Colorado public health employees have come out with a new study purporting to show that providing free intrauterine devices (IUDs) and contraceptive under-the-skin implants for young women reduces abortion rates and births. The idea is that simply removing the cost barrier to such long-acting methods taps a well of demand among girls and women, leading to decreased undesired conception and hence to decreased abortion and decreased childbirth, especially among the young, unmarried, and uneducated. (Here I will set aside the contested empirical questions about whether these methods cause implantation failure and hence are abortifacient in effect. When I use the term “abortion” in this post I mean “deliberate abortion of a known pregnancy.”)
I have critiqued in detail a previous study of this type carried out in St. Louis. One important difference between the Colorado program and that in St. Louis: in Colorado, there was no specific study group. So some of the criteria for the St. Louis study group might have made the group unrepresentative of the women who are the target of anti-fertility policies. (Participants, for example, had to be willing to change their method of contraception.) The absence of a specific study group, however, has a disadvantage in that it was impossible even in principle to examine birth rates and abortion rates among actual study participants. The Colorado study was an “ecological” study, meaning its initiative was introduced in a region, and outcomes were estimated from public health records. The study used counties as proxies for a study group and a control group, comparing outcomes in counties where the long-acting contraceptive initiative was carried out at family-planning clinics with outcomes in non-program counties.
Some Study Flaws
I have obtained a full copy of the published article, and several points are noteworthy. Two outcomes—abortion rates overall and birth rates to young, unmarried, and uneducated women—did appear to go down at a greater rate per 1,000 women in the program counties than in the non-program counties. It is important here to remember that correlation does not equal causation, a point blatantly overlooked in breathless news stories like this one, which said Colorado “has found a way to significantly reduce teen pregnancy rates.” It is possible that there was such a causal effect, but the inference should not be drawn automatically. Were there other relevant changes in those counties during the same time period, such as, say, increased pro-life presence and counseling?
Moreover, it’s hard to make any evaluation because all results are normed to “per 1,000 women,” no raw data is given, and we don’t know the sample sizes. In program counties, the abortion rate for girls 15-19 years old went down between 2008 and 2011 from 10.9 per 1,000 to 7.2 per 1,000. In non-program counties, the rate also decreased—from 14.4 per 1,000 to 10.2 per thousand. The former is a higher rate of decrease (34 percent decrease versus 29 percent decrease) partly because the abortion rate was lower in program counties to begin with. The difference between rates of decrease is not large, and to tell how statistically significant the difference of rate of decrease is, it would be quite helpful to know the sample size. Among 20-24-year-olds, the abortion rate decreased in the same time period by 18 percent in program counties and rose slightly (by 6 percent) in non-program counties. But again, to gauge even the statistical significance of this change it would be much better if we had some idea of sample size, and drawing conclusions about causes is more difficult still.
As for birth rates, they declined in Colorado as a whole, and the study authors do not say how they came to the conclusion, touted in the press, that a large percentage of this decline was a result of declines in births among poor women in the target counties. The authors break out what they call “high-risk births”—that is, births to girls and women under age 25, unmarried, with less than a high school education—by county, showing the rates both in program counties and in non-program counties, and there was indeed a difference there between the rate of decrease in births (24 percent decrease versus 6 percent decrease). But they do not do such a breakdown for all births. So the conclusion that the majority of the decline in all births in Colorado is a result of a decline among poor women in target counties is unjustified in the publication.
What About High-Pressure Sales Techniques?
A question that arose concerning the St. Louis study is pertinent here as well: Were high-pressure sales techniques used to convince girls and women to have IUDs or implants inserted? As I documented, the St. Louis study architects were quite frank about their aggressive attempts to induce women to accept these methods. There was no question of simply making the methods available and telling them that there would be no cost while giving unbiased, balanced information about risks and benefits. Their goal was to promote the methods. For example, the St. Louis study report says things like this:
All participants were read a brief script informing them of the effectiveness and safety of LARC methods at initial contact…
Contraceptive counseling included all reversible methods but emphasized the superior effectiveness of LARC methods (IUDs and implants).
In addition, the project provided education to promote the use of the most effective contraceptive methods, IUDS, and implants in an effort to alter population outcomes.
Was the same true in the Colorado study? The Colorado authors say less about how the methods were offered, but there are indications that similar techniques may have been used. They cite the St. Louis study and imply that it was similar to theirs (p. 1). They refer to “policies to make LARC methods more acceptable and accessible to young women, including targeted outreach using social media…” (pp. 1-2). At one point, they say,
The initiative’s efforts intensified over its first year and a half, when more than 150 Title X–funded staff were trained in insertion and counseling techniques, outreach efforts increased, and word of mouth about the availability and acceptability of the methods spread (p. 2).
This language suggests a more aggressive approach than merely making the methods available at no cost. The question is important, because the conclusion we are supposed to draw is that merely removing the cost barrier induces many women to use these methods. That, in turn, is supposed to be relevant to public policy. But if the methods were aggressively up-sold, removing the cost barrier was not the only factor, and any generalization of the program would need to involve generalizing the biased promotion as well, which should be controversial as policy.
Did Sexual Diseases Rise?
Another question completely unaddressed is whether sexually transmitted disease (STD) rates among women rose more in the target counties than in non-target counties. It seems at least possible that there would be such an increase if women using these methods either engage in more promiscuous intercourse or stop having their male partners use condoms. Similarly, the study does not address rates of side effects or possible side effects of the methods themselves, such as pelvic inflammatory disease, which can be caused or exacerbated by inserting an IUD.
The study authors do allow themselves a moment of lyricism about the alleged benefits of the policies they favor:
Women’s ability to avoid unplanned pregnancy through the widespread use of the most effective methods can provide a level of well-being unimagined even a decade ago, as young women may be better able to continue and complete their education and enter the labor force without having to care for young children at the same time. Furthermore, the Affordable Care Act has the potential to replicate the success of the Colorado experience across the nation (p. 7).
Any conclusions about “a level of well-being unimagined even a decade ago” should at least take into account side effects of the methods and possible unintended consequences, such as STD increase.
Suppose that we imagine a society in which promiscuity is widely accepted and even encouraged by the powers that be, in which children are given detailed sexual information without parental consent from a young age, and in which abortion is legal and widely available as an accepted form of birth control. In other words, imagine America today, complete with policies that social conservatives have always strongly opposed. Now suppose that we hold constant those factors and at the same time coercively (albeit temporarily) sterilize all girls and women, or even all unmarried girls and women, ages 15-24. No doubt both abortions and births will decrease in the target population. But this sounds more like the plot of a dystopian novel than like the dawning of an era of previously unimagined female well-being. Nor is it a test of socially conservative versus liberal sexual politics, since the factors held constant, which increase abortions and unwed pregnancy, are heavily advocated by liberals and opposed by conservatives. By the same token, programs in which sterilization takes place not by full coercion but by pressure and promotion, while the other issues are left in place, do not constitute such a test.
Policies like the Colorado initiative ultimately fail to address the underlying causes of unmarried pregnancy and of abortion, and they are cavalier about the negative effects on women of both promiscuity and long-acting sterilization. Even if a study were sufficiently well-designed and contained enough published information to provide strong evidence that aggressively pushing long-acting contraception to vulnerable girls and young women reduces these particular outcomes (which hasn’t happened yet), society would not have a good reason to adopt the liberal policy package.