Back in 2001, John Donohue and Steven Levitt wrote a famous paper linking the decline of crime to the rise of abortion. Put in blunt terms, it made some intuitive sense: unwanted children are more likely to grow up in circumstances that lead them into criminal behavior, so aborting these unwanted children is likely to reduce the number of criminals 20 years later. The paper came in for a lot of criticism, including some serious coding errors that were swept away with a bit of statistical sleight of hand, but D&L did make a prediction:
Roughly half of the crimes committed in the United States are done by individuals born prior to the legalization of abortion. As these older cohorts age out of criminality and are replaced by younger offenders born after abortion became legal, we would predict that crime rates will continue to fall. When a steady state is reached roughly twenty years from now, the impact of abortion will be roughly twice as great as the impact felt so far. Our results suggest that all else equal, legalized abortion will account for persistent declines of 1 percent a year in crime over the next two decades.
Now they’re back to check out this prediction, and they say that not only were they right, but the impact of abortion on crime was even higher than they initially thought. Now, it’s no secret that I think D&L are wrong: the important thing that happened in the 70s was the reduction of lead in gasoline, which by coincidence happened at around the same time as the increase in abortion. That said, if they’re still serious about the abortion-crime hypothesis, I have a question and a suggestion. First off, here’s a chart showing abortion rates, crime rates, and a stylized version of expected crime rates based on abortion rates:
Don’t take the precise levels of the “Expected Crime” line seriously. It’s just a mirror image of the abortion rate (the dotted gray line) and shows the rough shape you’d expect if abortion rates affect crime with a lag of about 20 years. What you’d expect is:
- A long period of high crime when abortion rates were low.
- A short period of falling crime as abortion rates rise.
- A long period of slowly rising crime as abortion rates decline.
The only one of these that checks out is the middle one. In the pre-Roe era, crime wasn’t steadily high. It was steadily low and then began rising in the mid-60s. And in the 2000-present era, crime hasn’t slowly risen, it’s just kept on declining.
It’s true that D&L predicted that when a “steady state” was reached, crime would continue to fall. But we haven’t reached a steady state. The abortion rate has steadily fallen ever since its peak in 1980. This should predict a slow rise in crime, but in fact crime has continued to decline.
So that’s my question: how well do the abortion and crime data really match up? And here’s my suggestion: one thing I’ve learned from my lead-crime research is that a single national-level correlation just isn’t enough to prove anything. The first paper showing a national lead-crime correlation was published in 1999, and the first paper looking at state-level differences (a strategy also used by D&L) was published in 2007. But it wasn’t until 2012 that I wrote about it. Why? Because those papers were intriguing but not conclusive. Much more evidence was needed, and it wasn’t worth a story until that evidence began to accumulate.
Here’s one way of accumulating more evidence: check out the abortion-crime correlation in other countries. The crime decline, after all, has been worldwide (though not identical in all countries). Here are four interesting countries to look at:
These are interesting because they aren’t similar to the US experience:
- Slovakia had a high abortion rate all through the 60s, with a peak in the 80s and then a decline.
- Australia had a sudden increase in abortion in the mid-80s.
- Poland had a high abortion rate that fell steadily for three decades before effectively reaching zero.
- Portugal had a low abortion rate that increased suddenly around 2005.
I don’t know how accurate these abortion figures are. That would require some serious research. But if you do this research, and then match up abortion rates with violent crime rates, you could find out if the abortion-crime hypothesis seems to pan out in different countries. It wouldn’t be a silver bullet, and nobody should expect perfect correlations, but we should expect to see rough shapes that match the hypothesis.
If D&L (or anyone else) are serious about the abortion-crime hypothesis, this is the kind of thing they need to do. Their theory really doesn’t deserve anyone’s serious attention until there’s more than just a modest correlation in a single country.