This month, a study was released finding that a Colorado state-government program to provide free contraception of all kinds for low-income women had reduced teen pregnancies and abortions in the state by an incredible amount. The Washington Post reported that the program was “how Colorado’s teen birthrate dropped 40% in four years.” It turns out, though, that while those indicators are improving dramatically in Colorado, it’s hard to credit the program in question and a lot of liberal praise for the program is way overblown.
In 2009, the state began a program called the Colorado Family Planning Initiative (CPFI) that gave low-income women free or low-cost IUDs and subdermal contraceptive implants, both highly effective but relatively expensive long-acting reversible contraceptives (LARCs).
Said study on the program, published by the Guttmacher Institute, an influential think tank that studies abortion and reproductive health, reported that between 2008 and 2011, the birth rate for low-income teens in the state dropped by 29 percent, and the teen abortion rate dropped by 34 percent. A separate CDC report noted that Colorado’s teen birth rate has decreased by 39 percent over the past four years, while the state government found a 40 percent drop in teen births from 2009 to 2013. “The state attributes three-quarters of the overall decline in the Colorado teen birthrate to the program and said its success had a ripple effect,” the Washington Post reported.
Guttmacher summarized its findings as follows:
The Colorado Family Planning Initiative produced a radical game change in the state: The [long-acting contraceptive] methods it promoted and paid for appeared to contribute to a large decline in fertility among the young, low-income patient population and to a decline in the overall fertility rate among women younger than 25. At the same time, measurable declines occurred in abortion rates, births to young unmarried women with limited education and numbers of infants receiving WIC services.
While that’s carefully worded, this all overstates the program’s success and influence and ignores the fact that much of these effects probably would have happened anyway.
There were big decreases in both teen abortions and births in the Colorado countries benefiting from the program during its duration — but to say the program directly caused the huge decreases is a simplification that overstates the complicated relationship between contraception, abortions, and births.
Why? The teen abortion rate had been falling dramatically for a significant period of time, and with CFPI, it just kept falling. The study compared 2008 and 2011 abortion rates in counties where CFPI was available — in that time period, abortion rates for 15–19 year olds in those counties decreased from 10.9 per 1,000 women to 7.2 per 1,000, which is indeed a 34 percent decrease.
In that same time frame, abortion rates in counties without the program decreased from 14.4 to 10.2 per 1,000, a 29 percent decrease.
So how could one attribute the 34 percent decline in abortion rates to the CFPI? Almost the same reduction — about 85 percent of the reduction we saw in CFPI counties — still happened in places where the program wasn’t available.
This makes sense because abortion rates have been dropping steadily for years (including among younger women):
It’s more curious, that the abortion rate for women 20–24 rose slightly in the non-CFPI counties while that statistic dropped noticeably where the program was available. But with the very limited evidence the study presents, we have no idea if this is just due to random variation.
This study does not examine comparable women who happened to get LARCs through the program and those who didn’t, which would make for a rigorous study (of course, such things are often infeasible). Instead, it compares whole counties, where only a smaller subsection of women would have access to the program, and then attributes the changes in their overall abortion statistics to the program.
Moreover, the CFPI and non-CFPI counties aren’t remotely comparable: CFPI was in place in 37 of Colorado’s 64 counties, and those 37 counties contained 95 percent of the state’s population. The non-CFPI counties are quite rural, covering 37 percent of the state’s land mass but containing only 5 percent of the total population. This, in the chart above, is what stands in for a control group, essentially — it shouldn’t be considered any such thing.
Colorado’s teen abortion rate did drop more between 2008 and 2010 than the national teen abortion rate (unfortunately 2011 data isn’t available yet), but just barely — the rest of the country saw a big drop too. According to Guttmacher data, the national teen abortion rate dropped by 17.4 percent while Colorado saw a 25 percent decrease from 2008 to 2010.
The way the program was credited for a drop in teen birth rates is a little bit more complicated, but basically just as bad.
To arrive at the conclusion that the program reduced the teen birth rate for low-income teens by 29 percent, the Guttmacher authors projected rates of low-income teen pregnancy based on previous years, essentially drawing a straight line out into the future (“linear trend lines,” in statistics speak), and then counted any performance below that line as resulting from CFPI:
But these “linear trend lines” based on the three previous years of data aren’t really useful. The authors’ projection shows that births would actually increase a bit during the period CFPI was put in place, despite the fact that, like the abortion rate, the teen birth rate is declining nationally, noticeably and steadily:
So sure, the low-income teen birth rate did decrease relative to previous years, but without a control group, it’s impossible to know what percent of that decrease the contraception program is responsible for.
And it’s clear that Colorado’s program can’t responsible for most of the state’s drop in birth rates when you compare the size of the reported effects to CFPI’s scope. Overall, only 8,435 low-income women received a LARC during the duration of the program. Some back-of-the-envelope calculations given the fertility rates the authors reported for low-income teens show that a maximum of 700 or 800 out of 11,000 or so predicted births, in a given year, were likely to be prevented by the CFPI, which would translate into a maximum 6.8 percent reduction of births in CFPI counties. The birth rate in the counties with the program, remember, dropped by 29 percent — four times as much as the free LARCs could have accounted for.
In fact, Colorado’s overall teen birth rate dropped by 40 percent from 2009 to 2013, 11 percent more than the low-income teen birth rate did from 2009 to 2011, suggesting there are other factors at work here.
There is likely some combination of factors driving the low teen birth rate in Colorado. The state also ramped up training and technical assistance for family-planning clinics and helped them to expand outreach, in addition to the new free provision of LARCs (which was paid for with a $23 million anonymous private donation). It’s also possible that some of the factors decreasing birth rates nationally have had a larger impact in Colorado, though there’s really no way to tell. As Sarah Kliff discusses in Vox, the cause of the national decrease in the teen birth rate is still a mystery, with theories ranging from increased IUD uptake overall to the popularity of the MTV show 16 and Pregnant.
Of course, researchers have limitations: The authors tried to measure the impact of a pre-planned public-health initiative after the fact, not conduct a gold-standard statewide experiment. In that sense, they did a pretty good job, but the study does not justify the headlines it got – it seems likely that the CFPI must have had some positive effects on abortion and birth rates, but it’s far from the policy panacea the headlines depicted. For a study titled “Game Change in Colorado,” it provided little evidence that the CFPI changed the teen-pregnancy-prevention game at all.
There also remain plenty of unanswered questions about CFPI as health policy: This study doesn’t cover data on discontinuance rates, reinsertion rates, changes in STI transmission, or on many other factors that are important. For instance, it’s possible that since LARCs are effective for a number of years, birth rates could increase again in a few years when the devices expire, especially if women forget to replace them or delay replacement due to cost. Another concern is that LARCs could increase STI transmission because they replace the need for traditional barrier methods to prevent pregnancy.
So when weighing the impact of the results, we should be careful to take the results for what they actually are: an indication that programs like the CFPI increases the uptake of LARCs, in the short term, to a limited degree, which might have positive effects on birth and abortion rates for some people, in some populations, in some places.