7 ms·

A Basic Cohort Test of the Lead-Crime Hypothesis

49 points by curtis 9y ago

InclinedPlane 9y ago
Undoubtedly lead exposure has had a profound effect on society, and on crime levels. Lead exposure leads to reduced IQ and increased aggression. But the crime landscape during the relevant period (the late 20th century) has too many confounding factors to extract data from it. For one, there's the "crack epidemic" crime wave of the late '80s and early '90s. During this period the rate of homicides committed by teens tripled and then went back to previous levels. Additionally, violent crime is correlated strongly with population density. And during the late 20th century in the US cities first began to empty out (due to "white flight") and then began to be repopulated as "yuppies" working office jobs and living in cities became more a more desirable "american dream" than the suburban family home and then later as urban centers became the core of the new tech economy. At the same time, police enforcement changed dramatically through the '80s, '90s, and later. The "war on (some) drugs (used by some people)", mass incarceration, etc. These are all tremendous confounding factors while the underlying "signal" from lead exposure is expected to be a gradual change over a fairly long period.
- vpribish 9y ago
  also the legalization of abortion, increase in food and medical subsidies for expectant mothers and young families, and the general rise in wealth.
- curtis 9y ago
  Separating out the confounding factors is definitely a big problem, but lots of effort has gone into just that. Drum has been following the Lead-Crime Hypothesis science since at least 2012. This blog post: https://www.motherjones.com/kevin-drum/2018/02/an-updated-lead-crime-roundup-for-2018/ https://www.motherjones.com/kevin-drum/2018/02/an-updated-le... provides an overview, but maybe not an ideal overview just because it's so long. Section 3 has links to some actual papers, and some of the other links may ultimately lead to papers, but there's so much of it that I didn't explore it deeply.
jupiter90000 9y ago
“Lead exposure in childhood may have played a small role in rising and falling crime rates in the USA but it is unlikely to account for the very high percentage of the decline suggested by the ecological studies” [1] [1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3829390/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3829390/
- alexandercrohde 9y ago
  I'm not sure why you're posting that link here. In academia, no single study is treated as conclusive. For example, when scientists observed a 50% decrease in men's sperm count in the last 30 years, they cited a meta-study which reviewed data from 185 other studies. [1] If you look on google scholar. It's trivial to find well over 100 lead-crime studies. The one you linked to seems to be only published-online, and not a meta-study, so it's not clear on what basis you (cherry?) picked it. https://doi.org/10.1093/humupd/dmx022 https://doi.org/10.1093/humupd/dmx022
  - jupiter90000 9y ago
    For me it's good to have a healthy dose of skepticism about these things, especially when it comes to social studies. I didn't like the conclusive tone of the mother jones article personally. The mother jones author does a fairly amateur analysis and concludes: "I’d say that basic cohort analysis supports the lead-crime hypothesis, while age analysis might not." The article I posted is a counterpoint to this in which an academic cohort study is reviewed finding lead explains less than 1% of the variation in crime in a cohort, and another cohort study there reviewed found no association between lead exposure and murder -- so a relevant article to post here, I'd say. Do you consider the results in the MJ article to be as robust as a meta-analysis and if not, do you have the same criticism of that article? (in which case, fair enough)
vpribish 9y ago
Do people generally find mother jones a quality information source? it has always seemed to me to be a over the line into left-wing-nut territory - willfully misleading to appeal to their market. I'd put it as basically a mirror of fox news on this chart : https://www.codlrc.org/content/news-sources-journalistic-quality-partisan-bias https://www.codlrc.org/content/news-sources-journalistic-qua...
- tyingq 9y ago
  Is there any news source truly in the middle, high quality and not low volume? I find myself having to read the left version and right version, then approximate the truth. Your referenced chart has NPR in the middle, for example. I do enjoy them, but the left bias seems obvious to me.
  - ikeboy 9y ago
    Reason.com
    - tyingq 9y ago
      Thanks...spent 5 minutes on a very brief flyover, but it looks promising. Lightweight site too, didn't notice excessive JavaScript or ads. I really am interested in better sources. Maybe I'm naive, but there seems like there's a reasonable, if not really lucrative, opportunity for someone to produce actual unbiased news in the US. It's worth a few dollars a month to me.
      - fleitz 9y ago
        It's hard to produce unbiased news, news is what's noteworthy, that implies at least curation if not explicit 'bias'. Hence whenever an article gets published, a group of people say 'they should have published something else'
      - dragonwriter 9y ago
        > Maybe I'm naive, but there seems like there's a reasonable, if not really lucrative, opportunity for someone to produce actual unbiased news in the US. Unbiased news is an impossibility. Centrist bias relative to the current position of the Overton Window is possible, but probably not economically optimal since political views aren't normally, or even unimodally, distributed.
curtis 9y ago
If you're looking for more reading on the Lead-Crime Hypothesis, Drum's "An Updated Lead-Crime Roundup for 2018" (https://www.motherjones.com/kevin-drum/2018/02/an-updated-lead-crime-roundup-for-2018/ https://www.motherjones.com/kevin-drum/2018/02/an-updated-le...) has a lot more information. Section 3 has some direct links to research papers, if that's your thing.
jcranmer 9y ago
This appears to be a case of "I have my preconceived interpretation of the data, and I'm not going to see how well other hypotheses match the data." The main thrust is that the cohort test supports the Lead-Crime Hypothesis because, on average, 70s kids were more violent than 80s kids were more violent than 90s. But later in the post, he mentions that there's an important confounding factor: the crack epidemic, which starts in 1987 and winds down in the mid-1990s. This would suggest that the increase in crimes over age for 1970s from 17 to 21 or so is fueled in large part (if we look at the later graph) by the crack epidemic. If you guesstimate that the trend should be flat-to-declining, then the trend line for the 1970s absent the crack epidemic is actually rather closer to the 1980s. Similarly the numbers for 1997 and 1998 (for 1980s) seem anomalously high, and should be lowered a bit. Put it all together, and you'd estimate that removing the suggested effect of the crack epidemic would be that there is no substantial difference between the criminality of people born in 1970, 1980, and 1990.
- cimmanom 9y ago
  This hypothesis is often put forth to in the US. The strongest studies of the lead hypothesis, however, compare data across multiple countries, which regulated lead-containing paint and gasoline at different times. These studies demonstrate that crime drops in each country at a similar period after the decrease in lead exposure, and at a rate that corresponds to the rate of decrease in lead exposure.
chris_va 9y ago
"Anyway, this is just a quick, amateur look at the data." ... Yes, yes it is. Statistics can tell you pretty much any answer you want to hear, if you are not careful.

