Using nineteenth century legal information combined with census information, I examine the effect of state laws that restricted American women’s access to abortion on the ratio of children to women. I estimate an increase in the birthrate of 4 % to 12 % when abortion is restricted. In the absence of anti-abortion laws, fertility would have been 5 % to 12 % lower in the early twentieth century.
The early twenty-first century United States has seen a wave of state legislation aimed at curtailing the availability of abortion that in many ways mirrors earlier legislative efforts from the nineteenth century (Boland and Katzive 2008). At the same time, scholars have shown renewed interest in nineteenth and early twentieth century fertility patterns, both in terms of the fertility transition when the United States experienced one of the world’s largest declines in fertility, and in terms of women’s access to fertility control (e.g., Bailey 2010; Guinnane 2011). Although good estimates are available on the effects of access to fertility control on twentieth century populations in the United States (e.g., Bailey 2010, 2012; Gruber et al. 1999; Levine 2004; Levine et al. 1999; and many others) and in modern developing countries (see Guinnane 2011 for a literature review), supply-side hypotheses are largely absent from the nineteenth century U.S. fertility transition literature (Guinnane 2011; Haines 1986, 1987; Haines and Guest 2008), with a few exceptions (David and Sanderson 1986; Degler 1980; Haines and Hacker 2006; Reed 1978; Sanderson 1979; Tolnay and Guest 1984; Wahl 1986). This article brings these literatures together, examining the effect on fertility of legislation controlling access to legal abortion in the nineteenth century.
Researchers have explored many demand-side hypotheses that affect fertility through the quantity-quality trade-off for children or through the opportunity costs of women’s time. They have found relationships between nineteenth century fertility and the changing roles of children, including child labor and increased educational opportunities (Carter et al. 2004; Degler 1980; Easterlin 1976a, b; Easterlin et al. 1978; Forster and Tucker 1972; Guest 1981; Guest and Tolnay 1983a, b; Guinnane 2011; McLaren 1990; Reed 1978; Tolnay and Guest 1982, 1984; Yasuba 1962), related needs for social insurance (Guinnane 2011; Haines and Hacker 2006; Steckel 1992), and bequest motives (Carter et al. 2004; Easterlin 1976a, b; Easterlin et al. 1978; Haines 1987; Steckel 1992; Sundstrom and David 1988). Women’s education and their greater labor market opportunities outside the home are also important (Easterlin 1976a; Guest 1981; Guest and Tolnay 1983a, b; Guinnane 2011; Tolnay and Guest 1982), as is literacy in general (Carter et al. 2004; Degler 1980; Guest 1981; Guest and Tolnay 1983a; Guinnane 2011; Haines 1987, 2000; McLaren 1990). Other work has stressed structural correlates, such as increasing urbanization and industrialization (Guest 1981; Haines 1987, 2000; Tolnay and Guest 1982), the role of agriculture and land availability (Carter et al. 2004; Degler 1980; Easterlin 1971, 1976a, b; Easterlin et al. 1978; Forster and Tucker 1972; Guest 1981; Guest and Tolnay 1983a, b; Guinnane 2011; Haines 2000; Haines and Guest 2008; Sundstrom and David 1988; Yasuba 1962), and changing female/male ratios (Guest 1981; Haines 2000). The percentage of the population that is immigrant is also known to be related to increased fertility (King and Ruggles 1990).
The literature on the effects of the supply of fertility control on fertility in the twentieth century is rich and growing. Bailey (2010) found that 1960s fertility would have been 8 % lower in states with nineteenth century Comstock laws curtailing access to fertility control and 4 % lower for the entire United States. Based on analysis of Roe v. Wade in the 1970s, Levine et al. (1999) found a decrease in fertility of at least 4 %; similarly, Ananat et al. (2009) found an initial decrease of 5.7 % in birth rates overall because of abortion law liberalization in the early 1970s. Angrist and Evans (1999) suggested that liberalizing abortion laws in the 1970s decreased teen fertility between 2.2 % for white teens and 18 % for black teens—groups that are particularly vulnerable and may have less access to outside options. Guldi (2008) explored the effects of 1970s abortion and birth control laws affecting minors directly and found a stronger drop for white teens, concluding that access to abortion leads to an 8 % to 15 % decrease in birth rates for minors in addition to the effects of adult laws. Also looking at more vulnerable groups, Bailey (2012) found that family planning programs from the 1960s and 1970s reduced childbearing among poor women by 19 % to 30 %.
