I haven't written much about the lead-crime link lately because there hasn't been much to write about. Today there is. Beware: a righteous rant is coming.
Many years ago Steven Levitt and John Donohue—Levitt is one half of the Freakonomics duo—postulated that the crime decline of the '90s was due to Roe v. Wade. "Children who are unwanted at birth are at risk of a range of adverse life outcomes and commit much more crime later in life," they explain, so if there were fewer unwanted babies after 1973 they posit that there would be less crime 20 years later.
Today, Andrew Gelman sent me off the deep end by quoting a recent update from Levitt and Donohue:
Though there is not complete acceptance of our hypothesis among academics, all agree that if our paper is not correct, then there is no viable explanation for the enormous drop in crime in America that started in the early 1990s. Indeed, there is not even an arguable theory to supplant the abortion-crime link. . .
Jesus Christ. Levitt and Donohue are well aware that violent crime dropped all over the world in the '90s but abortion was legalized only in the US. Their theory plainly explains nothing. What does, despite their cavalier dismissal, is lead.
Levitt and Donohue point to a simple correlation: abortion was legalized in the US in 1973 and crime started dropping about 18 years later. That's not much, but even at that their correlation doesn't hold up:
The abortion rate went up very suddenly over the space of five years, but crime didn't drop suddenly from 1990-95. It took 20 years. Furthermore, the abortion rate started decreasing sharply in 1990,¹ but crime didn't go back up starting in 2010.
This whole thing is just a coincidence that Levitt and Donohue stubbornly refuse to admit. 1973 is also right when lead use in gasoline started to drop. This is why their theory makes predictions that are mostly correct: because abortion increases happen to line up with lead reductions.
And there's more. Levitt and Donohue have a single nationwide correlation and a few differential state correlations. That's it. But so does lead, and as I've said many times before, if that's all there were I never would have written about it in the first place. But there's far more. There are global studies. There are neighborhood studies. There are studies from the early 20th century. There are prospective studies that track individuals from birth to adulthood. There are brain imaging studies that explain precisely what lead exposure does to developing brains.
In other words, not only is there an "arguable theory" for the crime drop of the '90s, there's a theory with absolutely mountains of evidence behind it. It was gasoline lead.
But wait. Are there any studies that suggest the lead hypothesis is wrong? Well, there are some dumb ones. For example, a study that looked only at homicide and only during the '80s. That's so stupid it beggars the imagination. Or there's one that uses a different measure of crime, but also shows a 96% correlation (!) between, for example, carjacking and assault. This also beggars the imagination.
To my knowledge, there's only one serious data point that pushes against the lead hypothesis: a meta-analysis that concluded (a) there was strong evidence for lead causing crime, but (b) there was also evidence of publication bias. That is, studies that found lead-crime links got published but studies that didn't were tossed out.
This has to be taken seriously, but I'm skeptical of it. Measures of publication bias are necessarily fairly crude, and there are lots of reasons to think that lead studies are unlikely to be rejected just because they showed small or no effects. That would be very publication worthy!
I'm usually a little more relaxed about this, but I'm pretty fed up with criminologists who steadfastly refuse to admit the obvious. No one thinks that lead is exclusively responsible for crime, but if you're looking for an explanation of the crime drop of the '90s and aughts, gasoline lead is it. The evidence by now has piled up so high it's all but irrefutable. What's more, aside from lead Levitt and Donohue are correct that there really aren't any other plausible theories.
The dead enders need to pull their heads out of the sand. Gasoline lead caused crime to go up in the '60s and '70s, and the end of gasoline lead allowed it go back down in the '90s and aughts. There's not much else to it.
¹Ironically, this was probably due to lead. The same mechanism that explains crime also explains less teen pregnancy and therefore fewer abortions.
Remember when sociobiology was all the rage in the 1980s? Where just about every human activity* was some sort of triumph of reproductive success? All it took was a narrative that sounded plausible.
Oxford shirts, with their vertical row of buttons, is subconsciously seen by women as the spine of a man in good health and therefor a preferred mate. That's why a dude wearing Van Heusen is successful at a single's bar.
Levitt and Donohue with their abortion - fewer unwanted - stable upbringing - less crime, seem to be using that kind of logic.
* including stamp collecting (really)
Demonstrating that cryptozoologists refer to themselves as scientists is not evidence that science is bogus. Likewise, demonstrating that idiots refer to themselves as sociobiologists is not evidence that sociobiology is bogus. On the other hand, the idea that human nature, as distinct from human nurture, has little influence on human behavior is as bogus as a three dollar bill.
i still think it has more to do with young men spending more time playing video games and using porn
previous generations were bored shitless so they would drink and drive around getting into trouble
(and also get their girlfriends into trouble)
hours playing video games is also highest on the weekend, which is when pre-90's kids were bored and getting into trouble
https://us-static.z-dn.net/files/ddf/dbbd83b54559cf914e1c102ec5c15231.gif
video game revenue since 1995
rise in the number of aggressive video games
There are a number of problems with the video games hypothesis, but the most glaring is that, unlike leaded gasoline, video games don’t provide a reason for an *increase* in crime as well as a decrease.
