Free Contraception Can't End the Abortion Debate
Ross Douthat, the resident social conservative of the New York Times opinion page, doesn't usually get angry. But in a blog post Wednesday, he unloaded on liberal columnists who suggest that defunding Planned Parenthood is going to cause more abortions by depriving women of access to birth control. Such columns, he argues, rest on the assumption, stated or not, that America's abortion rate is driven by a lack of that access -- and that if only we made birth control more widely and cheaply available, we could resolve America's rancorous abortion dispute by rendering the issue moot.
I ended up in a long Facebook discussion with one of the columnists he attacked by name, Damon Linker. Before I go further, I should say that both Ross and Damon are former colleagues, and I think they're both brilliant columnists. But on this issue -- not whether abortion should be legal, but whether the main thing standing between America and fewer abortions is better government birth-control policy -- I think Douthat has the better argument.
Damon, and many others who participated in that Facebook thread, voiced the belief that America has far more abortions than it otherwise would because conservatives are not doing enough to give women better access to contraception. This belief seems intuitive -- after all, birth control keeps you from getting pregnant, and it's hard to get an abortion if you're not pregnant. But when I look at the data, I am reminded of Ambrose Bierce's definition of prejudice: a vagrant opinion without visible means of support.
Consider abortion rates in various developed nations:
The U.S. rate is certainly high compared with, say Germany or the Netherlands. On the other hand, it's lower than Sweden, and right around that of New Zealand and the UK -- countries with comprehensive national health-care services that provide birth control. And who had one of the lowest abortion rates? Ireland, where it was illegal. (Irish women travel abroad to get abortions, but the rate still seems to be quite low by international standards.)
Now, obviously, we could theoretically do something to reduce our abortion rates to more German levels without going so far as to ban it. But this data doesn't really suggest that "something" is necessarily "provide more affordable reproductive health care services to women," or indeed, anything else that lends itself to government intervention, such as "better sex ed." Extremely high abortion rates can coexist with extremely comprehensive health-care systems and liberal social norms.
You see a similar pattern in the U.S. when you look at the variation in abortion rates between states: Liberal blue states with liberal abortion laws and liberal attitudes about birth control seem to have the highest, not the lowest, rates of abortion. What drives this? I can come up with a number of plausible theories, but I couldn't tell you which one is right. On the other hand, I think we can reject the hypothesis that liberal attitudes toward sex and birth control are a surefire way to get the abortion rate down.
In fact, the evidence for this thesis was never very good. Even William Saletan, who used to be a leading advocate of the squishy pro-choicer thesis that abortion is terrible so we need to give people lots of free birth control, ended up abandoning this thesis when he concluded that there's just not good data showing that the high price of birth control, or the inability to get your hands on the stuff, is the major reason people end up having abortions.
But wait, I hear you cry! What about Colorado?
Ah, yes, Colorado. A little while back, some Colorado researchers published a study on the Colorado Family Planning Initiative (CFPI), a program to provide long-acting reversible contraceptives (LARCs) to low-income women, called . Here are the results: "In CFPI counties, the proportion of births that were high-risk declined by 24% between 2009 and 2011; abortion rates fell 34% and 18%, respectively, among women aged 15–19 and 20–24. Statewide, infant enrollment in WIC declined 23% between 2010 and 2013. CONCLUSIONS: Programs that increase LARC use among young, low-income women may contribute to declines in fertility rates, abortion rates and births among high-risk women." This made columnists very excited. A 30 percent decline in the teen birthrate, and a huge decline in the teen abortion rate (with a smaller decline in abortions for women aged 20 to 24), for such a cheap program? Yes, please.
Here's the problem: That study is not very good. This is the part where I spend a lot of time wading through the numbers. If that sort of thing bores you, just accept arguendo that the numbers don't add up very well, and skip to the concluding paragraphs.
