https://orcid.org/0000-0003-0890-9692
Abstract
We examine whether changes in U.S. pregnancy, birth, and abortion rates between 2009 and 2015 reflect underlying change in the incidence of pregnancies classified by retrospectively reported pregnancy desires: pregnancies reported as having occurred at about the right time, later than wanted, too soon, or not wanted at all, and those for which individuals expressed other feelings, including uncertainty, ambivalence, or indifference. We calculate the proportionate distributions of these pregnancies and rates among U.S. women aged 15–44, as well as change over time, overall and among age groups. Characterization of desires for a past pregnancy shifted in a number of ways between 2009 and 2015, and changes across age groups were not uniform. Rates of pregnancies reported as occurring later than wanted increased among older women, while rates of pregnancies reported as occurring too soon decreased among all women. These findings shed light on previous research documenting an increasing age at first birth, increasing rates of pregnancy and childbearing among the oldest age groups, and changes in patterns of contraceptive use, particularly among young women. Our analysis explores limitations and challenges of two major sources of data on pregnancies in the United States and their measures of retrospectively reported pregnancy desires.
Introduction
Pregnancy incidence, when examined according to how individuals felt about a pregnancy they had, provides a snapshot in time of how individuals in a population characterized the pregnancy when asked if they had wanted to become pregnant and whether it occurred at the right time. Measures of past pregnancy experiences should reflect a broad and inclusive range of orientations individuals had prior to their pregnancy. Measuring whether individuals report their pregnancies as having occurred at the right time is as important as tracking indicators of pregnancies that did not or those that an individual did not want to have at all. Further, some individuals who became pregnant may not have had clearly formed plans, intentions, or desires to become pregnant or to avoid it; their experiences are crucial for developing improved understanding of all pregnancy experiences.1 Still, analyses focused on pregnancy incidence according to retrospectively reported desires for those pregnancies do not capture all pregnancy desires that individuals in the population had. A focus on pregnancies, rather than individuals, excludes the desires and experiences of those who did not become pregnant. Research on pregnancy desires among those who became pregnant as well as those who did not is needed to support efforts to advance reproductive autonomy.2
Acknowledging these limitations, our analysis focuses on pregnancy incidence in the United States and on multiple classifications of desires that individuals reported having had toward a pregnancy, moving beyond a focus primarily on unintended pregnancies (see Finer and Henshaw 2006; Finer and Zolna 2011, 2014, 2016). This approach contributes to better understanding of the content and limitations of currently available population-based measures of retrospectively reported pregnancy desires and to more accurate interpretation of their relationship to demographic trends. Specifically, we examine whether recent changes in pregnancy incidence in the United States reflect underlying changes in the incidence of pregnancies among particular groups or all groups of pregnancies classified by pregnancy desire.
According to the most recently published national estimates of pregnancy incidence, 2017 marked the lowest pregnancy rate recorded among U.S. women aged 15–44 since at least 1973, with a steady and continual decline from 2007 to 2017; both birth and abortion rates among women aged 15–44 also reached their lowest level in 2017 (Maddow-Zimet and Kost 2021). In that year, women aged 24 or younger experienced their lowest recorded pregnancy rate since year-to-year comparable estimates have been available (Maddow-Zimet and Kost 2021). In contrast, pregnancy rates among those aged 30 or older continued to increase between 2007 and 2017, a trend that has persisted over at least four decades.3
Rising pregnancy rates among older women may reflect increases in rates of births reported as having occurred at about the right time among these women—perhaps because of successful postponement of childbearing—rather than, say, increasing rates of births or abortions resulting from pregnancies not wanted at the time they occurred or ever. Similarly, with the steep declines in pregnancy rates among women aged 24 or younger, we might expect a concomitant increase in the proportion of their pregnancies ending in births reported as having occurred at about the right time if, overall, women in these age groups have become more successful at avoiding pregnancy unless they want to become pregnant.
In this study, we examine pregnancies, births, and abortions across four time points within this most recently documented period of decline in pregnancy rates in the United States—2009, 2011, 2013, and 2015—according to five categorizations of pregnancy desire, as well as change over this period overall and within age groups. In response to a large body of research critiquing the measurement of pregnancy intentions (see Kost and Zolna 2019), we break with long-standing and widely used nomenclature typically ascribed to respondents' survey responses—“intended,” “unintended,” and “mistimed”—and attempt to more accurately reflect how women in the United States reported their feelings toward pregnancies they experienced. To accomplish this, we reexamine data sources and use terminology more closely aligned with the wording of the survey questions and the responses individuals gave, an approach we have used in similar studies of pregnancy desires (Kost et al. 2018; Kost et al. 2021; Maddow-Zimet and Kost 2020). We also break with approaches used in prior studies of this kind (Finer and Zolna 2014, 2016) by examining each pregnancy desire category separately, rather than by combined groups (e.g., those typically referred to as “intended”—pregnancies that occurred at about the right time or later than wanted; those referred to as “unintended”—pregnancies that occurred too soon or were not wanted then or ever). Finally, our approach also differs in that we explicitly highlight constraints of the data sources available and create pregnancy desire measures that include a separate “other” group with responses characterized by uncertainty, ambivalence, indifference, and any response that did not express a preference for the pregnancy's timing.
Data Sources and Methods
Pregnancies and Populations
There is no single data source with individual records and a complete count of pregnancies occurring annually in the United States; therefore, we calculated pregnancy incidence by summing the incidence of births, abortions, and fetal losses from various sources. The annual incidence of births to all women and by their age at delivery was obtained from vital statistics reports from the National Center for Health Statistics (NCHS), based on registered U.S. birth certificates (U.S. Department of Health and Human Services 2020). Data on abortion incidence were obtained from the Guttmacher Institute, which periodically conducts a national census of all known abortion providers in the United States to track the numbers of procedural and medication abortions performed annually (Jones et al. 2022). The numbers of abortions by age are from Maddow-Zimet and Kost (2021), who combined information from the Centers for Disease Control and Prevention, the Guttmacher Institute's national census of abortion providers, and state health departments. There are no population-based sources with complete data on the incidence of fetal loss for the United States; we estimated its incidence overall as the sum of a fixed proportion of births (20%) and of abortions (10%), for the population as a whole and by age. This approach was originally suggested by Leridon (1977); a detailed rationale is also provided in Maddow-Zimet and Kost (2021).
