Jan 14 2013

Lead and Crime

A recent article in Mother Jones discusses the potential role of lead in the increase in crime from the 1960s to the early 1990s, and the subsequent steady decline in crime rates since then. I have received numerous questions about this article and this possible connection between lead and crime. It is a well-written article, and an interesting question. Could one toxin really be responsible, among all the other possible causes, of the rise and fall in crime rates in the US?

Before we get to that, it’s interesting that this is not the first time in history this question has come up. There is a theory that the fall of Rome was due, in part, to chronic lead toxicity. The Romans used lead to make their water pipes (the origin of the word “plumber” as the root “plumb” refers to lead). But this was probably not the most significant source of lead for the Roman aristocracy, who also sweetened their wine and some of their food deliberately with lead. Analysis of lead levels in bones of burials from the time do show variable but often elevated lead levels, but the data is too scant to draw any firm conclusions.

They apparently knew of the toxic effects of lead, but thought that this was limited to acute lead toxicity – something that slave lead miners had to worry about, but not citizens. They did not think that low level chronic exposure was a risk, and apparently they were wrong.

However – there are problems with this nice story. Some scholars doubt that lead poisoning was as endemic in Rome as others claim. The evidence is complex. Lead was probably not added deliberately to wine, but wine was occasionally heated in lead containers. Terra cotta pipes were often used instead of lead for carrying water specifically because the risk of too many lead pipes was understood. And contemporary references in medical and veterinary writing make scant mention of lead poisoning, even though the syndrome was well recognized.

Beyond disagreements over the extent of lead poisoning in the Roman Empire, there is the far thornier problem of assigning cause to the fall of Rome. Most scholars criticize any attempt to find “the” cause of such a complex historical event, whether lead or anything else.

We find ourselves in a similar situation today. The Mother Jones articles centers around the rise and fall of crime from the 1960s to early 1990s in the US, and discusses research that argues that this rise and fall in the crime rate closely mirrors (although shifted by 23 years) the rise and fall of the use of leaded gasoline.

The article correctly points out that such ecological data is very difficulty to interpret. There are many potential factors that could correlate with the rise and fall in the crime rate. Confidence increases, however, when the correlation holds up in many possible ways. The article cites research by Rick Nevin and others showing that the correlation with environmental lead levels, mostly from leaded gasoline usage, holds up when compared state by state, city by city, and even by neighborhood. and also in countries other than the US. This makes the correlation much more compelling.

The literature seems to support a real connection between lead and crime. A recent review and study concluded that there is a correlation, and that lead levels in the air are predictive of crime rates, but especially when linked with resource availability. In other words, the effect of lead exposure on crime rate is greater when those individuals live in a neighborhood where they are less likely to be diagnosed and treated for lead poisoning, which makes sense.

There remains the possibility that lead exposure is just a marker for other variables that are truly the cause of crime. In other words – it’s possible that crime really has sociological causes, factors that also predict lead exposure. While this type of connection is probably true to some extent, it does appear that lead is an independent variable correlating with crime. Also a connection has been demonstrated with a prospective study, which partly controls for such confounding factors.

The lead connection is also biologically plausible, as it is well established that lead causes neurological changes including and increase in violence and decreased executive function.

While it is difficult to account for all possible confounding factors, the story that environmental lead exposure increases crime by causing chronic symptoms of violence and attention deficit is plausible and reasonably supported by existing evidence. The real question, it seems to me, is the magnitude of this effect, especially compared to other effects on crime.

The Mother Jones article, by Kevin Drum, cited a figure that 90% of the increase in crime since WWII might be due to lead. He was called out on this figure by blogger Deborah Blum, and Drum later printed a correction. He said the 90% figure is at the upper limit of the range of estimates, and that 50% is likely closer to the truth.

In the review I cited above, reference is made to research showing that “as much as 20%” of crime is “lead related.” One small point – Drum’s now 50% figure, as he points out, is the rise in crime, not the cause of all crime. The 20% figure cited in research is all crime – so these numbers may be compatible. Either way, the 90% figure likely overstates the connection.

Therefore, even accepting the 20% figure, that means 80% of crime has nothing to do (at least directly) with lead, and the sociologists are free to continue to speculate and study about the myriad of social causes of crime.


The connection between chronic lead exposure and neurological effects, including those that plausibly contribute to crime, is both plausible and reasonably supported by existing evidence. The magnitude of this effect is difficult to tease apart from the many variables that can potentially affect the crime rate. If we accept the 20% figure (crime that is lead related), which seems plausible, then this indicates a significant role for lead, but lead is certainly not the only important factor.