49 points by curtis 9y ago

Undoubtedly lead exposure has had a profound effect on society, and on crime levels. Lead exposure leads to reduced IQ and increased aggression. But the crime landscape during the relevant period (the late 20th century) has too many confounding factors to extract data from it. For one, there's the "crack epidemic" crime wave of the late '80s and early '90s. During this period the rate of homicides committed by teens tripled and then went back to previous levels. Additionally, violent crime is correlated strongly with population density. And during the late 20th century in the US cities first began to empty out (due to "white flight") and then began to be repopulated as "yuppies" working office jobs and living in cities became more a more desirable "american dream" than the suburban family home and then later as urban centers became the core of the new tech economy. At the same time, police enforcement changed dramatically through the '80s, '90s, and later. The "war on (some) drugs (used by some people)", mass incarceration, etc. These are all tremendous confounding factors while the underlying "signal" from lead exposure is expected to be a gradual change over a fairly long period.

vpribish 9y ago
also the legalization of abortion, increase in food and medical subsidies for expectant mothers and young families, and the general rise in wealth.
curtis 9y ago
Separating out the confounding factors is definitely a big problem, but lots of effort has gone into just that. Drum has been following the Lead-Crime Hypothesis science since at least 2012. This blog post: https://www.motherjones.com/kevin-drum/2018/02/an-updated-lead-crime-roundup-for-2018/ https://www.motherjones.com/kevin-drum/2018/02/an-updated-le... provides an overview, but maybe not an ideal overview just because it's so long. Section 3 has links to some actual papers, and some of the other links may ultimately lead to papers, but there's so much of it that I didn't explore it deeply.