For each year, we estimated the total number of pregnancies, Ω, for a particular desire category, d, as a function of the number of births and abortions in that desire category:
where B is the number of births, A is the number of abortions, θd represents the proportion of births in desire category d, and αd represents the proportion of abortions in desire category d. To compare estimates from one year to another, we calculated rates expressed as the incidence of an event occurring per 1,000 individuals. Population counts of women aged 15–44 and by age for each year in our analysis were obtained from the U.S. Census Bureau (2021).
Pregnancy Desires
To obtain national counts of pregnancies ending in birth and in abortion according to each pregnancy desire category, we obtained proportionate distributions of pregnancy desires from nationally representative survey data: the NCHS's National Survey of Family Growth (NSFG) and the Guttmacher Institute's Abortion Patient Surveys (APS). Distributions from the NSFG were then applied to annual counts of pregnancies ending in birth, and distributions from the APS were applied to annual counts of abortions.
NSFG and APS respondents were asked similar questions to recall how they had felt about having a baby right before they became pregnant and how they felt about the timing of the pregnancy (see the online appendix for question wording).4 Both surveys were designed to construct a pregnancy desire measure that combines wantedness and timing dimensions (Campbell and Mosher 2000; Santelli et al. 2003). As described later, these questions are oriented toward a measure of pregnancy desires that primarily distinguishes groups along the timing dimension. Thus, our approach also prioritizes the respondents' assessment of the pregnancy's timing. Our measure of pregnancy desires includes five categories: pregnancies reported as having occurred at about the right time, having occurred later than wanted, or having occurred too soon, those reported as not wanted at the time they occurred or ever, and those in an “other” category described below (programming code for our measure is available here: https://osf.io/h3zeu/).
Pregnancy Desires Associated With Births
In face-to-face interviews, female respondents in the NSFG are asked for a complete pregnancy history, including the date of each birth and information on the pregnancy's wantedness and timing (NCHS 2011, 2014, 2016, 2018, 2020). The questions on pregnancy wantedness do not explicitly offer answer options reflecting uncertainty, ambivalence, indifference, or anything other than whether a baby was wanted or not, although some respondents voluntarily provided such responses.
Our measure of pregnancy desires prioritizes the respondents' reported feelings about timing. We categorized all births according to the timing preference, if these data were available: occurred at about the right time, occurred later than wanted, occurred too soon, or other.
NSFG respondents who said they had not wanted a baby at that time or ever when first asked about pregnancy wantedness are not asked about the pregnancy's timing. Those who answered the wantedness question with uncertainty or any other volunteered response were directed to follow-up questions that encouraged a “probably yes” or “probably no” response to wantedness; those who responded “probably yes” were directed to the timing question, as were those who initially responded “yes” when asked if they had wanted to have a baby just before the pregnancy occurred. Those who responded to the follow-up question with “probably no” or continued expressions of uncertainty, ambivalence, or any other response are included in our “other” pregnancies category because they were skipped past the timing question.5
In our measure, the “other” category includes responses across the two dimensions—wantedness and timing. This category includes both respondents who were asked about timing and reported “other” (e.g., “don't know” or “don't care”) and those who reported anything other than “yes” or “no” desires for whether their pregnancy was wanted and were not subsequently asked about the timing.
Finally, we do not assume that all pregnancies reported as having occurred “at about the right time,” “later than wanted,” or “too soon” represent “wanted” births (i.e., those for which the respondent answered “yes” when asked if she had wanted to have a baby just before she became pregnant). For example, a pregnancy classified as having occurred “at about the right time” may be one for which the respondent first reported that she wasn't sure if she had wanted a baby, but then reported that she probably did in the follow-up question and that the timing was about right.6
To ensure robust and representative estimates, we followed NCHS guidance and analyzed data from only those births reported to have occurred within three calendar years prior to the first calendar year of interview for each survey period and weighted the data so the estimates represent births to women during the midpoint year of that survey period (NCHS 2020).7 For example, we analyzed all births that occurred in 2014–2016 among all respondents interviewed in the most recent 2017–2019 survey period and applied the full sample's population weights to estimate births to the population of women aged 15–44 in July 2018.8
We then assumed that estimates based on a three-year period approximated an annual estimate corresponding to the midyear. Continuing from the foregoing example, we applied the weighted percentage distribution of all births that occurred in 2014–2016 by pregnancy desires from the NSFG to the counts of births in 2015 from vital statistics to obtain the number of births that occurred in 2015 in each pregnancy desire group. We followed the same approach to calculate the total number of births in each pregnancy desire group for 2009, 2011, and 2013.
Pregnancy Desires Associated With Abortions
Unlike births, most abortions are not reported in the NSFG and the distribution of missing abortions is not random (Desai et al. 2021; Lindberg, Kost et al. 2020; Lindberg and Scott 2018; NCHS 2020). We obtained the distribution of pregnancy desires associated with abortions from the Guttmacher Institute's APS, which periodically collects information from a representative sample of abortion patients from all known U.S. clinics (Jerman et al. 2016; Jones et al. 2010). The two most recent surveys were conducted in 2008 (n = 9,493) and 2014 (n = 4,181).