Also, because of the nature of this research there remains reasonable doubt about lead’s true role in the crime rate. This doubt, however, is not sufficient to argue that we should not pay attention to lead exposure or even take specific measures to limit it. The research is remarkably consistent in pointing to a real role for lead exposure. Multiple studies have also looked at the potential benefit of further reducing lead exposure (from soil and remaining lead paint, especially in window frames of old buildings). The research I can find all concludes that the benefit of lead reduction measures would be cost effective because of the potential benefits that would result.


15 responses so far

15 Responses to “Lead and Crime”

  1. locutusbrgon 14 Jan 2013 at 1:25 pm

    20% still seems overly high to me there are so many compounding individual factors related to lead exposure itself.

    The effect of lead on an individual has a great deal of variability; age/size of the individual concentration of the exposure, duration, variation of serum lead levels within affected dwellings.

    When it comes to Rome I always get the sense that once we have a good grasp of a modern problem, we start post-hoc fitting it to an old mystery. I see this a lot in paleontology and astrological theory. Too often I see scientific theory including greenhouse effect into absolutely every disaster scenario. Once we have a good scientific grasp of an issue I think theorists start seeing it everywhere. That is just my opinion. I am not advocating ignoring the lead problem. I am concerned that this type of hypothesis distracts people from the major factors affecting crime ridden area’s. Lead abatement is a relatively easy problem to deal with compared to poverty and community violence.

    Interesting, but this type of behavioral link always makes my skeptical spidey sense twitch. Sometimes inappropriately.

  2. Steven Novellaon 14 Jan 2013 at 1:41 pm

    locutus – I agree, but this is the information I found. I don’t know how reliable the 20% figure is, but I do note it dates from the peak of crime in the early 1990s, which means it is probably a smaller contribution to today’ crime. (taking all this at face value).

    In any case, as I said, that still leaves 80% or so due to other things.

    If we take the narrow question – should we invest money in reducing lead in the environment, I think the answer (even giving the uncertainties in current data) is yes. But as always let’s also track the results.

  3. Mark Ericksonon 14 Jan 2013 at 1:45 pm

    Good post. I’d like to see a comparison with other variables over the time period (60′s to today). Most likely they are sociological, such as household income, single-parent families, social cohesion, drug trade trends, legal regimes, etc. However, there must be other chemicals that are neurologically harmful that rose and fell over that time period as well. If such a source was found, and it very closely correlated with leaded gasoline use, would that mean the two together would account for about 20% of the variation or could the two together account for more using the same framework?

  4. SARAon 14 Jan 2013 at 2:30 pm

    I expected you to debunk it as so much bad science. That was my initial reaction to the story before reading your thoughts. It just seems wrong.

    Seems is such a bad way to base conclusions, isn’t it?

    Anyway, surely they should do more studies on individuals who have higher exposure, but non-toxic exposure and see if their behaviors are more violent / criminal.

  5. Woodyon 14 Jan 2013 at 2:54 pm

    You will have to pry my vintage lead D&D miniatures from my cold dead hands!

  6. Heptronon 14 Jan 2013 at 6:29 pm

    Does the lead stay in someone’s system, or does its exposure leave behind any signs in the body? That seems to be like it might be worth looking at.

    Also, whether or not there is a link to crime, isn’t it worth removing lead from the environment because there’s lead in the environment?

  7. Davdoodleson 14 Jan 2013 at 7:37 pm

    Chelation therapy for everybody!

  8. tmac57on 14 Jan 2013 at 8:49 pm

    Davdoodles- Too late ,I’m afraid the damage is already done :(
    The PreCrime Unit will be there soon…run!!!

  9. SimonWon 18 Jan 2013 at 5:21 am

    Oh a chance to remind Steve that for some toxins the dose, whilst important, doesn’t make the poison, since no safe dose has been established for lead exposure (or Ionizing radiation, and a handful of other poisons). Can you tell that expression bugs me?

  10. SimonWon 18 Jan 2013 at 5:47 am

    Heptron – wikipedia has a great article on lead poisoning.

    The effects of lead are myriad, it has no known (positive) role in physiology. I believe the big concern are over fetal and early childhood development, one study (Schwarz J 1994) showed 2.6 IQ points dropped for each 10 mg/dl in blood serum. That said this is a lot more lead than one would hope to find in one’s blood, only 7% of US school kids had more than one thousandth of that much lead in one study. So whilst I’m keen to point out no safe level has been established, the levels found in most people since lead was removed from petrol are not a significant health concern.

