So in my prior post, I observed that the conventional wisdom that “the only people who get abortions are those experiencing hardship, and that means poor women” is not necessarily true. I pulled out some statistics from a study I had read which suggested that middle class women are just as likely to abort, but, after having published that post, I wanted to check the reliability of that study (which draws the conclusion, “poor women aren’t having as many abortions as they should, so for equality reasons we should make the procedure more widely available and free!” so there could have been a bias toward finding what they wanted to find. With that in mind, here’s how the Brookings scholars did their math in “Sex, contraception, or abortion? Explaining class gaps in unintended childbearing.”
To begin with, their data source was the National Survey of Family Growth from 2011 – 2013. This survey includes 10,416 men and women from ages 15 – 44. For their study, they actually wanted to look specifically at women at risk of unintended pregnancy, so they narrowed this group down to “single women not trying to get pregnant,” a group of 3,885 women in total. It’s reasonable enough to exclude women who describe themselves as actively trying to get pregnant, and they rationalize the exclusion of married women because they may be “less cautious about contraception” and “more likely to continue with an unintended pregnancy” and because “marriage is also correlated with income.”
They then divide the women into multiple groups by income level. They don’t explicitly state whether this is household income or individual income and, looking at the questionnaire, I can’t actually find the applicable question. Of course, for a single woman on her own it doesn’t matter, and maybe, in any case, survey-takers don’t count as “household income” that of an unmarried cohabitating partner, but you’d think it would make a difference for teens and young adults who are still dependents.
They assess whether there is a difference in frequency of sexual activity by income group, and conclude the answer is “no” (though, as I griped in my original 2015 blog post, they use as the marker having had sex at least once in the past year when, of course, more frequent sex is duh connected up with more chances for pregnancy; they use this statistic because, so far as I can tell, this is all the survey asks, or, rather, it asks about number of partners but that doesn’t really get at frequency with a given partner.
They then identify that, in this subgroup, poor women are considerably less likely to say that they had used a method of contraception — 15.9% of poverty-level women and 14.6% of near-poor (100 – 200% of poverty level) said they had used “no method” of contraception in the past year; all other income groups were at less than 10%. This appears to me to be a strangely-high number, because “some method of contraception at some point in the past year” seems a pretty low bar to clear, but, again, since the survey group includes anyone who reported having sex even a single time at any point in the past year, maybe not.
Then then calculate the pregnancy rates for this group. Again, this is not a rate for all women of childbearing age, but for all single women who report not actively seeking to get pregnant, and is in response to the question of whether they were pregnant at any point in the past year.And, finally, of the women who said they were pregnant, they calculate the proportion of them for whom the pregnancy ended in abortion. My chart from the prior post backed out of this division with multiplication.
But there is another study which is cited much more often, using Guttmacher data, which comes up with a much higher prevalence of abortion among poor women.
The most recent set of calculations were published in the American Journal of Public Health in 2017 (also available at the NIH website) but use 2014 data (which is a bit later than the NSFG from 2011 – 2013 and maybe there’s a bit of a difference due to changes in just these few years, as the abortion rate has been decreasing, but I don’t know that this helps understand things).
This study used data from three sources: the Guttmacher Institute’s 2014 Abortion Patient Survey (APS), the American Community Survey (ACS), and the NSFG. The APS is based on surveys of abortion clinics, who were themselves asked to survey women getting abortions, by asking women to privately fill out questionnaires. These questions asked about demographic characteristics, union status, and “poverty” (it doesn’t say income level, but the numbers don’t make sense unless they also asked about income level and family size more generally so as to be able to classify people in income buckets). Because the study also needed to know the prevalence of women in each category, they used the ACS and the NSFG for these overall characteristics.
The results for this study are rather different — I’ve added them to the prior table (the third rate is the rate for all women with incomes over 200% FPL).
Also, the total abortion rate for all women of childbearing years was 1.46%.
So why on earth are these two statistics so different?
The statistical breakdowns in the AJPH data explain the motivation the Brookings scholars had in excluding married women, as it is this group which had the lowest abortion prevalence:
- Married, 0.55%
- Cohabitating unmarried 3.1%
- Neither married nor cohabitating, 1.7%
Because women of middle-class and upper-middle-class income levels are far more likely to be married, this could explain to at least some degree why the rates for “percent of single women not trying to get pregnant” vs. “all women” differ. According to the American Community Survey, as analyzed by the Institute for Family Studies, 26% of poor adults ages 18 – 55, 39% of working class adults, and 56% of middle and upper-class adults are married; and 13%/10%/5% respectively are cohabitating. But this adjustment accounts for some of the differences between the data, not all. Are there also differences in how women are classified, with respect to income group, with the APS overclassifying women as “poor” or the NSFG underclassifying them, for example, with women at abortion clinics more likely to exclude income from other family members in answering the question on “family income” or even with different definitions of what “family income” even is?
And none of this explains the broader discrepancy between these two data sources in terms of the overall abortion rates they calculate – 1.46% vs. something less than 1%..
So this is where I leave readers; I have no answer here, which is not particularly helpful, except to document that not every study gives the pat answer you’d like to have, which is itself a reminder to be skeptical of those studies which do claim simple answers.