A Brief Analysis of the Donohue-Levitt Hypothesis
As we saw previously, there is much you should know before you get to enjoy the cozy sensation of being fully prepared and protected when you enter a discussion or debate about abortion.
Some of this knowledge takes the form of a simple number. Even in such cases, it is usually not easy to obtain and verify the piece of information, let alone analyze its impact on your arguments.
To make matters more complicated, much of the knowledge needed relies on a relationship between two or more variables. Let’s call these complex facts and deep-dive into one representative example: the relationship between abortion and crime.
1. The Donohue-Levitt hypothesis
Its scientific parents used several datasets in their attempt to explain the drop in crime in the 1990s and then the 2010s. To be clear: they never stated that abortion is the only factor behind the decrease in crime, but they did state its is among the most important ones.
2. How is it supposed to work?
The relationship between abortion and crime is not one where most of us would go ‘yeah, that one’s obvious’. So how is it supposed to work?
Here’s the bare bones version.
Unwanted children are more likely to engage in crime. Legal and safe abortion leads to fewer unwanted children, and thus lower crime.
But what exactly is ‘unwantedness’? While the phrase may be unfamiliar, we have, in fact, already covered all these reasons under the Circumstances pillar of our comprehensive view on abortion.
In brief, women may not want children because they
- don’t have enough money to provide for them;
- live in a difficult home environment (e.g. with abuse, violence, toxic relationships);
- are too young.
As a result of these or other unfavorable life situations, unwanted children may receive less financial, emotional, or even physical (e.g. breastfeeding) support. Unintended pregnancies are also associated with a higher incidence of substance use (smoking, drinking, drugs) by mothers.
What happens when abortion is an option? As Levitt himself describes it in this summary article of a Freakonomics radio episode,
In our hypothesis, what happens is that abortion becomes legal; women are given the right to choose; and what our data suggests is that women are pretty good at choosing when they can bring kids in the world.
3. Does it work?
According to Donohue and Levitt, it absolutely does. In fact, they concluded their original paper by stating that
legalized abortion can account for about half the observed decline in crime in the United States between 1991 and 1997.
Moreover, they found a similar relationship between abortion and crime between 1997 and 2014. In that period, states with high abortion rates saw crime rates fall 60 percent more than states with the lowest abortion rates.
The overall effect size also happened to age rather well with abortion accounting for 45% of the fall in crime in the 2010.
4. But surely, it’s something else than abortion!
The statistical relationship seems solid beyond dispute. As the number of abortions goes up, crime goes down. Correlation, however, does not equal causation. It could very well be be something other than abortion driving the effect.
If this is your first exposure to the idea that abortion may reduce crime, I’m almost certain you had similar thoughts. So did the authors. Applying experimental control was out of the question for obvious reasons.
So they did the next best thing and applied statistical control. Turns out, the effect between abortion and crime remains strong controlling for the effect of
- level of incarceration
- the number of police
- concealed handgun laws
- measures of states’ economic well-being: unemployment rate, income per capita, poverty rate, and state welfare generosity
- per capita beer consumption
- and age (only in their 2019 paper).
In addition, strong confirmation came six years after the original publication in a 2007 paper by Jessica W. Reyes. In her attempt to tell a different story, namely how the phase-out of lead from gasoline resulted in lower crime, she found that in her own models,
the effect of legalized abortion reported by Donohue and Levitt  is largely unaffected.
Okay, but does it only work in the US?
Additional evidence came from the Prague Study, which followed 220 children born to mothers whose request for termination of an unwanted pregnancy had been twice denied. Controlling for a host of variables, the authors of the study found that unwanted children were significantly more likely to end up in prison.
Even more to our point, François et al. (2014) provided evidence from 1990–2007 for 16 Countries in Western Europe. Their findings were consistent with the hypothesis.
5. Final conclusion
Make no mistake, the causal relationship between abortion and crime has not been ‘proven’. That’s why we call it a hypothesis, not fact.
As Cal Gabrin, my stats teacher in my early Ph.D. years hammered into us, in science it all comes down to replication and convergence.
Much of which still needs to be done but based on the evidence I presented above, I’m already reasonably (say, 95%) convinced that the negative relationship between abortion and crime is real.
I’m less confident that it’s as strong as they posit. Some of Donohue’s and Levitt’s own models and some included in their replications produced a smaller effect size. At this stage I’m quite (75%) confident that legalized abortion accounts for at least 25% of the drop in crime. As always, I’m happy to be proven wrong now or in the future.
6. What does it all mean?
What are the policy implications of abortion’s negative impact on crime?
Upon a first and shallow reading, the Donohue-Levitt hypothesis seems to support a pro-choice position on abortion. After all, it shows that legal and safe abortions have a positive impact on society. Therefore, the political right should hate it and the left love it, right?
Wrong. In fact, when it first came out, both political sides hated the hypothesis. Those on the left accused its authors of (at least implicitly) promoting eugenics along the lines of ‘Well, you’re killing these fetuses, so they never get a chance to grow up to be criminals.’
As Levitt recalled,
The number of death threats that I got from the left was actually greater than the number of death threats I got from the right.
In reality, the policy implications of the Donohue-Levitt hypothesis are not that simple. Recall for a moment the causal avenues through which it is purported to work.
If we somehow made sure that no child grows up in adverse conditions, and that all of them have access to key resources like love and money; not only would the relationship likely go away but so would most of crime.
As crime has the potential to impact everyone, and abortion disproportionately impacts the mother and the child, the arguments demonstrated in this article are very important to address when making up your mind on abortion.
Exactly how you do that is, of course, up to you. If you don’t trust me (and I’m no expert on abortion or crime), please do your own research and arrive at your own confidence intervals.
All I ask is you do that after reviewing the evidence presented by the authors and their critics, using the deep critical thinking tools you’ve developed in your professional and personal life.