Kevin is right. It’s a story of lead in gasoline, on a huge scale.
homicide rates were much higher in the past, and they didn't use leaded gasoline
Fourteenth-century Florence experienced the highest known annual homicide rate of 150 per 100,000, and estimated homicide rates in 16th-century Rome ranged from 30 to 80 per 100,000.
https://humanprogress.org/trends/global-murder-rate-is-falling/
violent crime in the United States has been in decline since colonial times. The homicide rate has been estimated to be over 30 per 100,000 people in 1700, dropping to under 20 by 1800, and to under 10 by 1900.
https://en.wikipedia.org/wiki/Crime_in_the_United_States
In general we can see a marked decrease in violence in many human societies over the long term. In very broad terms, humans have indeed been growing less murderous over the millennia, thank heavens.
One notable, short term exception to this was the four odd decades years after World War II in the United States* of America—a trend that mysteriously seemed to fade as children born in the post unleaded gasoline era began to hit their prime crime years. Gee, I wonder what might have caused this?
*Observable in numerous other countries, as well, with adjustments on timeline consistent with the local situation pertaining to leaded gasoline usage.
I guess all human beings can't find the same things intuitive. But to me the LCH is overwhelmingly so.
and you have data that the lead-poisoned cohort is identifiable over time?
Uhm, what? Do you seriously think that tens of million of efficient, lead-pollution machines were sent onto our streets and it wouldn't have created an increase in humans exposure to environmental lead? How could this have been avoided? The millions of gas masks issued by the government to families with young children? LOL.
Also, you still haven't responded to the failure of the video/porn hypothesis to explain the sharp increase in crime in the postwar period. Again, the LCH does this perfectly—in myriad countries.
violent lead-poisoned 20 year olds should eventually be violent lead-poisoned 30 year olds, then violent lead-poisoned 40 year olds, 50, 60, etc
so over time there should be a bump in the charts associated with the lead-poisoned cohort showing that they were more violent at each age than the non-poisoned generations at the same age
Levitt seems on the spectrum. (Not a judgment, just an observation.)
Boy do people get wedded to their positions.
Kevin's claim that there is no evidence against the lead-crime hypothesis is nonsense. There is above all the known ages of the criminals. Kevin thinks the crime wave is explained because lead in pre-schoolers peaked in 1970 while crime peaked in 1992-3. If the pre-schoolers were born in the late 60's, then they would be in their late 20's in 1993, which is normally a prime age for some types of crime. But the FBI data show that the violent criminals during the crime wave were consistently 18-19 years old. They were born in the middle 70's, well past the time of peak lead content. Furthermore the modal age of criminals did not increase during the crime years - it consistently stayed at 18-19, which is not the behavior of a consort. There is no evidence whatsoever of any age group, whether born in the late 60's or middle 70's, committing crime at an elevated rate in later years, contrary to what would be expected from a criminal consort. Kevin admitted once that the age data is a fatal flaw, and another time he presented some of the age distributions which illustrate it, but apparently he wants people to forget that.
There are many other inconsistencies, such as the fact that the crime wave did not occur in other countries which had leaded gas:
https://kieranhealy.org/blog/archives/2015/10/01/assault-death-rates-1960-2013/
and the fact that the NAEP and other school testing shows no effect of the supposed lead impairment. Those who are really interested in this need to read from other sources written by scientists who have really examined all or most of the evidence. What Kevin presents is not science, it is one-sided argument - he collects things which he thinks support his hypothesis, rather than subjecting the hypothesis to meaningful tests. Of course others have done the same thing, such as abortion-hypothesis supporters.
Kevin is good at spotting inconsistencies in media claims when he can look up the relevant data in a day or two, but he got in over his head with the crime-wave problem and apparently missed all the inconsistencies. Now he refuses to admit he was wrong.
Perhaps, but there is international evidence about the crime drop! https://youtu.be/8c3PoD_qW1Y?si=AhbVqrCzkJnGSsl9
The lead-crime hypothesis, the idea that leaded gas explains changes in the violent crime rate over time, conflicts with the null hypothesis, the idea that the lead-crime hypothesis cannot be proven beyond a reasonable doubt. The latter is confirmed by default, and the former is confirmed by exception.
Has relevant evidence been excluded? If yes, then the lead-crime hypothesis is subject to confirmation bias and definitively refuted. Otherwise, there are identified and unaddressed logic flaws in the assumptions used to connect the evidence to the lead-crime hypothesis, and then it is definitively refuted, or it is verified. And if it happens to be the only verified explanatory hypothesis, then it is conclusively confirmed.