In contrast to the interviewer-led NSFG, the APS is a self-administered survey and offers “not sure, don't know” and “didn't care” as answer options to the question on pregnancy wantedness. All respondents—except those who chose “no” when asked whether they had wanted to become pregnant—are then directed to the timing question and can choose among four answer options: “too soon,” “at the right time,” “later than I wanted,” or “didn't care.” As in our approach with the NSFG data, we prioritized responses to the pregnancy timing question to create five categories of pregnancy desire, including the “other” category. The latter category includes respondents who answered the timing question and selected “didn't care” and those who selected “yes,” “don't know, not sure,” or “didn't care” for whether their pregnancy was wanted, then did not answer the timing question.
We used linear interpolation of the weighted distributions of pregnancy desires in the 2008 APS and the 2014 APS to obtain proportions of abortions by pregnancy desire for 2009, 2011, and 2013. For 2015 estimates, we used weighted distributions in 2014, the latest year for which data were available.
Analytic Approach and Statistical Significance
Approach
We examined both percentage distributions and rates of pregnancies by pregnancy desire. Percentage distributions describe the composition of all pregnancies in terms of the relative number of pregnancies in each desire category. Changes in proportions over time can reflect underlying shifts in the relative prevalence of pregnancies characterized by the categories. For example, a trend in increasing proportions of pregnancies reported as having occurred at the right time might indicate that pregnancies are increasingly occurring among those who had previously been successful at avoiding a pregnancy they did not want. In contrast, rates provide information on the incidence of pregnancy among the population of women of reproductive age, and changes over time reflect shifts in the absolute levels of pregnancy incidence. Thus, proportions and rates allow for different insights into change over time.
We present overall distributions (Table 1) and rates (Table 3) of all U.S. pregnancies, births, and abortions across pregnancy desire categories for 2009, 2011, 2013, and 2015, as well as distributions (Table 2) and rates (Table 4) by age groups for 2009 and 2015. In the Results section, we report on patterns in the distributions and level of estimates across desire categories and then examine evidence of significant change over time. The period covered in our analysis, 2009–2015, corresponds to the most recent data available on births and abortions by pregnancy desire.
Four of the five pregnancy desire groups appear to parallel what have traditionally been labeled “intended” (occurred at about the right time or later than wanted) and “unintended” (occurred too soon or had not been wanted). However, we emphasize that our estimates are not comparable to prior published ones because our analysis relies on a different coding scheme for the pregnancy desire measure; further, our approach to selection of the analytic sample of births from the NSFG data and estimation of fetal loss also differs.
The fifth pregnancy desire group in our analysis—“other”—includes responses of uncertainty, indifference, and refusal across the survey questions. Because the opportunities to offer these responses differed in the NSFG and APS questionnaires, we were not able to code data into an “other” pregnancy desire category that was precisely comparable across the surveys. We therefore omit estimates for the “other” category when showing estimates among pregnancies, since these combine data from the two surveys. The “other” category is included for estimates for births and abortions shown separately.
Finally, we note that variation in the likelihood of experiencing fetal loss by age is well documented (Andersen et al. 2000; Magnus et al. 2019), but little is known about how it varies over time as the proportion of pregnancies that end in birth and abortion vary; consequently, we do not include external estimates of variation in fetal loss by age in our analyses. We use a constant proportion of births and abortions to calculate fetal losses for each age group, so our estimation method produces variation that depends only on the ratio of abortions to births in each age group. Our findings should be interpreted as primarily reflecting population-level patterns in the incidence of births and abortions and should not be used to draw inferences about fetal loss.
Tests of Statistical Significance
Prior studies of national estimates of pregnancy intentions have not included tests of statistical significance because estimates are obtained using multiple data sources, each containing its own sampling error. However, we sought to capture uncertainty around our estimates as well as the probability of observed changes over time. To accomplish this, we employed a simulation that independently accounts for two known sources of sampling error in the pregnancy rates—the distribution of births by pregnancy desire in the NSFG (θd for a particular desire, d ) and the distribution of abortions by pregnancy desire in the APS (αd for a particular desire, d ). This approach enabled us to calculate confidence intervals as well as the probability that any difference observed between a proportion (or a rate) for two different years was not the result of random chance.
For each year and desire category (and age group for age-specific analyses), we generated distributions of θ and α separately using the svy: mlogit command in Stata with sampling weights and strata specified. We then drew 10,000 samples for θ and α based on the calculated distribution. For each sample indexed, s, we estimated the total number of pregnancies for a particular desire category as a function of the number of abortions and births in that desire category: Ωd,s = 1.2(Βθd,s) +1.1(Ααd,s). The 95% confidence intervals presented are derived from the 2.5 and 97.5 percentiles of the distribution of the 10,000 samples indexed: 95% CI = [(Ωd, .025), (Ωd, .975)].
To obtain the probability of change over time, we calculated the mean value of the difference in indexed samples between year: Ωd,µΔ = Σ(Ωd,s[Δ]) / 10,000. The proportion of samples for which the difference over time was in the same direction as Ωd,µΔ represents the probability that the difference between our point estimates is not zero and is in the direction we observed. We highlight differences between two estimates when the probability of change was 95% or greater, indicating that 5% or less of the time, no difference in point estimates was observed, or the difference was in the opposite direction of what we observed; probabilities lower than 95% suggest that differences in estimates in the direction observed may be a result of at least a 5% random chance.
We used Stata version 16.1 throughout the analysis (StataCorp 2019). For analyses of the NSFG and APS data conducted separately, we used the svy command to account for each survey's sampling design.
Results
Distributions of Pregnancies, Births, and Abortions by Pregnancy Desire
Pregnancies
In all four years, the largest proportion of pregnancies were reported to have occurred at about the right time, ranging from 43.7% to 47.9% in 2009–2015 (Table 1). The next largest group across all years were pregnancies that occurred too soon (22.6–27.5%), followed by pregnancies that had not been wanted (17.1–18.0%) and pregnancies reported as having occurred later than wanted (6.6–10.9%).
