I have seen the Freakonomics idea that legalizing abortion caused a decrease in violence and I have been wary about it. They say that they control for lots of variables, but it still amounted to an observational study with their inherent problems. In particular confounding factors–just because you think you have thought of all the other variables, doesn’t mean you have. It now seems that exposure to lead (via here) is one such factor. I also have doubts about this study, but it seems stronger to me:
The centerpiece of Nevin’s research is an analysis of crime rates and lead poisoning levels across a century. The United States has had two spikes of lead poisoning: one at the turn of the 20th century, linked to lead in household paint, and one after World War II, when the use of leaded gasoline increased sharply. Both times, the violent crime rate went up and down in concert, with the violent crime peaks coming two decades after the lead poisoning peaks.
Other evidence has accumulated in recent years that lead is a neurotoxin that causes impulsivity and aggression, but these studies have also drawn little attention. In 2001, sociologist Paul B. Stretesky and criminologist Michael Lynch showed that U.S. counties with high lead levels had four times the murder rate of counties with low lead levels, after controlling for multiple environmental and socioeconomic factors.
In 2002, Herbert Needleman, a psychiatrist at the University of Pittsburgh, compared lead levels of 194 adolescents arrested in Pittsburgh with lead levels of 146 high school adolescents: The arrested youths had lead levels that were four times higher
This means that there is statistical evidence and medical evidence. If it’s thought that exposure to lead changes parts of the brain that make violence more likely, then that can be tested (they imply this has been done, but don’t really state that it has–they only point to more observational studies). If a medical theory predicts something and data shows that it happens that makes the conclusion stronger even if the data does come from an observational study. Still, I think this is stating things much too strongly:
Most of the theories have been long on intuition and short on evidence. Nevin says his data not only explain the decline in crime in the 1990s, but the rise in crime in the 1980s and other fluctuations going back a century. His data from multiple countries, which have different abortion rates, police strategies, demographics and economic conditions, indicate that lead is the only explanation that can account for international trends.
It’s only saying that of the variables they look at, lead is the only one that accounts for the trends. There could be other variables they haven’t looked at that explain things better.