That is science.
Deep Thoughts, by Jack Handey
That is not science. There is a very long range between "well confirmed" and "decisively refuted" and "beyond a reasonable doubt" is a legal standard, not a scientific one. If the data appears to confirm the hypothesis but there are issues with the data, that does not mean the hypothesis is decisively refuted, it simply means that the probability of it being true is less. In all cases where there is data that even weakly supports a hypothesis, the hypothesis has a probability of being true.
stuff and nonsense.
Lead didn't get off windowsills just because we took lead paint off the market.
Here in Maine, where the housing stock is older, nearly every home from before that time is sold with a lead-paint disclaimer.
Your numbers don't reflect the reality of how homes are lived in, and how slowly the lead has been taken out of them. It is still there, still in the dirt around the houses; and the decline in lead poisoning (and the violence it causes) reflects that reality.
>>Now he refuses to admit he was wrong.<<
Kevin has nothing to “admit”…..he is correct.
Here are some other things that correlate with crime.
https://www.tylervigen.com/spurious/variable?id=20220
This is a favorite of mine
https://www.tylervigen.com/spurious/correlation/11869_us-milk-fat-used-to-produce-fluid-beverage-milk_correlates-with_violent-crime-rates
This is just argument by ridicule and as such it is nothing to take seriously.
Everyone knows correlation is not causation, But if you have correlation and a plausible mechanism, which lead poisoning provides, you have a fine hypothesis.
Proceed from there. Laughter is not a rebuttal.
Earlier this week, Kevin asked about site performance. I, for one, have had a persistent issue. When a post is fairly new, and so there are relatively few comments, clicking on the line for them doesn't show them. Rather, it says I must be logged in to comment. If I wait an hour or two, clicking shows the thread as (surely larger by then) it then stands.
Next up, fluoride: https://www.theguardian.com/society/article/2024/aug/23/fluoride-water-study ...
But since lots of folks are not drinking tap water anymore but buying bottled water, doesn't that negate the intended benefits of fluoridating municipal water supplies?
Wonder if you can find a correlation between consumption of bottled water and reduction in whatever the effects of fluoride are supposed to be.
I've been following your analysis on lead crime for years and much appreciate it. I this the other day on Twitter and curious what your thoughts are: https://x.com/Noahpinion/status/1825584397363589312
Their first "Freakonomics" book was pretty good.
The second--they went off the the deep end a bit....
https://time.com/archive/6933942/are-the-freakonomics-folks-off-base-on-global-warming/
for those interested in Global Warming, see The Climate Wars done by the BBC--the first one (1/3) is found here:
https://topdocumentaryfilms.com/earth-the-climate-wars/
What’s the theory again? Lead exposure / poisoning makes people marginally more terrible? Potentially criminally insane? Violent? Lacking impulse control? Drugs made them do it? There’s a lot of crime and violence among humans. Republicans are just straight up assholes by definition. Then you have war and religious fanaticism.
So many problems. And we want to pin a small period of social dysfunction on lead exposure? What would you call the US civil war? The abortion theory sounds dumb too.
Sadly, damage caused by lead exposure is irreversible. So we'll simply have to wait for a new generation of pop economists who aren't as brain damaged as Levitt.
For a bit of historical perspective, working as a lead pipe artisan in Rome was a strongly mitigating factor in a Praetor's (criminal) court, a venue not noted for excessive leniency. The stuff makes you nuts.
Beats me why Kevin is so emotional about this. I read Levitt/Donohue's paper when it came out and Kevin/lead were nowhere to be seen.
It's because (a) this has become part of his personal brand, and (b) just within the last year the quality of his blogging has dropped off the roof.
"all agree that if our paper is not correct"
Is this the same "all" that wanted to get rid of Roe v Wade?
When trying to correlate social phenomena, such as crime, with environmental factors like lead, one should always look at more than one country. Leaded gasoline was used in other countries, and was prohibited at different times in different countries. Does the banning of leaded gas correlate with a decrease in crime 18 years later in other countries? If not, then look for explanatory factors unique to the US.
Does the banning of leaded gas correlate with a decrease in crime 18 years later in other countries?
Yes, it does. This is one of the findings that Kevin's work has highlighted. Indeed, this effect—as well as the increase in crime in myriad countries trailing the introduction of leaded fuels by 15-20 years—is one of the most powerful pieces of evidence for the lead crime hypothesis. It appears MoJo provides Kevin's original, ground-breaking work on this topic free of charge. Take a read if you've got twenty minutes:
https://www.motherjones.com/environment/2016/02/lead-exposure-gasoline-crime-increase-children-health/
Here's a paper by researcher Rick Nevin looking at multiple countries. The abstract reads:
https://pic.plover.com/Nevin/Nevin2007.pdf