Last week I talked about the data that suggests lead might have been partially to blame for the rise of violent crime around the world in the 1990s, and its removal from things like the gasoline we put in our cars is partially to blame for the subsequent drop in crime in the later ‘90s. If you don’t feel like watching that video, I’ll sum it up like this: lead is without a doubt VERY bad for humans, especially young children, but crime is a complicated subject and so we can only say that there’s very good evidence that lead contributed somewhat to that 1990s crime wave.
In response to that video, a WHOLE LOT of you wanted me to know that instead of or in addition to the lead hypothesis, abortion is to blame for those 1990s crime statistics. Like, this is very clearly an idea that has really caught on amongst laypeople, which I absolutely get: it’s a tidy explanation that will make someone at a party think you’re very clever and interesting, and it appeared in a wildly popular mainstream “science” book that presents it as a fact with a lot of evidence behind it. So I wanted to make this little follow-up video where I inform you that the abortion-crime hypothesis is NOT as well-supported as many people seem to believe. I also want you to know that the wildly popular mainstream “science” book it’s found in, Freakonomics, is overall pretty dodgy when it comes to actual science.
By the way, yes, it’s another video about economics. I’m really in my economics era, aren’t I? What an incredibly boring era to be in.
Anyway, in 2001 economist Stephen Levitt and law professor John J. Donohue III wrote an article arguing that Roe v. Wade, aka the legalization of abortion in the United States, was responsible for a full 50% of the reduction of crime in the 1990s.The idea is this: in 1970, five US states legalized abortion. In 1973, Roe v. Wade extended this to the rest of the country. Approximately 20 years later, crime dropped, first in those five states and then a few years later in the rest of the US. Levitt and Donohue believe this is causal, explaining that women now aborted unwanted fetuses; unwanted fetuses would otherwise turn into unwanted babies, who receive fewer resources and experience worse health outcomes than wanted babies, meaning they’re more likely to become criminals. Legalized abortion deleted that risky population and thus 20 years later we reap the benefits of Roe v. Wade.
While that article was published in 2001, Levitt would popularize it in 2005 with his book Freakonomics, written with New York Times journalist Stephen J. Dubner.
Now, there are a lot of problems with all this so I’ll try to just flag a few of the big ones. First of all you need to be very skeptical of that proposed effect size: 50% of the reduction in violent crime is due to ONE cause? If you’ll recall, very good data suggests that the reduction of lead in our environment MAYBE contributed anywhere from 7–28% of the drop in homicides. Some economists dropping one paper on the topic definitively stating that their pet hypothesis is responsible for fully half is ridiculous.
Second, that paper had some data issues. In 2005, some other economists discovered a computational error in Levitt and Donohue’s paper that artificially inflated the effect size, which Levitt and Donohue acknowledged but then said that doing some other statistical tweaking resulted in results that were just as strong IF NOT STRONGER than their original paper, so there. This is just one example of a pretty exhausting back and forth between various researchers arguing over how to crunch this data to get a true representation of reality. As Mark Twain once quipped, there are three kinds of lies: lies, damned lies, and statistics. This is very messy real world data and unlike the relatively clear data on lead, it kind of seems like you can get it to say whatever you want based on what you believe.
There are also some inconvenient facts that Levitt (I’ll refer to just him now since he authored both the original paper and the later book) skips over or hand waves away: for instance, if the decrease in violent crime was due to the population of “wanted” babies born after 1973 coming of age, we would expect the crime statistics to first decrease in the youngest adult population, and then following that population as they age up. But instead the data shows the reverse: violent crime decreased in older populations first, and then moved down into younger populations.
Arguments like this are why I honestly cannot understand why Levitt doesn’t just throw his hands in the air and say “Okay, maybe abortion legalization isn’t responsible for HALF of the decrease.” Because if, for example, he argued that it explains 7-28%, like the lead hypothesis, than you can say “well other factors are the reason why the age-related data looks like that.” And you can point to the many other reasons why experts think crime dropped, like the crack epidemic ending, or new strategies of policing, or new advances in medical interventions that result in fewer people dying.
That would also address some other common problems, like how this relationship between abortion and crime doesn’t really tend to show up in other countries that have changed their abortion laws. In Freakonomics, they argue that it does, but the actual data is really, really messy. Some studies find some effect in some countries, and many other studies find nothing, like this 2007 study of seven European countries that found it was demographics and incarceration rates, and NOT the legality of abortion, that explained crime rates.
So there are a lot of issues with the abortion-crime hypothesis, but one I rarely see brought up is the fact that there’s a very good chance that Levitt based this entire idea on an initial premise that is simply not true: that legalizing abortion drastically alters the number of abortions women get. Levitt calls this “true virtually by definition.” It is not.
This is actually a point I bring up a lot so I apologize if you’re hearing this from me for the 8,000th time, but we have a lot of data that suggests that restricting access to abortion doesn’t actually change the number of fetuses that are aborted. All it changes is the number of women who die trying to get the abortion.
For instance, in 2017 the Guttmacher Institute published a report that found that “Abortions occur as frequently in the two most restrictive categories of countries (banned outright or allowed only to save the woman’s life) as in the least-restrictive category (allowed without
restriction as to reason).” Meanwhile, the percentage of unsafe abortions was 1% in those least restrictive countries but 28% in the most restrictive.
We don’t have perfect data for the number of abortions performed prior to Roe v. Wade due to the overwhelming number of home abortions combined with the stigma that would limit self-reporting, but the data we DO have shows that abortions were extremely common, as detailed in a 2008 paper by Dr. Kate Gleeson of Macquarie University.
Legalizing abortion in the 1970s didn’t necessarily affect the birth rate in the United States, but do you know what did? The invention of oral contraception in the 1960s, which, as I discussed in my video about Nobel Prize winning economist Claudia Goldin, ABSOLUTELY prevented unwanted births and DRASTICALLY changed life for the better thanks to enabling women to have control over their lives and thus the ability to delay marriage, complete schooling, and enter the workforce.
But for some reason, Freakonomics ignores all of that. For a hypothesis based upon women’s actions, women’s own motivations seem curiously lacking.
So, was the legalization of abortion 50% responsible for the drop in crime in the 1990s? Absolutely not. No. But if you want to argue that it might have made SOME difference, I certainly wouldn’t tell you you’re definitely wrong. I’d just expect you to have the equivalent amount of extraordinary evidence to support an extraordinary claim with a lot of data going against it.