Crime and Punishment

Felix Salmon:

How Unleaded Gasoline Slashed the Violent Crime Rate - Finance Blog - Felix Salmon - Market Movers - Portfolio.com: The paper, from the NBER, is 70 pages long, but the conclusion, from Jessica Wolpaw Reyes, is simple, and stunning:

The main result of the paper is that changes in childhood lead exposure are responsible for a 56% drop in violent crime in the 1990s.

What are those "changes in childhood lead exposure"? Primarily the move to unleaded gasoline, which happened in the US between 1975 and 1985.

This result is not entirely surprising: I blogged a similar finding, by Rick Nevin, last summer. Nevin's paper is more international in scope: it covers the USA, Britain,Canada, France, Australia, Finland, Italy, West Germany, and New Zealand. But it also uses a less rich dataset: the new paper really nails this finding down.

What I learn from this paper is that sometimes the Law of Unintended Consequences can mean unintended positive consequences: the 1970 Clean Air Act had a much more beneficial effect on America than anybody guessed it would at the time. (Today, of course, we're living in a country where the federal government is suing California not to impose stricter emissions standards on automobiles, which is depressing.)

And as Shankar Vedantam of the Washington Post noted when writing about Nevin last year, these findings make politicians' claims to have reduced crime much less compelling, especially when you combine them with Steve Levitt's findings about the effect of abortion on crime. Here's Wolpaw Reyes:

The elasticity of violent crime with respect to childhood lead exposure is estimated to be approximately 0.8. This implies that, between 1992 and 2002, the phase-out of lead from gasoline was responsible for approximately a 56% decline in violent crime... The effect of legalized abortion reported by Donohue and Levitt [2001] is largely unaffected, so that abortion accounts for a 29% decline in violent crime (elasticity 0.23), and similar declines in murder and property crime. Overall, the phase-out of lead and the legalization of abortion appear to have been responsible for significant reductions in violent crime rates.

Significant? I'll say. 56% plus 29% is 85%, which means that the overwhelming majority of the reduction in crime can be attributed to exogenous factors for which local politicians can take no credit. Not unless they were involved in the Clean Air Act or Roe vs Wade, anyway.

Comments

[Bernard Yomtov]

Just to quibble - "56% plus 29% is 85%." The addition is correct, but adding isn't. It should be 69%, not 85%, as I figure.

1-(1-.56)*(1-.29)=.69.

[John]

Bernard Yomtov, darn you beat me to it. I was going to say something like: Sure, 56% + 29% is 85%, but 56% + 29% - 56%*29% is 69%.

Oh, well.

[Anderson]

And our world-champion percentage of citizens in prison figures where into this, exactly?

[CalDem]

its funny, the lead paper was in revision at QJE (in the top couple of economics journals) and then it dropped way down to a lower journal. I'd like to know the story.

[johne]

A Wikipedia article on the controversy provides a start on understanding it:
http://en.wikipedia.org/wiki/Legalized_abortion_and_crime_effect

[David Lloyd-Jones]

Ummm.

Nobody's saying what's in the paper. they're just quoting the exciting -- and extremely implausible -- finding.

Could somebody maybe post even a small hint on why this might be even remotely believable?

Like is all the rest of violent crime supposed to vanish once we get rid of chemicals in kids' candy? Gimme a break.

[John Emerson]

Abortion, non-leaded gas, some shit like that. This is science. It's in the numbers. Numbers don't lie.

Others say that it was the increase in cabbage consumption. But we know that there was cause -- probably one of these, or maybe something else, because we have used very sophisticated statistical tools.

You're not a creationist Luddite, are you? This is science. You're not against science, I hope.

NOTE: I am pro-choice. I am also pro-Clean Air Act. HOWEVER......

[Bloix]

John Emerson and David Lloyd Jones:
In both cases - abortion and lead - there are causal theories to explain the correlation.

In the case of abortion: each aborted fetus eliminates one unwanted child. Unwanted children are, more often than wanted children, unloved, uncared for, and abused. Abused children, more often than children raised in loving homes, become criminals and victims of crime.