Basically, what you would want to do with a study like this is randomize it: Divide women into two populations, give half of them increased LARC access, and track what happened. Failing that, you'd at least want to track what actually happened to the women who got LARCs. The study does neither of these things; instead, after the program had been operating for a while, it took a look at county-level data, either for the whole population of the county, or the population of low-income women in the counties where LARCs became more widely available. Then it shows you that various things, such as teen births and abortion rates decreased. There's not really much evidence of a causal link, but there are so many pretty charts and tables that you don't really notice its absence until you start to work through the underlying numbers.
By 2011, 8,435 women aged 15 to 24 had been given LARCs, out of a total population of about 420,000 women who fell into that age range for at least some of the study period. (At any given time, in Colorado, there are about 320,000 women aged 15 to 24.) Between 2008 and 2011, the study reports that the abortion rate in CFPI counties had fallen from 10.9 to 7.2 per 1,000 women aged 15 to 19, and from 22 to 18 per 1,000 for women aged 20 to 24. Doing some quick math, that gives me about 1,200 fewer abortions in a single year. To believe that the CFPI was responsible for all or most of the change, we have to believe that around 15 percent of the women who got LARCs would have gotten pregnant and had abortions in a single year -- and that's before we even try to account for the fact that some substantial number of women who got those 8,400 LARCs would have aged out of the range by 2011, and therefore would presumably not have been counted in the abortion data, so that the remaining women who were would need to have had even higher rates of pregnancy and abortion. Just for reference, the pregnancy rate in the U.S. for teen women aged 15 to 24 was about 6.5 percent, and about 15 percent for women aged 20 to 24. Obviously, not all of those women have abortions.
Now, pregnancy rates for low-income women are much higher than normal, and the unintended pregnancy rate is well north of 10 percent. That is still not high enough to account for those numbers unless you literally assume that the women who got LARCs would otherwise have had extremely high rates of unintended pregnancy, and that all of the ones who got pregnant would have had abortions.
It's not safe to assume that, however, because the study also contains a substantial section on birth rates during that time period. They fell substantially, and the LARC "game change" is also, by implication, credited with this decline. Now, abortions and births are pretty much mutually exclusive, so you have to add the two numbers together to get the implied number of pregnancies the researchers believe have been prevented in the target population. The paper says that "Between 2009 and 2011, the fertility rate for all Colorado women aged 15–19 declined 26%, from 37 to 28 births per 1,000. In the same period, the fertility rate among all women aged 20–24 declined 12%, from 89 to 78 births per 1,000. An estimated 77% and 74% of the decline among these age groups, respectively, can be attributed to the decline in births among low-income women in the CFPI counties." But if we do the math on those numbers, with the known number of Colorado women in those age groups, I get somewhere in the neighborhood of 2,200 fewer births among low income women in the CFPI counties. If we attribute that decline to the LARCs, we get an incredible 3,400 prevented pregnancies in 2011, or around 40 percent of the women who received them. That seems way too high. Of course, the LARCs might have contributed something to the decline -- in fact, I'd assume they did. But how much? The paper can't tell us.
Did Colorado's birthrate fall significantly faster than the nation's? Yes. So did Florida's, Georgia's, Arizona's, Maryland's, Delaware's, Mississippi's, and Massachusetts', to name just some of the biggest outliers. Probably all of these states had some program that you could suggest helped lower the birthrate. Probably many of the states that didn't see declines also had such programs, only since their birthrate didn't decline spectacularly, no one wrote that paper. This is the sort of thing that makes John Ioannidis suspect that most published research results are false.
The researchers attempt to control for this problem by looking at counties where the CFPI didn't operate, and comparing their abortion rates with the ones where it did. Problem: "Beginning in 2009, funding was provided to 28 Title X– funded agencies in 37 of Colorado’s 64 counties, which contained 95% of the state’s total population, including 95% of the low-income population (defined as individuals with incomes at or below 150% of the federal poverty level)." The remaining counties, a footnote tells us, were rural places with a lot of land mass and very few people. This is not a good control group for the counties with all the people in them. For one thing, if we assume that the age ranges are roughly similar, we are talking about only 16,000 women out of roughly 320,000. More importantly, rural people tend to differ from urban and suburban people in their experience of poverty, cultural and political values, and other ways.
