When states make it harder to get an abortion, the abortion rate falls.
A few months ago, SPPQ published an article by Michael New asking a simple question: Do anti-abortion laws have any effect? It’s a good question. Although the number of abortions performed in the U.S. fell by 22.2% between 1990 and 2005, it’s not clear what caused the decline. Maybe it was the abortion restrictions passed by many states during the 1990s. Maybe it was just a societal change in values. Maybe economic growth caused the decline, with more women deciding they could afford to raise a child. Maybe contraceptives became more effective, or used more often. It’s hard to say.
For a political scientist, of course, the political angle is most interesting. So New sets up a test to see whether new abortion restrictions played a role.
What kind of abortion restrictions are out there?
Prior to the Supreme Court’s 1992 Casey decision, states had very few options when it came to restricting abortion. Really, there were only two options. First, states could require parental notification or consent prior to a minor’s abortion, but those rules would affect only minors. Second, states could refuse to fund abortions via Medicaid, but those rules would affect only poorer women who used Medicaid.
After Casey, states were free to try other restrictions as long as they did not impose an “undue burden” on the woman’s right to pursue an abortion. So states came up with waiting periods, “informed consent” laws, and other innovations.
In his analysis, New looks at the effects of the two old restrictions (parental involvement and restrictions on Medicaid) and of one new restriction (informed consent).
What’s the effect of these restrictions?
New sets up the analysis well. He collects data on the abortion rate (# of abortions per 1000 women) and abortion ratio (# of abortions per 1000 births) in each state, each year. He uses state and year fixed effects, and also tries other panel corrections (such as panel corrected standard errors and other specifications). He includes a number of reasonable control variables. The analysis is set up well. Overall, here’s what he finds:
- Parental involvement laws do not have a statistically significant effect in the overall model, but when he looks only at the abortion rate for girls 13-17 years old, these laws reduce the in-state abortion rate among minors by around 15%.
- Medicaid restrictions reduce a state’s abortion rate by around 8-9%.
- Informed consent laws have roughly half the effect of Medicaid restrictions, reducing the state’s abortion rate by around 4-6%.
Are we sure that the laws caused these changes and not something else?
As a check on endogeneity, New slices the analysis a few other ways. For example, he shows that parental involvement laws affect abortion rates only for girls 13-17, while informed consent laws affect abortion rates only for women 18-45. If these laws affected all age groups equally, we might worry that underlying shifts in each state’s values were causing both (a) adoption of these laws and (b) declines in abortion. But since these laws have different results on different age groups (as we would expect), it seems plausible that the laws are actually causing the change.
Likewise, New shows that if two states pass similar laws, but the courts in one state nullify the laws, then the abortion rate declines only in the state where the restrictions stood. This is further evidence that the laws are causing the changes, not underlying societal shifts.
Although I am no expert on abortion policy, it appears that New has set up his models in an orthodox way. I see little reason not to accept his claim: When states make it harder to get an abortion, the abortion rate falls. We don’t know from this analysis what the causal mechanism is. It could be that women are more careful not to get pregnant, recognizing that it may be hard to abort the pregnancy. Or it may be that more women are carrying their children to term, becoming reluctant mothers. New’s analysis doesn’t get into that.
I have only a few concerns about New’s method. In a footnote (#14), New writes that he has chosen not to analyze waiting periods. Many states require women to visit a doctor twice to get an abortion. After their first visit, they must wait a set amount of time before returning for the actual procedure. New does not analyze waiting periods, apparently because they are multicollinear with other types of laws: “Waiting periods are typically enacted in conjunction with other types of anti-abortion laws, making them difficult to properly analyze.” If that’s true, then perhaps it’s the waiting periods–not the informed consent laws–that are driving the results in New’s models. Earlier (pg 32), he criticizes previous authors for failing “to examine the effects of individual types of anti-abortion laws in their analysis.” Perhaps the same could be said of New’s analysis.
Also, it would seem that New should interact his “Medicaid restrictions” dummy with an indicator of the state’s poverty rate. If Medicaid will not fund abortions in a particular state, the effect will be most pronounced among poorer women. He includes many economic control variables in his model (so many, in fact, that there may be a multicollinearity problem among these variables), but he ought to interact these poverty variables with the Medicaid dummy.