Wednesday, January 16, 2013

I do not think that correlation suggests what you think it suggests...

TL;DR version:
Kevin Drum:  Wow!  When you add in a 23-year time lag, atmospheric lead levels correlate wicked close with violent crime rates - in every nation on the planet where we've measured these.
Jim Manzi:  Hang on a second!  One correlation is hardly conclusive, and there are all kinds of finicky fudge factors at work here; this is sure interesting, but there's almost certainly more going on here than what we're seeing.
Kevin Drum:  Dude, you just kinda made my point for me - I'm saying that atmospheric lead is one of those background factors that we've been missing.

Kevin Drum wrote an article in the latest issue of Mother Jones, opening with Giuliani's running for mayor in 1993 on the promise that he'd get tough on crime.  True to his word in a kinda-sorta way, crime dropped.  There were about three flies in this ointment, though:  first, crime actually had been dropping before Giuliani took office, peaking in 1990 and falling from there; second, a foretold rise in crime due to a demographic boom of prime candidates for thievin' and murderin' simply failed to materialize; third, crime wasn't only dropping dramatically in New York City - it was dropping everywhere, because the nation was one giant dubstep show and crime was its bass line.

On the one hand, I feel kinda iffy about that one; on the other hand, I use metaphors to keep my posters rolled up when I move.  I think I'll stick with it.

Hypotheses were developed:  economics, drugs, demographics, and others.  All failed - every proposed explanation had many and serious counterexamples.  What do you know?  This shit is complicated.  Enter Rick Nevin and Jessica Wolpaw Reyes, a consultant and Harvard grad student (respectively).  These two were both pursuing lead-related statistical research:  Nevin was working up a cost-benefit analysis on removing lead paint from homes, and Reyes was working on her dissertation.  In the course of their investigations, they found both international and state-by-state correlations which weren't exactly tight as a fish's asshole (which is to say, water-tight), but they were damned close.  And there was nowhere - not one place in the world - that didn't follow the same pattern.

Jim Manzi wrote a critique of Drum's article for the National Review, also looking into the paper that Reyes ended up publishing on the subject.  He puts forth sound counterarguments, but his ultimate conclusion boils down to:  more research is needed (which is always true).  Nevertheless, his skepticism isn't unwarranted - surprising and interesting as the correlation between lead and crime is, it's still highly unlikely that a single factor explains such a complex phenomenon as a worldwide rise and fall of violent crime.  Even the kinds of crime, when looked at closely, resist the explanation:  property crimes and murders don't track with environmental lead levels.  Even if Manzi is dead wrong, it's good food for thought and a healthy dose of critical thinking, especially since Drum's proposed solution involves spending hundreds of billions of dollars.

But Drum responded, escalating the situation to a full-blown conversation, and the exchange is fascinating.  Drum begins with a crash course in the foibles of statistical analysis - say it with me, "There are lies, there are damned lies, and then there are statistics."  There's also a fair bit of discussion on what's counterintuitive and what constitutes "common sense" (Einstein thought it was just a bunch of prejudices acquired by age eighteen).  The conclusion, though, is that Manzi actually plays right into Drum's hands with his skepticism:  yes, of course there are other factors at work; and with the research of Reyes and Nevin, it looks like we found one of them.  More research is always needed, and we mortals must always act with incomplete data.  Everything humans ever plan can be summed up as, "This seems like it could work, now let's see if it does."  It has to boil down to that, because we can never have all the facts, yet we must act anyway.

And in any case, it's not like we haven't been looking for other explanations - we've proposed a bazillion, and they've all failed to track as tightly as the rise and fall of leaded gasoline consumption.  The correlation between lead and crime is so much closer, so much bigger, so much more widespread, that it demands attention.  We've already got loads of research on the effects of lead on individuals, and it looks near as dammit that what we're seeing here is the effects on society.  The Hell of it is, even if Reyes and Nevin are dead wrong, their explanation is still the best we've got so far, and by a long shot.


Paul Sunstone said...

Fascinating post! While I agree there is a current need for further study of the relationship between crime and environmental lead, I also agree with you that we should not demand perfect knowledge of the relationship as a precondition of taking action. Thanks for a great read!

D said...

Thanks for your comment, Paul! :) I'm glad I was able to provide an entertaining read.