other countries handled pandemic aid entirely differently; I even linked to an article highlighting specific cases in the comment you replied to.
"It went up in the US, and not elsewhere" is a good sign the US's recent spikes had nothing to do with lead, in either direction.
Everyone is always looking for a political angle, but while everyone is arguing over the efficacy of their team’s favorite crime policies, the answer is probably: give every restless teenager an Xbox and some macaroni and cheese.
The lead hypothesis has not been debunked, and if you read the article in the OP you'll find a number of unanswered questions.
https://en.m.wikipedia.org/wiki/Lead%E2%80%93crime_hypothesi...
Edit: Ah, apparently it was mercury, not lead. Cute anecdote anyway.
Previously, urine was used.
She wasn't aggressive in any way, just not really "here".
Basically, during the 20th Century, lead were used in a lot of products, including gasoline and paint. To shocking surprise, it turned out that this lead slowly ended up in our environment, and built up in our bodies.
The hypothesis is that because virtually the entire population got lead poisoned, we became more affected by "learning disabilities, decreased I.Q., attention deficit hyperactivity disorder, and problems with impulse control", which again leads to more crime.
Lead was eventually phased out (no thanks to the companies behind the products), and this correlated with a drop in US crime-rates in the early 1990s.
Since then there's been lots of discussion on why the crime rate dropped. The removal of lead from the environment is one (an alternative, but not mutually exclusive, hypothesis is the legalized abortion and crime effect[2]).
[Personally, I've been wondering how this also affects other qualities of the people affected... Would people born before the 1970s have been more intelligent than they are currently? Could this have given us better political leadership, and better secondary and tertiary effects?]
1: https://en.wikipedia.org/wiki/Lead–crime_hypothesis
2: https://en.wikipedia.org/wiki/Legalized_abortion_and_crime_e...
https://www.motherjones.com/environment/2016/02/lead-exposur...
poor education & lower IQ = locked out of the labor force and more likely to make bad decisions in life
not rocket science.
> Nor are they the kind of study where you round up a hundred undergraduates and perform an experiment on them.
Do we have good clinical evidence showing: lead = lower IQ? (I'm guessing nobody volunteered for lead poisoning).
Thing is that crime reduction could be cultural, or simple widespread video games :)
(Maybe it's a bit everything, and lead probably isn't the only environmental improvement we've made)
Confirmed, sigh.
The abortion model is simple. Unwanted babies to mothers who don't want them will not grow up well.
We know lots of things target IQ, but the lead model is really complex, it hypothetically reduces impulse control across the entire population and with really tiny amounts in the environment.
When I see it proven in mice not humans I'll be more convinced. (And not large amount of lead in their water bowl)