Births
The distribution of births by desire group in 2009–2015 followed patterns similar to those of total pregnancies, but a larger proportion of births occurred at the right time (54.0–57.6%). The proportion of births from pregnancies we characterized as “other” ranged from 0.7% to 2.4%.
Abortions
In contrast to pregnancies and births, the largest proportion of abortions were pregnancies that occurred too soon (47.6–49.1%), followed by those that had not been wanted (41.5–42.7%); smaller proportions were “other” pregnancies (5.5–7.1%), those that occurred at about the right time (1.5–2.4%), and those that occurred later than wanted (1.1–1.5%).
Change in Percentage Distributions of Pregnancies, Births, and Abortions by Pregnancy Desire Since 2009
Pregnancies
There was no significant change over time in the proportion of pregnancies that occurred at the right time, nor in those that had not been wanted (Table 1). However, the proportion of pregnancies that occurred later than wanted increased 59% from 6.9% in 2009 to 10.9% in 2011, and increased 37% between 2009 and 2015 (to 9.4%). In contrast, the proportion that occurred too soon declined 12% from 27.5% in 2009 to 24.2% in 2011, and declined 18% between 2009 and 2015 (to 22.6%).
Births
Changes over time in the distributions of pregnancy desires among births mirror those among total pregnancies. The proportion of births from pregnancies that occurred later than wanted increased 61% from 8.3% in 2009 to 13.3% in 2011, and rose by 35% between 2009 and 2015 (to 11.1%). The proportion of births that occurred too soon fell 17% from 2009 (22.0%) to 2011 (18.3%), and 21% from 2009 to 2015 (to 17.5%). The proportion of births with “other” pregnancy desires was small, although it increased from 1.0% in 2009 to 2.4% in 2011. There were no significant changes over time in the proportion of births that occurred at the right time or were not wanted at all.
Abortions
In contrast to the increases in proportions of pregnancies and births occurring later than wanted and decreases in the proportions that occurred too soon, the proportion of abortions in both desire categories did not change significantly across the four points in time. And unlike the patterns in all pregnancies and births, the proportion of abortions characterized as pregnancies that occurred at about the right time decreased 39% from 2.4% in 2009 to 1.5% in 2015. The proportion of abortions from pregnancies characterized by “other” desires increased 23% between 2009 and 2013 (from 5.5% to 6.8%) and 29% between 2009 and 2015 (from 5.5% to 7.1%).
Percentage Distributions of Pregnancies, Births, and Abortions by Pregnancy Desire and Age, and Changes Between 2009 and 2015
In contrast to the lack of significant change among all women aged 15–44, women aged 35–44 experienced a 26% decline over the period 2009–2015 in the proportion of their pregnancies characterized as having occurred at the right time (Table 2). This age group also experienced a 39% decline in their proportion of pregnancies that occurred too soon, and an 84% increase in the proportion that occurred later than wanted. No other age group experienced significant changes in the proportion of pregnancies in these three desire categories, except women aged 15–19, who experienced a 95% decrease in the proportion of pregnancies that occurred later than wanted. There was no evidence of change between 2009 and 2015 in the proportion of pregnancies that had not been wanted for any of the age groups.
Changes in the proportionate distribution of births within each desire category for age groups mostly mirrored the changes for pregnancies among age groups, although the decrease in the proportion of births that occurred too soon among 35–44-year-olds was not significant. In 2009, women aged 20–24 characterized 71% fewer births with “other” pregnancy desires than they did in 2015, while the proportion of these births to 35–44-year-olds increased more than 10-fold.
Among pregnancies ending in abortion, there were 54% fewer in 2015 than in 2009 reported by 20–24-year-olds as having been pregnancies that occurred at the right time. There were no other significant changes in the proportion of pregnancies across the other pregnancy desire groups for any of the other age groups, except those abortions reported with “other” pregnancy desires. Three age groups—15–19-year-olds, 20–24-year-olds, and 30–34-year-olds—all had higher proportions of their pregnancies ending in abortion characterized by “other” pregnancy desires in 2015 than they did in 2009.
Rates of Pregnancies, Births, and Abortions by Pregnancy Desire
Pregnancies
Across all four time points, pregnancies with the highest rate among U.S. women were those that occurred at about the right time, ranging from 41.3 to 45.5 per 1,000 women aged 15–44 (Table 3). Pregnancies with the lowest incidence were those that occurred later than wanted, with rates ranging from 6.1 to 10.3. Falling between these two pregnancy desire groups were pregnancies that occurred too soon, with rates ranging from 20.6 to 27.5 per 1,000, and pregnancies not wanted at all, ranging from 15.5 to 18.0.
Births
Like patterns in pregnancy rates, the highest rate of births were those that occurred at the right time, with rates ranging from 34.1 to 37.5 per 1,000 women aged 15–44 across the four time points. Rates of births resulting from pregnancies that occurred too soon were the next highest (from 11.0 to 14.6). Rates of births from pregnancies that occurred later than wanted (4.9 to 8.4) and that were unwanted (7.3 to 8.0) were lower. Births from “other” pregnancies had the lowest rates, ranging from 0.4 to 1.5.
Abortions
Rates of abortion were highest for pregnancies that occurred too soon, ranging from 7.3 to 9.1 per 1,000 women aged 15–44 across the four points in time. Somewhat lower abortion rates were for pregnancies that occurred when individuals did not want to have a baby (6.1 to 7.7).9 The lowest abortion rates were for “other” pregnancies (1.0 per 1,000 women aged 15–44 across all years), for pregnancies that occurred at about the right time (0.21 to 0.45), and for those that occurred later than wanted (0.16 to 0.27).
