Get the Lead Out, Part 1

The rate of violent crime in this country has been declining for decades, and although there several plausible hypotheses, no one really knows why. Kevin Drum argues in the January Mother Jones that changes in lead emissions from automobiles have been the primary influence on violent crime in this country. Why didn't I think of that?

Here's the argument, in brief. Violent crime began to increase in the '60s, peaked in the early '90s, and has been declining ever since. Neither demographic changes, i.e., increases and decreases in the number of young men, nor changes in the economy can fully explain this pattern. Lead in the environment comes from two major sources: lead paint, which was gradually phased out during the last century, and gasoline. Following World War II, Americans began driving a lot more, and the oil companies added lead to gasoline, allegedly to improve engine performance. In the mid-'70s, due to evidence that lead exposure reduced I. Q., government forced the oil companies to switch to lead-free gasoline. Lead has its greatest influence on the developing brains of children. The results of lead on violent behavior becomes apparent when people are in their early twenties. Here are the data (originally compiled by by Rick Nevin) showing the relationship between the concentration of lead in the environment and the rate of violent crime 23 years later.

It's obvious that there is a correlation. But like many social scientists, I've spent my career telling students that “correlation does not mean causation.” Whenever a variable, A (lead), is correlated with another variable, B (violent crime), there are three possibilities: A causes B, B causes A, or some third variable, C, is responsible for the apparent relationship between them. Since no one is arguing that violent crime causes lead to be deposited in the environment, the real issue is whether confounding variables (Cs) have been ruled out.

Correlational arguments can be strong or weak. They are relatively strong if they have been replicated using several different data sets and research methods, and if alternative explanations can be ruled out. The two primary methods of determining whether a social policy, such as lead abatement, influences a behavioral variable are time series and comparison group designs. In a time series design, you look at whether a change in the policy is followed, at an appropriate interval, by a change in the behavior. Those are the data in the above graph.

In a comparison group design, you compare different spatial locations that have different levels of the presumed cause to see whether they also have different incidences of the presumed effect. The switch to unleaded gasoline was not uniform among the 50 states. Reyes found that states that switched to unleaded sooner saw their violent crime rate drop sooner. Nevin has examined the relationship between lead and crime in several countries, and has found that lead predicts differences in crime rates both within and between nations. Mielke has compared lead concentrations in various U. S. cities with the same results. Mielke has also measured lead concentration in New Orleans soil samples—which is quite unevenly distributed. He finds that it predicts crime rates at the neighborhood level. There are also data relating the crime rate to the distance one lives from a major highway.

Turning to alternative explanations, all these studies do a reasonable job of statistically controlling for confounding variables such as race, gender, socioeconomic status, family demographics, and unemployment rates. Of course, the number of possible alternative explanations is theoretically infinite, so you can never anticipate all of them. It is important to note that these researchers are not suggesting that lead is the only variable that influences violent crime, only that it is much more important than has been generally realized.

The relationship between lead and violent crime has been confirmed at the individual level in longitudinal studies. Researchers at the University of Cincinnati have followed a cohort of children for 30 years. Those with higher levels of lead in their bloodstreams as children are more likely to be arrested for violent crimes as adults. Lead exposure is also related to lower I. Q., attention deficit disorder, and higher rates of teen pregnancy.

Furthermore, plausible physiological mechanisms to explain the lead hypothesis have been proposed. The Cincinnati group compared the MRI brain scans of adults who had high or low lead exposure as children. Those exposed to more lead had less gray matter in the prefrontal cortex, the area of the brain associated with “executive function”—attention, verbal reasoning and impulse control. They also have a thinner myelin sheath around the synapses which connect adjacent neurons, suggesting that their communication channels within the brain are slower and less reliable.

Tomorrow:  Part 2