In the case of lead exposure: Lead behaves chemically in a way somewhat similar to calcium. Calcium is an important component in nerve cells and plays a role in transmission of neurological signals between nerves. When an infant breathes or eats lead, the lead in the bloodstream is taken up into developing nerve cells in the places reserved for calcium, but the lead does not function precisely like calcium, so the nerve cells never develop properly. High levels of lead will kill a child for this reason. Moderate levels result in blindness, severe retardation, paralysis of the limbs. Low levels are linked to mild retardation, poor memory, poor impulse control, lack of coordination, late speech development. Kids who have been exposed to low levels of lead are more likely to be weird kids, stupid kids, problem kids. They are more likely to get in trouble at school, to flunk out, and eventually to wind up in trouble with the law.

[Jonathan King]

David Lloyd-Jones writes:
>
> Nobody's saying what's in the paper. they're just quoting the exciting -- and extremely
> implausible -- finding. Could somebody maybe post even a small hint on why this might be even remotely
> believable?

I am not in a position to validate the results of the current paper for several reasons, but I would strongly suggest you poke around in PubMed a bit. So maybe start with:

http://www.ncbi.nlm.nih.gov/sites/entrez?db=pubmed&cmd=search&term=childhood+lead+exposure

or slightly more precisely:

http://www.ncbi.nlm.nih.gov/sites/entrez?Db=pubmed&Cmd=DetailsSearch&Term=childhood%5BAll+Fields%5D+AND+%22lead%22%5BMeSH+Terms%5D+AND+exposure%5BAll+Fields%5D

Clearly, not all of this research is equally good or relevant, but in broad strokes, the current burning research question is whether lead exposure has a direct effect on problem behaviors in childhood and adolescence or whether most of the effect is caused by lead's deleterious effects on cognitive ability (usually measured via some IQ test). Now, epidemiology does not always hold up when you do actual experiments that use random assignment; for obvious reasons, we aren't ever going to do the critical experiments in people. But I think most of the latest papers are exploiting the fact that various places outlawed leaded gasoline (or paint, or whatever) at different times which gives you something you can work with at least similarly.

To make a long and very hairy story short, lead is a very, very, *very* bad pollutant, and early exposure is associated with dose-dependent bad outcomes of several different kinds. I am not sure how much work has been done on gene x environment interactions in this area, but it's a decent guess that there could also be groups that are differentially susceptible to lead's toxic effects.

In short, I'm not sure whether the current paper is solid, but lead would be one of the environmental exposures where your Bayesian prior probably should have non-negligible mass on spectacularly bad outcomes.

[rockyjoe]

Wouldn't it be additive? The postulated mechanisms are not independent. The lead only makes violent criminals out of those those who are born, not those who are aborted. But even if we assume that Donohue and Leavitt can be ignored we run into trouble. According to http://www.ojp.usdoj.gov/bjs/glance/viort.htm the violent crime rate went from about 47.9 per 1000 to 22.8 per 1000 during that period -- that would be about a 52% decrease. Other measures show decreases in the 50-60% range. The quote from the report says , "the phase-out of lead from gasoline was responsible for approximately a 56% decline in violent crime" NOT "56% of the decline". Are they purporting to explain (nearly) all of the decline as due to removing lead? This seems implausible, and almost certainly contradicts the assertion that there is an additional effect due to legalizing abortion. Perhaps the quotation was taken out of context?

[melissa]

Though we've known about the harmful effects for 40 years, lead remains in soil (which toddlers eat readily) in poor neighborhoods. A month ago, the AP reported efforts to clean it ups by adding "fertilizer made from human and industrial wastes on yards in poor, black neighborhoods."
http://www.boston.com/news/nation/articles/2008/04/14/sludge_tested_as_lead_protection_in_poor_areas/

I recall a Washington Post article from the late 60s about peeling lead paint and a defeated resolution to force landlords to repaint with unleaded substances. This was also in poor black neighborhoods.

[John Emerson]

Bloix:

I understand the causal theories. I am pro choice. I am anti-lead poisoning. But the results are reported too confidently:

"The main result of the paper is that changes in childhood lead exposure are responsible for a 56% drop in violent crime in the 1990s."

Especially because the same drop had already been confidently explained in terms of a different cause, abortion.

Furthermore, what I've read about lead poisoning in the past mostly implicated lead-based paint chips in decaying houses. Leaded gasoline was routinely used starting in the 20s, but during the period 1920-1980 the crime rate fluctuated. There was an increase in violent crime during the sixties and seventies which had nothing to do with lead, and then there was a decrease.