Change in Rates of Pregnancies, Births, and Abortions by Pregnancy Desire Since 2009
Pregnancies
Between 2009 and 2011, the rate of pregnancies that occurred at the right time dropped 9% (from 45.5 to 41.3 pregnancies per 1,000 women aged 15–44; Table 3). The rate of pregnancies that occurred too soon also decreased, falling 17% between 2009 and 2011 (from 27.5 to 22.9). During this same time, the rate of pregnancies that occurred later than wanted increased 50%, from 6.9 to 10.3.
From 2009 to 2015, the rate of pregnancies that occurred too soon decreased 25% (from 27.5 to 20.6 per 1,000 women aged 15–44). The rate of pregnancies that occurred when a baby was not wanted decreased 14% between 2009 (18.0) and 2015 (15.5).
Births
Changes in birth rates between 2009 and 2011 mostly mirror changes in the pregnancy rates. The birth rate for pregnancies that occurred at about the right time decreased by 9% from 2009 to 2011 (from 37.5 to 34.1 per 1,000 women aged 15–44), the rate for those that occurred too soon decreased by 21% (from 14.6 to 11.6), and the rate for pregnancies that occurred later than wanted increased 54% (from 5.5 to 8.4). Birth rates for “other” pregnancies more than doubled from 0.7 in 2009 to 1.5 in 2011.
Like the change observed between 2009 and 2015 among pregnancies, births from pregnancies occurring too soon decreased 25%, falling from a rate of 14.6 to 11.0 per 1,000 women aged 15–44. Unlike pregnancy rates, birth rates for pregnancies that were not wanted did not change significantly across the four time points.
Abortions
Decreases in abortion rates were evident in every pregnancy desire category, except for the “other” category, where the rate stayed constant at 1.0 per 1,000 women aged 15–44. Even the very small rate of abortions for pregnancies that occurred at the right time decreased 44% between 2009 and 2013 (from 0.45 to 0.25), and 54% between 2009 and 2015 (from 0.45 to 0.21). Large percentage decreases in the already low rates of abortions resulting from pregnancies that occurred later than wanted were also evident, with the rate falling 35% between 2009 (0.27) and 2013 (0.18), and 43% between 2009 and 2015 (to 0.16).
Abortion rates for pregnancies that occurred too soon or those for which a baby was not wanted were significantly lower in 2011, 2013, and 2015 relative to 2009. The rate for pregnancies that occurred too soon fell 10% between 2009 (9.1) and 2011 (8.2), 20% between 2009 and 2013 (to 7.3), and 26% between 2009 and 2015 (to 6.7). A similar pattern was found for the abortion rate of pregnancies for which the individual had not wanted a baby: the rate fell 7% between 2009 (7.7) and 2011 (7.1), 16% between 2009 and 2013 (to 6.5), and 21% between 2009 and 2015 (to 6.1).
Pregnancy, Birth, and Abortion Rates by Pregnancy Desire and Age, and Changes Between 2009 and 2015
Pregnancy Occurred at About the Right Time
In 2015, rates of pregnancies that occurred at about the right time ranged from 9.2 to 82.9 per 1,000 women aged 15–44, depending on the age group (Table 4). Rates were higher than for all other pregnancy desire categories only for women in age groups 25–29 or older; among the two youngest age groups, rates of pregnancies that occurred at the right time in 2015 were lower than those that occurred too soon.
Between 2009 and 2015, only women aged 20–24 experienced significant change in the rate of pregnancy that occurred at the right time, declining 20% over this period. Similarly, the birth rate of pregnancies that occurred at the right time for 20–24-year-olds declined 19% from 2009 to 2015. And the abortion rate for pregnancies that occurred at the right time among 20–24-year-olds declined by 68%, although rates were very low (0.8 in 2009 and 0.3 in 2015). Individuals in the 15–19 age group also experienced significant declines in the very small abortion rate for pregnancies that occurred at about the right time, falling 82% between 2009 and 2015.
Pregnancy Occurred Later Than Wanted
In 2015, the rate of pregnancies that occurred later than wanted ranged from 5.6 to 13.7 per 1,000 women, depending on the age group. Only women in the oldest group experienced a significant change in the rate of these pregnancies between 2009 and 2015, with the rate in 2015 twice as high as it was in 2009. The birth rate of pregnancies that occurred later than wanted among women aged 35–44 also doubled between 2009 and 2015, although there was no change in their abortion rate associated with these pregnancies. The abortion rate for pregnancies that occurred later than wanted decreased only among 25–29-year-olds, falling 56% between 2009 and 2015.
Pregnancy Occurred Too Soon
In 2015, rates of pregnancies that occurred too soon ranged from 2.5 to 47.1 per 1,000 women, depending on the age group. For 15–19-year-olds, rates of pregnancies, births, and abortions that occurred too soon were higher than rates for any other pregnancy desire category, and these rates each fell 46–48% between 2009 and 2015. Women aged 20–24 also experienced declines in pregnancies that occurred too soon between 2009 and 2015, with rates falling 23%. No other age groups experienced significant declines over this time period in pregnancy or birth rates of pregnancies in this desire category. However, abortion rates for pregnancies that occurred too soon decreased significantly between 2009 and 2015 for all age groups.
Did Not Want to Have a Baby
In 2015, rates of pregnancies that occurred to individuals who did not want to have a baby ranged from 8.1 to 22.7 per 1,000 women. These pregnancy rates fell significantly between 2009 and 2015 for the two youngest age groups only: among 15–19-year-olds, the rate declined by 46%, and among 20–24-year-olds, it dropped by 29%. Like the abortion rates for pregnancies that occurred too soon, the abortion rates of pregnancies occurring when a baby was not wanted decreased significantly for nearly every age group between 2009 and 2015, except for the oldest. And as for pregnancies that occurred too soon, these decreases in abortion rates occurred alongside either a decrease or no significant change in the pregnancy or birth rates.
