>Exposure to lead in the environment might, perhaps, be the next best hypothesis. But as we showed in our 2020 paper, when one controls for both environmental lead and abortion, the coefficient on abortion remains large while the coefficient on environmental lead is greatly reduced and loses statistical significance.
Their analysis about lead relies on a single study by Reyes. Far from robust. Additionally, if lead results in more violent people, then it's no wonder that reducing the births of people in lead polluted environments would have an effect on crime. What they would need to do to tease our lead, would be to find places with no to little lead pollution and show that abortions resulted in lower crime, which they do not seem to do.
Studies by Mielke and Zahran (2012), and Feigenbaum and Muller (2016) provide further evidence that lead is the main culprit in crime reduction.
Lead seems to be the reason, not abortion rates. Levitt and Donohue don't engage seriously with the Lead hypothesis as there are many compelling studies supporting Lead. They only engage with the weakest research. I'm not impressed. I worry that Levitt and Donohue are just adding wood to racist fires and their work will be used by bigots, fascists, and racists around the world. It's irresponsible work.
Why do you think their work is irresponsible? Is it still irresponsible if they're right? Should people only engage in anti-racist inquiry, or do you think they should just self-censor if they get the 'wrong' result?
I don't find Levitt and Donohue very persuasive, but Levitt does seem very data-driven.
He answers your question in the same paragraph that he states they're irresponsible:
> Lead seems to be the reason, not abortion rates. Levitt and Donohue don't engage seriously with the Lead hypothesis as there are many compelling studies supporting Lead. They only engage with the weakest research.
The accusation is that they are ignoring research that might disprove their point, while at the same time are using weak research that they can easily refute to try to downplay the stronger arguments they refuse to engage with. Now I personally haven't read the paper, so I do not know if this is true, but it does answer the question you asked as to what OP thinks is irresponsible.
OP's accusation is that they Levitt and Donohue are irresponsible because the ideas they're investigating could be "used" by "fascists" and "racists".
Parent is questioning the idea that scientific inquiry should be restricted because it might reveal facts or open ideas that harm a preferred political program.
You are not answering that question. You're just talking about the quality of research, which isn't really the point here. It's whether the research should be done/allowed at all if it could harm a particular political ideology.
> Like the timid lady who, after a lecture on the Darwinian doctrine of man, said to her daughter, “Let us hope it is not true, my dear; or if it is, let us hush it up.”
> It's whether the research should be done/allowed at all if it could harm a particular political ideology.
I mean, I think that's kind of what OP was saying. Just go a little more extreme: Would it be responsible to publish science that you knew for certain would be used by others to justify genocide, even when there were other, better supported ideas to explore? I think we can agree that it would not be responsible. So where does the line get drawn? I don't know--I'm not going to draw it.
And let's not reduce big baddies like racism and fascism down to mere "political ideology". We're not talking about tax policy here.
Tools like statistical significance and correlation. These tools were developed by a known eugenicist R.A. Fisher to compare populations which is what Levitt and Donohue use here. Their table referencing Reyes work to tease out Lead and show abortion has more statistical significance uses the EXACT TOOLS that Fisher would use. These are tools used by charlatans who pushed IQ and other complete quack ideas on the general population.
All you have to do is study the mathematics to understand why. For example, why use correlation at all here? Correlation assumes a LINEAR relationship. No such justification is made in their paper of why a linear relationship should exist here.
Those pushing things like correlation to IQ have the same problem as IQ is a non-linear relationship. Are these ideas appropriate to other problems? Yes, correlation and statistical significance are useful tools when applied to the right problems. But they are not appropriate here.
Unfortunately most have a basic course in statistics and have no knowledge about the history of these tools. So they get convinced by charlatans who wow them with statistics.
Statistics are great tools to lie with because so few people understand what things like 'correlation' mean and things like statistical significance. So they see papers like this and get wowed.
In other words, justifying abortion as a way to control crime is disgusting and you can thread the needle from eugenics to Levitt and Donahue. Abortion should be justified on it's merits of reducing suffering of women, not some way to reduce crime.
Or the so-called crack wars[1], whose effect on criminal violence is so huge that whatever the effect size of lead or abortions was, it’s absolutely dwarfed to the point where it isn’t practically analyzable.
Between 1984 and 1989, the homicide rate for Black males aged 14 to 17 more than doubled, and the homicide rate for Black males aged 18 to 24 increased nearly as much. During this period, the Black community also experienced a 20–100% increase in fetal death rates, low birth-weight babies, weapons arrests, and the number of children in foster care.
The problem is the claim that the crack wars would have been even worse without aborting future crack warriors is impossible to size. It ruins any baseline.
Ideally, we would be able to tease out the effect of abortion, lead-paint, AND the interaction (lead-paint x abortion). If an interaction model better explains the data, then we should be open to not rejecting it.
They didn't try that. They also didn't engage with other literature showing link with lead and crime at a micro level. In other words, they didn't seriously counter the lead hypothesis, which is stronger than their work.
From the article.