This article adds information on one topic that could affect the supply of fertility control technology in the nineteenth century: laws restricting abortion access. When abortion is restricted, the birthrate increases 4 % to 12 %. In the absence of anti-abortion laws, fertility would have been 5 % to 12 % lower in the early twentieth century.
Nineteenth century abortion technologies were potentially dangerous. For early-term abortions, herbal remedies of unknown efficacy and varying safety were common. Later-term abortions included membrane rupture and dilation and curettage (D&C) (Brodie 1994; Ernst 2002; King 1992; Smith-Rosenberg 1985).
Very early abortion laws functioned as malpractice laws designed to protect women but became more restrictive by the 1850s, increasingly punishing earlier-term abortions and prosecuting rather than protecting women (Brodie 1994; Degler 1980; Lader 1966; Mohr 1978; Polsky 1970). Theories for the reasons for these increased restrictions include power consolidation by the American Medical Association (AMA) (Degler 1980; Mohr 1978; Reed 1978) and worries about the high fertility of immigrants compared with natives (Smith-Rosenberg 1985).
Data and Methodology
Abortion laws were compiled from secondary sources (Dellapenna 2006; Dennett 1926; Lader 1966; Mohr 1978; Quay 1961; Storer 1860; Storer and Heard 1868) and from superseded state statutes preserved at the Harvard Law Library and the University of California–Los Angeles (UCLA) law library. Additional laws were obtained via Google Books and state law librarians. A list of the dates of these laws can be found in Table 1. In cases in which a court ruling prior to the law is mentioned in the law books, the earlier ruling is used, something that affects only Kentucky (1879 vs. 1880) and Pennsylvania (1846 vs. 1860). Results are robust to coding only the law.
Census information by state during this time period is limited. For that reason, I use one standard measure of nineteenth century fertility—the child-to-woman ratio, calculated as the ratio of the number of children aged 0–9 to the number of women of childbearing age, or 15–44. The original data come from the Haines census tables in the Historical Statistics of the United States (Carter et al. 2006), and I replace the Haines data with collapsed cells from the Integrated Public Use Microdata Series (IPUMS) 100 % sample for the 1880 census (Ruggles et al. 2010). I correct in the denominator for enumeration differences across decades with the Yasuba interpolation for 40- to 44-year-olds from data for 30- to 39-year-olds, 40- to 49-year-olds, and 50- to 59-year-olds.1 This measure is highly correlated with total fertility (Haines and Hacker 2006) but cannot capture the nonlinear nature of annual fertility changes and is sensitive to migration and mortality. Easterlin (1976a), Guinnane (2011), Haines and Hacker (2006), Tolnay and Guest (1982), Tolnay et al. (1982), and Yasuba (1962) provided more thorough discussions of the benefits and limitations of these measures.
Controls for percentage immigrant and the ratio of women to men come from the same Haines data. The control for percentage of the population that is urban was collected from historical census tables published in 2010.2 The percentage of those literate over age 20 by state/year was created from IPUMS (Ruggles et al. 2010).
I estimate the impact of abortion laws on the child-to-woman ratio for children between the ages of 0 and 9 and women between the ages of 15 and 44 in the years from 1850 to 1910. I do not include Arizona, New Mexico, North Dakota, Oklahoma, South Dakota, or Utah because either they were not yet states or they lack population data. States that had legal and population information collected while they were still territories are included.
Because of limitations with decennial data, the independent variable of interest for having a law is a share of the decade for which the law was relevant to childbearing, with a one-year lag because abortions in year 0 cause a change in births in year 1. For example, for the decade ending in 1880, a law passed in 1876 would be coded as 0.4, and a law passed in 1870 or earlier would be coded as a 1.
Table 2 reports the results from Eq. (1). All these results include state fixed effects and year fixed effects to control for differences between states and years in fertility and likelihood of obtaining a law. In the base regression with the (children aged 0–9)/ (women aged 15–44) ratio as a level, shown in column 1, having a law increases the child-to-woman ratio by 111 per 1,000 women. Similarly, column 2 repeats this regression with the log of the child-to-woman ratio as the dependent variable and finds an increase of around 12 %.
Previous researchers, as noted in the literature review, have found effects of percentage immigrant, percentage urban, schooling, and female-to-male ratios on the child-to-woman ratios. Therefore, it is interesting to examine the effect of these controls in conjunction with abortion laws, keeping in mind that any variation must be at the state × year level because variation at the state level and variation at the year level are washed out by state and year fixed effects, respectively. A higher percentage of immigrants in the population should increase fertility and may encourage nativists to pass anti-abortion legislation. Column 3 adds the percentage of immigrants by state/year to the baseline regression. Controlling for the percentage of immigrants increases the effect of having an abortion law on fertility to 14 % and itself, as expected, has a positive and significant effect on the child-to-woman ratio, with a 1 percentage point change in the percentage of immigrants increasing the child-to-woman ratio by 1 %.