    The use of graphs in the original article is I think misleading, and something we should as skeptics be wary of (and critical?). They’ve shuffled axises both vertical and chronologically to overlap two peaks. If we are allowing overlaying such graphs, then this one is much more convincing.


    Clearly one would expect a delay between childhood poisoning and adult criminality, but since lead affects the brain’s development, and people went through exposure at different times, and typically don’t suddenly become non-criminal (unless imprisoned) the effect should be more spread out in time than the cause. Since the graphs suggest a more immediate overlap they are probably evidence that lead had a very small or no effect on crime levels, and any correlation of the graphs is probably because of the way they have been scaled and manipulated rather than due to any effect from lead. In this case correlation implies lack of causation (I must get extra skeptic points for that).

  11. SimonWon 18 Jan 2013 at 5:54 am

    Oops Faust was per ug, unit confusion, so that would be more plausible.

    However I still think the graphs in original article are misleading.

  12. daedalus2uon 18 Jan 2013 at 9:21 am

    For all compounds exhibiting toxic effects there is a dose response curve.

    In the low dose regime, the effects are from the hormetic distortion of the normal regulation of physiology. For that, there is no threshold because the pathways that are normally regulating physiology to be a certain way are already in the active range, so perturbing them moves them to a different operating point of that active range.

    We know there are physiological pathways that regulate the degree of violence an individual exhibits. Exposure to violence increases this, as in the cycle of violence. That is most likely mediated through pathways activated during fight-or-flight. A final common pathway is very likely oxidative stress. Lead does cause oxidative stress, so it very likely does couple to other pathways controlled by oxidative stress.

  13. howdinion 18 Jan 2013 at 11:01 pm

    Hey Steve,

    l think this might provide some context and a preliminary rebuttal to Nevin’s article.

  14. steve12on 26 Jan 2013 at 9:55 pm


    Steve Sailer gets 0 points for dealing with Nevin’s paper. He doesn’t seem to understand that regardless of the age cohorts chosen or resolution thereof, the variation in crime rate is predicted very nicely (uncannily, in fact) by time lagged lead exposure.

    Also, he keeps talking about this whole “60′s freak out” thing which is irrelevant: the relationship is not weakened because it doesn’t account for some portion of the data you pick out of a hat as important. Predicting variation in some dependent variable (dv) from some independent variable (iv) does not mean that peaks, troughs, etc are somehow more important than any other part of the dataset. Obviously if the two are related, as one gets up the other tends to, but there are clearly other forces at work on the DV, so the lack of explanation of prediction of some peak X does not invalidate the statistical relationship between the 2 variables. Especially considering the political upheaval of the time that was also influencing behavior.

    He also avoids a lot of the other geo localized lead exposure evidence that does show a relationship. Ignoring evidence is his strong suit.

    IOW, nothing he writes really refutes the data in Nevin. Nevin has either screwed up (possible) or there is another variable that rises and falls with lead exposure.

    This could be due to the fact that Sailer has a dog in the fight, i.e., he is a racist ass hole. Sailer believes that blacks are genetically less intelligent than whites, so he needs to argue that environmental factors, epigenetic factors, etc. are not explanative in testing differences between races (of course intelligence is linked to criminality). He seems reasonable when you read him, but he ignores a shitload of data he doesn’t want to deal with to maintain his racist ideology, which is ugly unscientific nonsense.

    Of course this isn’t WHY he’s wrong re: Nevin. He’s wrong because of what I wrote above. But the reason he makes those arguments, IMO, is because he’s a racist ass hole. Lest someone accuse me of ad hom reasoning….

  15. steve12on 26 Jan 2013 at 10:19 pm

    One other point:

    There’s some confusion above about claims re: % of violent crime accounted for by lead. Nevin was reporting coefficient of determination (called R squared, literally the r value squared) which tells you the proportion of variability of your DV that’s accounted for by you IV(s).

    Nevin did indeed find that 90% of the variation (R squared =.9) in violent crime was accounted for by variation in childhood lead exposure, but this is not nearly the same thing as saying that 90% of crime is caused by lead exposure. We’re still looking at a correlation, so other factors that changed with lead exposure could be at play. Many things were improving in society over this time period, including exposure to other toxins, better teaching methods, etc.

    Another issue is that we’re we’re modeling the change in crime over time around some baseline. You could have a baseline crime rate (with many causes) that would be present no matter what, but we’re modeling the variation around that baseline over time, so it would be wrong to claim that this finding is accounting for the percentage of all violent crime. It’s modeling the change in crime rates over time, which is different thing.

    That said, the number IS really high, and does raise a furrowing of the brow to be sure.

Trackback URI | Comments RSS

Leave a Reply

You must be logged in to post a comment.