Perhaps most importantly, the teen abortion rate also declined precipitously in the non-CFPI counties -- by 29 percent, compared with 34 percent in the CFPI counties. The idea of a big difference in the abortion rate ends up resting entirely on a difference between the abortion rate among women aged 20 to 24 in non-CFPI counties, which ticked up slightly, and those in CFPI counties, where it fell by 18 percent. Given how few people lived in the non-CFPI counties, the fact that they already had substantially higher abortion rates, and the fact that this effect only seems to operate on older women even though the clients of the clinics are roughly split between the two age groups, this starts looking more like Texas Sharpshooter Fallacy than a robust result.
That's the whole study. They tell you that Colorado gave out LARCs; they tell you that birthrates and abortions fell. They don't dwell on the simultaneous fall in these areas at the national level, which is somewhat mysterious but may be fallout from the financial crisis. They have no way to establish causality except for an inadequate control group that doesn't even show a substantial difference in teen abortion rates, a fact that they appear to have forgotten in other sections of the paper. (They also mention that the overall decline in Colorado's birthrates was concentrated among low-income women in the studied counties, but that's not actually very interesting, because early motherhood and unintended pregnancy are also concentrated among low-income women.)
The authors basically concede that they cannot come close to establishing causality, because the summary conclusion is a weasel: LARCs may contribute to a decline in fertility and abortion among high-risk women. How much? We don't know, but look at the pretty graphs! Needless to say, most of the people who have hyped the study missed the weasel words and treated the results as if the study had shown a direct causal link, when in fact it does no such thing.
The universe contains many more correlations than causes. Is this correlation plausible? Sure. But that doesn't mean it's there. And this paper simply doesn't show that it is.
Yet even if this paper actually did make a reasonable stab at establishing causality, and we accept that Colorado produced some sort of miracle, it seems worth pointing out that for all this, their abortion rate is still not zero. It's almost 2 percent of women aged 20 to 24, and almost 1 percent of women aged 15 to 19. Overall, women in this age range had something north of 4,000 abortions in Colorado in 2011. If you think that abortion is murder, it's obviously good news that this was 1,200 fewer than they would have had in 2008. But the 4,000 murders are probably still going to bother you quite a lot. And this is the result of giving away free or low-cost birth control that is basically impossible to screw up to a target population of the poorest women in the state.
Every time I've written about this issue before, I've ended up concluding the same thing: the primary barrier to people using birth control is simply not the price, or the distribution. There are a host of reasons that people don't use birth control, from concern about side effects, to cultural values, to simply not anticipating their need for it. And that matters for policy, because the government is relatively good at providing free stuff, and relatively bad at getting people to change their most intimate behaviors. I'd certainly guess that offering women cheap LARCs instead of other birth control methods does something to prevent abortions and unwanted births. But I certainly wouldn't assert that it was large based on this research.
That's not to say that the government shouldn't help low-income women get access to birth control. It's just to say that there's not a lot of great evidence that doing so will knock our abortion rate down to German or Austrian levels. When I pointed this out yesterday, someone I was arguing with said that even if this study was weak, he believed that well-crafted policy could make a big difference. And that may well be so. But unless you can actually articulate that policy, and provide some evidence that it works the way you believe, you're in a weak position to criticize pro-lifers for not embracing it.
For squishy pro-choicers -- a category I have spent many years in myself -- believing that free birth control can reduce abortion to negligible levels is very convenient. It is the "torture doesn't work" of the feminist movement. Instead of forcing a hard moral choice between the autonomy of the woman and the value of the potential life that is terminated, belief in the birth control fairy lets us off the hook. It would certainly be lovely if programs such as the Colorado Family Planning Initiative could make abortion so rare that it no longer required us to make those moral and political choices. But hoping for something to be right doesn't make it so -- and neither does bad data.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author on this story:
Megan McArdle at email@example.com
To contact the editor on this story:
James Gibney at firstname.lastname@example.org