Column 4 provides one measure of education: the percentage of those over age 20 who are literate. Although it is insignificant and does not change the law’s effect, its sign is as expected, with greater education decreasing population. Column 5 includes the female-to-male ratio and shows, as in prior literature, that a higher proportion of women decreases the child-to-woman ratio but does not change the effect of abortion laws on the child-to-woman ratio.
Similarly, previous literature would lead us to expect that a higher percentage of urban would decrease the child-to-woman ratio, although it is unclear a priori whether a high urban population would have an effect on passing anti-abortion legislation. Column 6 demonstrates this negative effect on fertility and also decreases the effect of having an abortion law on fertility to 5.7 %.
Column 7 shows all the controls together in one regression, for a predicted increase of 5.5 %.3 Column 8 adds in a state-specific time trend to control for linear variation over time at the state level. These results attenuate to 4.4 % with significance at the 10 % level, although it is unclear whether this attenuation is because of the change in specification, the addition of more variables to a regression with an already small sample size, or the effects of improperly including a state-specific time trend (for a discussion, see Wolfers 2006).4 It is important to note that the results in column 8 are not statistically different from those in columns 6 or 7.
Discussion and Conclusion
Anti-abortion legislation increased nineteenth century fertility causing the child-to-woman ratio to increase 4 % to 12 % in states with laws. As in previous literature, the percentage of the population that is immigrant is associated with increases in fertility. However, percentage of immigrants does not affect the coefficient on having an abortion law; if anything, it increases the effect of the law. Literacy is negatively correlated with fertility, although not significantly so with my measure of literacy. The female-to-male ratio is also negatively associated with fertility. Percentage of urban decreases the fertility ratio and attenuates the effect of abortion laws. These findings are all consistent with previous literature on the demand for lower fertility.
The nineteenth century demographic transition would have been even more dramatic in the absence of these laws. Table 3 provides a rough calculation that predicts what the child-to-woman ratio would have been in the absence of laws restricting abortion. Fertility would have been 5.2 % to 10.9 % lower without anti-abortion laws in 1900 and 5.7 % to 11.9 % lower in 1910. These estimates are consistent with modern results from the twentieth century, which find that greater access to birth control, abortions, and family planning during the 1960s and 1970s led to fertility increases of 2 % to 6 % overall, and up to 15 % among groups with greater need and less flexibility (e.g., teenagers and minorities). As with these worse-off groups, nineteenth century populations may have also had less ability to travel to access abortions and more dire consequences to an additional unwanted child than many modern populations. The results in this article suggest that in the nineteenth century, just as in the twentieth century, laws prohibiting fertility control had real effects on population growth. This evidence of unmet demand for fertility control suggests that the supply side is important in the analysis of fertility in the nineteenth century just as it is today.
Thanks to Elizabeth Ananat, Martha Bailey, Dora L. Costa, Laura Dague, Claudia Goldin, J. David Hacker, Michael Haines, Ann McCants, Bob Margo, Karen Norberg, and Lori Taylor for helpful discussions; Michael Haines for assistance with census tables; and state law librarians in many states for assistance with historical abortion laws. Thanks also to seminar participants at APPAM, the University of California at Riverside, the University of Kansas, the NBER Development of the American Economy, Northwestern University, Pomona College, and SSHA; to Jillian Boles, Alejandro Bras, Erin Harrison, Rebecca Willis, and YiDing Yu for excellent research assistance; and to M. Rose Barlow and Mary Cozad for comments. Finally, thanks to three anonymous reviewers for helpful comments. The author thanks the National Institute on Aging NBER Grant # T32-AG00186 for funding and support.
The interpolation formula is (A + 8B – C) / 16, where A is the 30–39 cohort, B is 40–49, and C is 50–59.
The entire set of citations for these data is not included for brevity’s sake. An example of the basic format of these citations can be found in the entry for the U.S. Census Bureau (2012) in the reference list.
None of the controls listed significantly predict the passage of a law within the next decade using Eq. (1) and law passage as a dependent variable.
Charts and tables examining the nonlinearity of the post period (available from the author) suggest this latter explanation may be correct but are not conclusive given the fragility of the data and the length of the period studied.