No. Sorry, I didn’t notice there was a link there.
Another serious confounder in that study is that the author is an economist.
Hah. This seems like a fully general counter-argument since I could replace economist with almost any other discipline and have similar issues. I don’t see anything actually wrong with their data or stats in this case.
I was under the impression that probability to murder has gone down among all age brackets.
Oh, come on. Everyone takes as their baseline the crack epidemic.
The graph you had shows the current dips to be below that of the start of the crack epidemic. (Treating 1980 here as start. This would be almost but not quite true if we started at 1984 which is sometimes used as the start point of the epidemic.)
Yes, young people are slightly less murderous than in 1980. I think that’s close enough for my claim of “as murderous as ever.” They aren’t driving the overall numbers, except by not existing. The point is that this is exactly the opposite of what is predicted by the first two theories. And not very compatible with the third.
One thing is for sure: though the national homicide rate is comparable to that of 1960, the situation is completely different. The ride up was different than the ride down. Here is a graph of homicide victimization rates 1968-2013, from CDC data. The previous graph shows how that is close to but not the same as offending rates. The very rank order of age buckets has changed.
When you replicate the paper, then you will not see anything wrong the data and stats. Or you will. At that point, I will be happy to look at your data and code.
Yes, young people are slightly less murderous than in 1980.
But much less murderous than 1988, and that’s about 20 years after the height of lead levels.
They aren’t driving the overall numbers, except by not existing. The point is that this is exactly the opposite of what is predicted by the first two theories.
How so? Abortion is the first theory, and making young people not exist seems to be one impact it has. Part of the standard version of the abortion argument focuses on their being fewer young people in bad environments, but general reduction in young people would have a similar impact. (Although this doesn’t explain the reduction in crime in other age brackets).
The very rank order of age buckets has changed.
Huh. That is a very interesting point. I don’t see how any of the explanations in question would make that likely. I recognize I am confused.
When you replicate the paper, then you will not see anything wrong the data and stats. Or you will. At that point, I will be happy to look at your data and code.
What I meant by that comment was that there were no obvious points where looking at it there was an obvious problem with the stats. This is depressingly common problem where one can look at a paper and just be like “yeah, no” to something in the way they handled the statistics.
I do think there is another possible issue that the paper doesn’t address: different states may have been measuring lead levels in different ways and I could see that as being a proxy for something else, like general attention to health of the population.
1988? The year before it skyrockets, somewhere in 87-89, looks to me the same as 1980. Are we talking about crack or not? That was way too abrupt to blame on lead, especially because it hit 14-17 and 18-24 simultaneously. I’d prefer to stay far away from crack, at 1980 or even the previous peak of 1975.
I suppose that you could have a much more complicated theory that lead makes people more susceptible to drug fads and gang wars, which is why the data is noisier than it used to be. But that’s a very different theory that cannot be tested by simple lagged regressions. And I think that the noise is only in America.
When people say “abortion reduced crime” they’re referring to Levitt’s theory that it produced individuals that are less criminal. Of course, that isn’t the theory that he tested, something that you can’t tell from his paper. The question of whether abortion reduced total fertility is controversial. I suppose that the claim that abortion lengthened generation time is not controversial. Anyhow, abortion is a drop in the bucket of demographic change.
No. Sorry, I didn’t notice there was a link there.
Hah. This seems like a fully general counter-argument since I could replace economist with almost any other discipline and have similar issues. I don’t see anything actually wrong with their data or stats in this case.
The graph you had shows the current dips to be below that of the start of the crack epidemic. (Treating 1980 here as start. This would be almost but not quite true if we started at 1984 which is sometimes used as the start point of the epidemic.)
Yes, young people are slightly less murderous than in 1980. I think that’s close enough for my claim of “as murderous as ever.” They aren’t driving the overall numbers, except by not existing. The point is that this is exactly the opposite of what is predicted by the first two theories. And not very compatible with the third.
One thing is for sure: though the national homicide rate is comparable to that of 1960, the situation is completely different. The ride up was different than the ride down. Here is a graph of homicide victimization rates 1968-2013, from CDC data. The previous graph shows how that is close to but not the same as offending rates. The very rank order of age buckets has changed.
When you replicate the paper, then you will not see anything wrong the data and stats. Or you will. At that point, I will be happy to look at your data and code.
But much less murderous than 1988, and that’s about 20 years after the height of lead levels.
How so? Abortion is the first theory, and making young people not exist seems to be one impact it has. Part of the standard version of the abortion argument focuses on their being fewer young people in bad environments, but general reduction in young people would have a similar impact. (Although this doesn’t explain the reduction in crime in other age brackets).
Huh. That is a very interesting point. I don’t see how any of the explanations in question would make that likely. I recognize I am confused.
What I meant by that comment was that there were no obvious points where looking at it there was an obvious problem with the stats. This is depressingly common problem where one can look at a paper and just be like “yeah, no” to something in the way they handled the statistics.
I do think there is another possible issue that the paper doesn’t address: different states may have been measuring lead levels in different ways and I could see that as being a proxy for something else, like general attention to health of the population.
1988? The year before it skyrockets, somewhere in 87-89, looks to me the same as 1980. Are we talking about crack or not? That was way too abrupt to blame on lead, especially because it hit 14-17 and 18-24 simultaneously. I’d prefer to stay far away from crack, at 1980 or even the previous peak of 1975.
I suppose that you could have a much more complicated theory that lead makes people more susceptible to drug fads and gang wars, which is why the data is noisier than it used to be. But that’s a very different theory that cannot be tested by simple lagged regressions. And I think that the noise is only in America.
When people say “abortion reduced crime” they’re referring to Levitt’s theory that it produced individuals that are less criminal. Of course, that isn’t the theory that he tested, something that you can’t tell from his paper. The question of whether abortion reduced total fertility is controversial. I suppose that the claim that abortion lengthened generation time is not controversial. Anyhow, abortion is a drop in the bucket of demographic change.
Hmm, that’s a good set of points. So it really does look very hard to explain what was going on and lead doesn’t look like it has enough data.
Having discussions with you has to be one of my favorite things on LW because whenever I do, I find out that I’m probably wrong about something.
I guess I should more often post evidence agreeing with people.