Abortion Reporting in the United States: An Assessment of Three National Fertility Surveys

Despite its frequency, abortion remains a highly sensitive, stigmatized, and difficult-to-measure behavior. We present estimates of abortion underreporting for three of the most commonly used national fertility surveys in the United States: the National Survey of Family Growth, the National Longitudinal Survey of Youth 1997, and the National Longitudinal Study of Adolescent to Adult Health. Numbers of abortions reported in each survey were compared with external abortion counts obtained from a census of all U.S. abortion providers, with adjustments for comparable respondent ages and periods of each data source. We examined the influence of survey design factors, including survey mode, sampling frame, and length of recall, on abortion underreporting. We used Monte Carlo simulations to estimate potential measurement biases in relationships between abortion and other variables. Underreporting of abortion in the United States compromises the ability to study abortion—and, consequently, almost any pregnancy-related experience—using national fertility surveys. Electronic supplementary material The online version of this article (10.1007/s13524-020-00886-4) contains supplementary material, which is available to authorized users.

Two adjustments must be made to the external counts of abortions to match the abortions that could have been reported by NSFG respondents: an age adjustment and an interview year adjustment. We adjust the external counts of abortions using a series of assumptions about the distributions of respondents' actual age at abortion in each of the five years prior to, and not including, the year of the survey. In the method details below, these are referred to as "age adjustments". In addition, while the NSFG is designed to be a cross-sectional, nationally representative survey, data collection since 2006 is continuous and interviews from a single year are not representative of all women aged 15-44 (in addition, the distribution of women across interview years is unequal). We refer to adjustments made to account for the distribution of interviews across years within a "round" of the survey (e.g. the 2006-2010NSFG and the 2011-2015 as "interview year" adjustments.

Age adjustment
The NSFG respondents are a random sample of women ages 15-44 at time of interview.
Respondents' retrospective reports of abortions-as well as abortions they may have had but did not report-refer to increasingly younger ages as the time between interview and the year of the abortion increases. As a consequence, when we rely on NSFG data, we are missing some abortions for women at older ages. For example, a woman who had an abortion at age 44 in 2003 would be age 47 in 2006, and so would not be included in the NSFG sampling frame. We thus have to adjust the external count of abortions for prior years to remove abortions to women outside of the sampling frame. Failure to omit those abortions from the external count would result in an overestimation of abortion underreporting in the NSFG, affecting both the estimates of underreporting among the oldest women as well as the total among all women. Thus, for each of the five years prior to the interview year, we examine how old women who had abortions in the external counts would be at the time of the survey in order to include only those who match to NSFG respondents.
This correction is slightly more complicated than it first appears because of the interplay in each year and the birthdays of the women having those abortions. Unfortunately, we do not have external data to this level of granularity, so we make several simplifying assumptions. For the most part, NSFG interviews occurred throughout the year (with the exception of the first and last years of survey interviews; more on this below). We assume that birthdates and dates of abortion are also equally distributed throughout each year and throughout each month. 1 Finally, we assume that for abortions that occurred in the same month as a woman's birthday, ½ occurred to women at the younger age (before her birthday) and ½ occurred at the older age (after her birthday). These assumptions allowed us to calculate adjustment factors to apply to the number of abortions to women at each individual age in the external counts.
the NSFG, we sum the adjusted counts to obtain a new, adjusted total number of annual abortions appropriate for matching to the fertility survey.

Interview year adjustment
A final adjustment is needed because the NSFG surveys from 2006 on are designed as a nationally-representative, cross-sectional survey, but all interviews are not conducted in a single year or even a short range of years (e.g.  Because the Add Health sample is restricted to high school students, and the sampling frame is not all U.S. women of a particular age, we did not make assumptions about the exact ages of the external sample (e.g. by accounting for month of abortion and month of birth). However, we did attempt to adjust for educational attainment since some women who did not complete high school would not have been included in the Add Health sample. To estimate external counts of abortions for women with at least 12 years of education, we estimated from the 2000 and 2008 Abortion Patient Survey the proportion of women in each of the age ranges above with a high school education or higher. We then used linear interpolation to estimate these proportions in the intervening years, and applied the relevant proportion to the abortion total for that age range in each year. For birth totals, we use a similar method, estimating the proportion of women in a given age range with more than 12 years of education using birth certificate data from each (year of interview-5) and December (year of interview-1), how many pregnancies did you have that ended in abortion? 13.5 12.9 12.2 11.6 10.9 10.3 9.6 9.6 9.6 9.6 9.6 9.6 9.6 9.6 9.6 9.7 9.8 9.9 10.0 10.