I think that one of the problems with extremely sophisticated statistical analyses is that they're only used when there are no blatantly obvious results discoverable by crude methods. From the virtuoso scientific point of view, the sophisticated results are infinitely more impressive, but in terms of reliable results, you really prefer blatantly obvious, unsophisticatef correlations. For example the one between tobacco and lung cancer or between the smallpox virus and smallpox. (If I'm, not mistaken, sophisticated statistical analysis of tobacco and lung cancer is mostly necessary because of tobacco industry quibbling. The crude initial results based on a simple analysis of a large number of death certificates, and a protocol asking doctors to record smoking habits, would be enough).

In the same way, sophisticated financial instruments seem pretty treacherous, as in LTCM, Enron, and the tranches in the most recent collapse.

Maybe I am a Luddite after all. I'm too old to care what people call me.

[Ken Muldrew]

John Emerson, Clauswitz enunciated the principle that you're looking for:
"Statistics is the continuation of experiment by other means." ;-)

Still, anyone who isn't skeptical of the unreasonably bold claim that was quoted needs to take some stop_being_so_trusting pills. Hardly a Luddite position.

[Valuethinker]

John Emmerson

But there were far more cars, and far more auto air pollution, in the 1950s and 1960s than there were (was?) before that-- also people began to drive much further. So exposure was anything but constant.

If you coincide the fall in leaded gasoline (1975-85) with the crime bust, and remember that violent crime is almost exclusively the province of men aged 15 to 30 then you have a good story.

The effects of leaded house paint were well documented in Australia in the 1940s, but not widely accepted in the US.

the toxic effects of lead on cognitive formation are well known.

[Valuethinker]

Does anyone else remember that the inventor of leaded gasoline AND of Freon (CFC for refrigerators) was one Thomas Midgeley, chief chemist at General Motors.

Midgely will go down in history as one of the most nefarious characters of the 20th century, in that he fiercely resisted any scientific evidence of the harm these products were doing. He thought he was doing social good.

[John Emerson]

Maybe it can be generalized across cultures. Colombians must put lead in children's breakfast cereal, judging by their murder rates. We should check lead levels in the Tutsis and Hutus, too, and most of the ex-USSR.

Sorry. There may be something there, but this kind of single-factor materialist explanation has been the bane of history for centuries. Lots of people are confident that the Mongol invasions were caused by changes in rainfall on the steppe. They don't know yet whether the rainfall increased before the invasions, or decreased, or first one and then the other, but it stands to reason, doesn't it? Because we're scientists, explaining things is what scientist do. And only anti-science Luddite obscurantist humanist enemies of reason and progress cavil about our great discoveries. Eventually quarks will have explained everything in one fell swoop.

[David Lloyd-Jones]

.

Thanks for all your good comments above, folks. Emerson, it's nice to have you around. Anderson's note is the one closest to my own heart: I'm sure that lead is your food or air is a Bad Thing, rots your brain, etc. etc. -- but criminality is socially defined.

It sees to me that the fact that huge numbers of Americans are criminals -- the typical teenager being thrown in jail for life the third time he snags some other kid's slice of pizza -- is a function of the fact that America has huge numbers of insane legislators, and has nothing whatsoever to do with the crud floating around in the air.

.

[c.l. ball]

The paper is available as a PDF here: http://www.amherst.edu/%7Ejwreyes/papers/LeadCrimeNBERWP13097.pdf

My stats are rusty so I'm unclear how she is deriving the percentage change in crime from the elasticity figures. What's the computation?

[anne]

This is pretend science, but what the heck anyone can play. Me, I think it is all cabbage consumption.

[smaug]

"Midgely will go down in history as one of the most nefarious characters of the 20th century, in that he fiercely resisted any scientific evidence of the harm these products were doing. He thought he was doing social good."

Midgley died in 1944. The danger of CFCs was not known until the early 1970s. Unlike TEL, CFCs are not toxic.

[kris]

Correlation does not imply causation, however sophisticated the correlation analysis may be. This is why I thought that abortion rates being linked to crime was a crock, and why I think this is a crock as well. There are (were) multiple factors different then and now and many of them are fairly important social, economic and legal factors. I find it surprising that one can make such a direct link such an improbable and *single* factor and something like crime which is dependent on so many economic and social variables. Somewhat similar is the research (much cited and admired) about the relationship between Hepatitis B and sex ratios in Asia (for which there is no medical evidence that hep b has an effect on sex of the baby as far as I can tell), which has apparently been acknowledged to be wrong.

This stuff is not science, it is statistics and I am sure people know the saying (there are lies, damn lies and statistics).