jupiter90000 9y ago

“Lead exposure in childhood may have played a small role in rising and falling crime rates in the USA but it is unlikely to account for the very high percentage of the decline suggested by the ecological studies” [1] [1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3829390/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3829390/

alexandercrohde 9y ago
I'm not sure why you're posting that link here. In academia, no single study is treated as conclusive. For example, when scientists observed a 50% decrease in men's sperm count in the last 30 years, they cited a meta-study which reviewed data from 185 other studies. [1] If you look on google scholar. It's trivial to find well over 100 lead-crime studies. The one you linked to seems to be only published-online, and not a meta-study, so it's not clear on what basis you (cherry?) picked it. https://doi.org/10.1093/humupd/dmx022 https://doi.org/10.1093/humupd/dmx022
- jupiter90000 9y ago
  For me it's good to have a healthy dose of skepticism about these things, especially when it comes to social studies. I didn't like the conclusive tone of the mother jones article personally. The mother jones author does a fairly amateur analysis and concludes: "I’d say that basic cohort analysis supports the lead-crime hypothesis, while age analysis might not." The article I posted is a counterpoint to this in which an academic cohort study is reviewed finding lead explains less than 1% of the variation in crime in a cohort, and another cohort study there reviewed found no association between lead exposure and murder -- so a relevant article to post here, I'd say. Do you consider the results in the MJ article to be as robust as a meta-analysis and if not, do you have the same criticism of that article? (in which case, fair enough)

vpribish 9y ago

Do people generally find mother jones a quality information source? it has always seemed to me to be a over the line into left-wing-nut territory - willfully misleading to appeal to their market. I'd put it as basically a mirror of fox news on this chart : https://www.codlrc.org/content/news-sources-journalistic-quality-partisan-bias https://www.codlrc.org/content/news-sources-journalistic-qua...

tyingq 9y ago
Is there any news source truly in the middle, high quality and not low volume? I find myself having to read the left version and right version, then approximate the truth. Your referenced chart has NPR in the middle, for example. I do enjoy them, but the left bias seems obvious to me.
- ikeboy 9y ago
  Reason.com
  - tyingq 9y ago
    Thanks...spent 5 minutes on a very brief flyover, but it looks promising. Lightweight site too, didn't notice excessive JavaScript or ads. I really am interested in better sources. Maybe I'm naive, but there seems like there's a reasonable, if not really lucrative, opportunity for someone to produce actual unbiased news in the US. It's worth a few dollars a month to me.
    - fleitz 9y ago
      It's hard to produce unbiased news, news is what's noteworthy, that implies at least curation if not explicit 'bias'. Hence whenever an article gets published, a group of people say 'they should have published something else'
    - dragonwriter 9y ago
      > Maybe I'm naive, but there seems like there's a reasonable, if not really lucrative, opportunity for someone to produce actual unbiased news in the US. Unbiased news is an impossibility. Centrist bias relative to the current position of the Overton Window is possible, but probably not economically optimal since political views aren't normally, or even unimodally, distributed.

curtis 9y ago

If you're looking for more reading on the Lead-Crime Hypothesis, Drum's "An Updated Lead-Crime Roundup for 2018" (https://www.motherjones.com/kevin-drum/2018/02/an-updated-lead-crime-roundup-for-2018/ https://www.motherjones.com/kevin-drum/2018/02/an-updated-le...) has a lot more information. Section 3 has some direct links to research papers, if that's your thing.

jcranmer 9y ago

This appears to be a case of "I have my preconceived interpretation of the data, and I'm not going to see how well other hypotheses match the data." The main thrust is that the cohort test supports the Lead-Crime Hypothesis because, on average, 70s kids were more violent than 80s kids were more violent than 90s. But later in the post, he mentions that there's an important confounding factor: the crack epidemic, which starts in 1987 and winds down in the mid-1990s. This would suggest that the increase in crimes over age for 1970s from 17 to 21 or so is fueled in large part (if we look at the later graph) by the crack epidemic. If you guesstimate that the trend should be flat-to-declining, then the trend line for the 1970s absent the crack epidemic is actually rather closer to the 1980s. Similarly the numbers for 1997 and 1998 (for 1980s) seem anomalously high, and should be lowered a bit. Put it all together, and you'd estimate that removing the suggested effect of the crack epidemic would be that there is no substantial difference between the criminality of people born in 1970, 1980, and 1990.

cimmanom 9y ago
This hypothesis is often put forth to in the US. The strongest studies of the lead hypothesis, however, compare data across multiple countries, which regulated lead-containing paint and gasoline at different times. These studies demonstrate that crime drops in each country at a similar period after the decrease in lead exposure, and at a rate that corresponds to the rate of decrease in lead exposure.

chris_va 9y ago

"Anyway, this is just a quick, amateur look at the data." ... Yes, yes it is. Statistics can tell you pretty much any answer you want to hear, if you are not careful.