Other Pregnancy Desires
In 2015, few births were characterized by “other” pregnancy desires, with rates of these births ranging from 0.4 to 2.7 per 1,000 women, depending on the age group. The rate decreased significantly among 20–24-year-olds between 2009 and 2015, falling 77%; in contrast, the rate increased more than 10-fold among 35–44-year-old women.
Rates of “other” pregnancies ending in abortion in 2015 ranged from 0.5 to 1.6 per 1,000. There was no evidence of significant change between 2009 and 2015 in the abortion rate of pregnancies in the “other” pregnancy desire category for any of the age groups.
Discussion
Implications of Findings
The increasing age at first birth in the United States (Guzzo and Payne 2018; Mathews and Hamilton 2016), as well as increasing rates of pregnancy and childbearing among those in the older age groups (Maddow-Zimet and Kost 2021), might suggest an increase in desire for pregnancy at older ages. While we did not find evidence of significant change between 2009 and 2015 in the proportion of pregnancies characterized as having occurred at about the right time among all women aged 15–44, we did find a significant decrease over the same period in the proportion of these pregnancies to women in the oldest age group (35–44-year-olds). This group also experienced a significant decline in the proportion of their pregnancies that occurred too soon, as well as a significant increase in the proportion reported as having occurred later than wanted. In other words, women may have avoided pregnancies they did not want until older ages, but older women were not increasingly reporting those pregnancies as having occurred at “about the right time.”
Previous analyses combined pregnancies reported as having occurred at about the right time with those that occurred later than wanted (“intended” pregnancies), with the latter often assumed to reflect pregnancies that individuals wanted at the time they occurred but simply took longer to conceive (Santelli et al. 2003). However, by examining these two desire groups separately, we found differing trends over time, with a significant overall increase in the proportion of pregnancies that individuals said occurred later than they had wanted. The increase between 2009 and 2011 may have been related to the economic recession (December 2007–June 2009), such that some of these pregnancies may have been postponed. Indeed, other research has sought to explain how economic downturns and financial stressors are likely to impact fertility trends (Brauner-Otto and Geist 2018; Seltzer 2019).
Still, the proportion of all pregnancies characterized as having occurred later than wanted continued to rise after 2011, increasing 37% between 2009 and 2015. Aside from recession factors, this could reflect patterns of increasing age at first birth and delayed childbearing in the U.S. population; some individuals seeking to become pregnant at increasingly older ages may find they are waiting longer to become pregnant than they had wanted. This increase could also reflect an opportunity to think back over an increasingly longer time period and report a desire for a birth to have occurred sooner than it did, when the individual was younger or when their life circumstances had been different (Gemmill 2019; Hayford 2009; Nitsche and Hayford 2020). In fact, we found a significant increase in the rates of pregnancies and of births characterized as having occurred later than wanted only among the oldest age group. These findings also suggest there could be substantial, and growing, unmet need for fertility treatment services.
We found evidence of significant decreases in the proportions and rates of all pregnancies and births reported as having occurred too soon between 2009 and 2015—trends that may have been influenced by documented increases in use of the most effective methods of contraception (Kavanaugh and Jerman 2018; Kavanaugh and Pliskin 2020) and facilitated by implementation of the contraceptive mandate of the Affordable Care Act signed into law in August 2012, requiring insurance plans to cover contraception without copays and comply by January 2013 (Becker and Polsky 2015). Such improvements in access to contraception likely mean that more pregnancies were successfully prevented among those who were specifically looking to postpone childbearing, whether for their first birth or subsequent ones.
The overall rate of pregnancies that individuals did not want to have at all decreased 14% between 2009 and 2015. A rising age at which women are beginning childbearing could mean that fewer women reach the point at which they have had the number of births they want and no longer want to become pregnant. However, rates of pregnancies that occurred when a baby was not wanted did not change between 2009 and 2015 among women in the oldest age groups (25–29 and older). Instead, the overall decline in these unwanted pregnancies appears to have been driven mostly by significant decreases in their incidence among women aged 15–19 and 20–24.
A large body of research has shed light on the wide range of attitudes toward pregnancy that individuals have, including ambivalent feelings, uncertainty, indifference, and fatalism (Gómez et al. 2019; Hernandez et al. 2020; Higgins et al. 2012; Miller et al. 2013; Schwarz et al. 2007; Yoo et al. 2014). We found that the proportions of births and abortions for which individuals expressed feelings about the pregnancy other than those specifically characterized by certainty in their wantedness or timing preferences increased between 2009 and 2015; regardless of the pregnancy outcome, an increasing proportion of individuals who became pregnant over this period characterized their feelings toward the pregnancy in ways that did not fit clear preferences to have or to avoid a pregnancy.
We found that age groups differed widely in the relative size of pregnancy rates for each desire category, and changes in these rates over time also differed across age groups. Among the youngest women of reproductive age (those aged 15–19 and 20–24), we found significant decreases in the rates of pregnancies reported as having occurred too soon. And while the decrease during this time among 25–29-year-olds did not meet our criteria for statistical significance, the combined decrease across these three age groups, together with the high proportion of all pregnancies that occur at these ages (Maddow-Zimet and Kost 2021), means that much of the decrease in overall pregnancy incidence from 2009 to 2015 is attributable to fewer pregnancies that occurred too soon.
In addition, decreases in rates of pregnancies that occurred too soon were proportionally larger for the youngest women—the 15–19 age group—dropping 47% between 2009 and 2015, resulting in a growing gap between the pregnancy rates across the three youngest age groups. To put these estimates of incidence in context, in 2009, the rates of pregnancies that occurred too soon among 15–19-year-old and 25–29-year-old women were similar at 34 and 39 per 1,000 women, respectively. In 2015, these rates were 18 and 31 per 1,000 women.
Kost and colleagues (2021) used state-level data for 2012–2017 to examine pregnancies and—because of data constraints—combined pregnancy desires into three categories: those that occurred too soon and those that had been unwanted as one group, those that occurred at about the right time and those that occurred later than wanted as another group, and a third group consisting of uncertain desires. They found wide geographic variation in rates for all three groups and no consistent trends—whether increasing or decreasing—in rates of pregnancies that had occurred too soon or had not been wanted (combined) or in those with uncertain pregnancy desires across the states. In this current analysis, however, we did find significant declines between 2009 and 2015 nationally, both in the rate of pregnancies that occurred too soon and in the rate of those that had not been wanted. And, while the state-level analysis found that the majority of states showed declining trends in the rate of pregnancies that had occurred at the right time or later than wanted between 2012 and 2017, this national-level analysis, which considered these groups separately, revealed that the two groups had opposing patterns of change from 2009 to 2015, underscoring the importance of distinguishing the two groups.
We found evidence that underlying the decline in the abortion rate between 2009 and 2015 (Jones et al. 2019; Maddow-Zimet and Kost 2021) were significant decreases in the national abortion rates across all pregnancy desire groups. Most pregnancies ending in abortion were those that occurred too soon or had not been wanted at all, together comprising 90–91% of all abortions in 2009 and 2015. We also found that the decline in the overall abortion rate was driven primarily by significant declines in the rates of abortion from pregnancies that occurred too soon or were not wanted, with nearly every age group experiencing significantly lower rates of both in 2015 than in 2009. In total, the abortion rate for pregnancies that occurred too soon declined 26% over this period, and the rate for pregnancies not wanted at all decreased by 21%. These lower abortion rates in 2015, taken alongside significantly lower birth rates for pregnancies reported as having occurred too soon and no evidence of increased birth rates over this time period for any pregnancy desire group, indicate that far fewer individuals were becoming pregnant in 2015 than in 2009, and that abortion incidence went down because individuals did not get pregnant, not because their pregnancies continued to a birth instead of an abortion.
Over the time period covered by this analysis, others have noted significant changes in patterns of contraceptive use that are consistent with the age group differences we find in pregnancy desires as well as declining pregnancy rates overall. Kavanaugh and Jerman (2018) examined national-level estimates of contraceptive use in 2008, 2012, and 2014 and found little change in the overall level of use among the entire U.S. population of women aged 15–44; use remained consistently high in all years (60% of all women, and 90% of those at risk of unintended pregnancy). However, they also found shifts in the composition of methods women were using, with increased use of highly effective reversible methods, especially among women aged 18–24 (Kavanaugh and Jerman 2018). These patterns likely underlie the decreases in pregnancy, birth, and abortion rates for those pregnancies that occurred too soon or had not been wanted at all.
It remains to be seen how the COVID-19 pandemic, related economic turmoil, and the new landscape of statewide abortion restrictions and bans will impact pregnancy rates and individuals' feelings about the pregnancies. However, initial evidence of substantial impacts on how individuals are thinking about their future fertility—and their reports of behavioral changes in relation to the pandemic (Lindberg, VandeVusse et al. 2020) and the obstacles to obtaining sexual and reproductive health care it has created (Kavanaugh et al. 2022)—indicates we may see further changes when data on pregnancy incidence and desires become available for the pandemic and postpandemic years.
A Need for Improved Measurement
This work provides new estimates of pregnancy desires at the national level and improves upon previously published estimates in several ways, including more accurate representation of survey item responses and retention of separate, rather than combined, pregnancy desire categories. Still, the available data for these national-level estimates continue to present significant limitations. Specifically, the survey questions fall short of capturing the complexity of pregnancy desires. Neither the NSFG nor the APS was designed to fully capture uncertainty, indifference, or ambivalence in either pregnancy wantedness or timing; our estimates would likely differ if they had been. Further, some respondents may have tailored their responses to fit the framing of the questions and response options, which would impact the reliability of the measure, the precision of our estimates, and interpretation of findings.
The “other” category among births in the NSFG is not likely sufficient as a proxy for the true proportion of births from pregnancies characterized by uncertainty, indifference, or ambivalence because these answer options were not explicitly offered in the face-to-face interviews. In fact, analyses of other surveys that presented uncertainty as an option to respondents found that as many as 20% of births were characterized this way (California Maternal, Child and Adolescent Health Program 2013; Hollingsworth 2010; Maddow-Zimet and Kost 2020). In our analysis, only about 1–2% of births from 2009 to 2015 resulted from pregnancies characterized as “other.” In the self-administered APS, we found higher proportions in our “other” category, ranging from about 6% to 7% of abortions across the four years. However, this proportion could also be an underestimate, given that the APS questionnaire did not offer answer options for expressing uncertainty or ambivalence about pregnancy timing, only for indifference.
Pregnancy desire questions in both surveys constrain individuals' reports of their feelings toward their past pregnancies. For example, in the NSFG, individuals who cannot report that they probably did or probably did not want to become pregnant are skipped past the question asking how they feel about the pregnancy's timing, although they may be able to answer this very different question. And, again, in the APS, individuals are not offered answer options reflecting uncertainty or ambivalence in relation to the pregnancy's timing.
Further, in both surveys, individuals who recalled that they had not wanted to have a baby at that time or ever are not asked how they feel about the pregnancy's timing. Given evidence that some individuals who had not wanted a baby later report happiness when they learn they are pregnant (Aiken et al. 2015; Blake et al. 2007; Hartnett 2012; Sable and Libbus 2000), all respondents, regardless of whether they report a pregnancy had been wanted or not, should be directed to the timing assessment question. Pregnancy desires are complex, and a fuller understanding of all respondents' timing assessments would allow for improved measurement of the experience and impact of these events on individuals' lives.
Finally, we found it difficult to describe the time frame individuals are referring to when they report feelings about the timing of their pregnancies in the NSFG and APS. In both surveys, the timing question seems to ask how they feel about the timing, not how they had felt: “Would you say you became pregnant too soon, at about the right time, or later than you wanted?”10 Without specifying a time frame, respondents may answer in relation to their current feelings and life circumstances, or they may interpret the question as asking about their feelings at the time of, or prior to, their pregnancy.
These issues emphasize the need for improvement in survey-based measures of pregnancy desires. Introducing a wider range of response options in questions on pregnancy wantedness and timing will shift our estimates of pregnancies by desire category, although we cannot anticipate the directions of change for each category. In a recent study of how the introduction of a “not sure” answer option impacted distributions of other answers offered, Maddow-Zimet and Kost (2020) found significant evidence that the effects were neither random nor predictable.
This study underscores the need for further research on how individuals characterize their feelings toward a pregnancy they had and for new quantitative, population-level measures that address shortcomings of measures the field has relied on for decades (Kost and Zolna 2019). Better measures of recalled pregnancy desires would capture a wider range of pregnancy experiences, improve researchers' ability to identify factors that influence these experiences, and more accurately identify pregnancies individuals did or did not want to have, either at the time they occurred or ever. Rigorously tracking the rate at which these pregnancies occur, and the proportion of pregnancies characterized by differing pregnancy desires, is important because these indicators can signal public health needs for contraceptive, abortion, and fertility treatment services, as well as for prenatal and maternal care. Indeed, retrospective measures of pregnancy desire are informed only by those who became pregnant; missing from this research is the experience of those who were trying to become pregnant but were not able to, as well as those who had been able to avoid a pregnancy they did not want to have. Still other measures should be developed to capture the pregnancy desires of all individuals, whether or not they became pregnant. Future research must center the individual's realization of their pregnancy and fertility desires, including measuring the extent to which people are successful in avoiding a pregnancy they do not want to have and unsuccessful in becoming pregnant when they want to. The development of new measures can help sharpen the focus on individuals for whom targeted resources are needed and strengthen efforts to support equitable access to reproductive health services and person-centered care.
Acknowledgments
The authors are grateful for research assistance from Sean Finn and Noah Zazanis, for comments received on manuscript drafts from Joyce Abma, Aletha Akers, Jonathan Bearak, Joerg Dreweke, Sean Finn, Megan Kavanaugh, Laura Lindberg, Kimberley Lufkin, Isaac Maddow-Zimet, Lisa Melhado, Elizabeth Nash, Adam Sonfield, and Noah Zazanis, and for the contributions of many other colleagues in the field of sexual and reproductive health and rights who contributed to generative discussions that informed this work. We are also grateful to Jonathan Bearak for guidance in the application of methods for tests for statistical significance. The authors are solely responsible for the content.
Notes
Trans men, gender nonbinary people, and gender-nonconforming people can become pregnant and give birth, have an abortion, or experience fetal loss. However, data collection systems often assume all individuals who can become pregnant are captured in surveys by including those who identify as a “woman” or “female.” To date, most data sources with measures of pregnancy experiences have not incorporated sufficient measures to allow the identification of these varied populations to determine whether their experiences are captured in data collection efforts focused on “women” and the extent to which they are excluded. Given these limitations, we use the term “women” while recognizing that it does not accurately describe, and the data may not reflect, the experiences of everyone who is able to become pregnant.
Upadhyay et al. (2014) defined reproductive autonomy as “having the power to decide about and control matters associated with contraceptive use, pregnancy and childbearing.”
Birth rates since 2017 have continued to decline; while the rate increased somewhat between 2020 and 2021—from 56.0 to 56.6 per 1,000 women aged 15–44 (Hamilton et al. 2022)—it remained well below the rate of 60.3 in 2017 (Martin et al. 2018). Rates of abortion also increased in 2019 and 2020 (Jones et al. 2022), although the large decreases in birth rates likely mean that pregnancy rates overall have continued to decline, given that far fewer pregnancies end in abortion than in birth (Maddow-Zimet and Kost 2021).
Some respondents in the NSFG reported not using, or having stopped using, contraception just prior to the pregnancy in order to become pregnant. For these respondents, the question about whether they had wanted a baby just prior to the pregnancy was not asked; instead, they were directed to the question about their feelings toward timing of the pregnancy (see the online appendix).
In the widely used recode variable for pregnancy intentions provided in the NSFG data set (“wantresp”), respondents who responded “probably no” were coded as having pregnancies not wanted at that time or ever. In our measure, those respondents’ pregnancies were coded as “other” because they were not asked about timing and their initial response to the question on pregnancy wantedness was something other than “yes” or “no.”
These timing categories could also include births to respondents who were not asked about whether they had wanted a baby but instead were directed to the timing question because they reported that they had stopped using contraception in order to become pregnant.
An alternative approach would be to choose all births reported in the three years prior to each survey interview, or those occurring in the years 2014–2018 for the most recent NSFG data release. However, this approach would treat each year of interviews as a representative sample, which it is not (NCHS 2020); each two-year survey period, spanning three calendar years, is designed to be representative as a whole, which guided our approach.
The 2015–2017 and 2017–2019 NSFG included individuals aged 15–49; however, because earlier rounds were limited to individuals aged 15–44, we limited analyses to those aged 15–44 to generate comparable estimates across time.
We did not test for differences in rates across desire categories, so we cannot say with certainty that rates of abortions characterized by pregnancies occurring too soon are higher than rates of those characterized by not wanting a baby. However, confidence intervals for the two estimates do not overlap.
The NSFG question about timing is preceded by other questions anchoring “Before you became pregnant”—and for the first reported pregnancy—to introductory text reinforcing for the respondent that the next few questions will be about the time just before the pregnancy. The interviewer training is designed to minimize the possibility that the respondent shifts the time reference when answering about timing (J. Abma, personal communication, January 20, 2023).
