Introduction
Induced abortion is not an obvious topic in a section on health economics. Although being a common procedure, abortion does not contribute to rising medical expenditures or inflation. There were 1.1 million surgical abortions in the US in 2008, but the number of abortions has fallen overtime, although the inflation-adjusted cost of a first trimester abortion has remained remarkably stable at approximately $450. Nor have there been dramatic technological breakthroughs in the delivery of abortion. The most significant innovation is RU-486, more commonly referred to as the ‘abortion pill.’ However, its impact on the demand for and availability of abortion services has been modest at best. Finally, abortions are extremely safe with only 0.7 deaths per year per 100 000 procedures between 1988 and 1997. In contrast, the maternal mortality rate in the US is 15 times greater.
So why include an article on abortion? Two reasons. First, induced abortion, a medical procedure performed only by physicians, is one of the most contentious and divisive issues in the politics of many countries today. In the US, clinicians who perform abortions and staff workers who assist them have been murdered and their clinics vandalized. Politicians are defined by their stance on abortion and Supreme Court nominees must tread carefully when discussing the precedent set by the Court’s decision in Roe versus Wade. Academic research on abortion has not been protected from this scrutiny. Donohue and Levitt’s (2001) study linking the legalization of abortion to the decrease in homicide rates 20 years later was extremely controversial, received widespread exposure in the popular press, and became a central chapter in the hugely successful book Freakonomics.
The second reason to include a review of abortion is because the indirect effect of abortion on health is potentially large but empirically challenging to document. Induced abortion, the focus of this article, also represents a conscious decision to end a pregnancy, unlike spontaneous abortion which is an involuntary and largely random termination of pregnancy. Arguably the most notable link between abortion and health or well-being is the hypothesized relationship between abortion and crime (Donohue and Levitt, 2001). If Donohue and Levitt are correct, then the legalization of abortion averted 15 000 homicides over a 10-year period (Joyce, 2009). But homicide is but one measure of well-being. If abortion has a profound effect on crime, then it likely affected other measures of wellbeing such as marriage, schooling, drug use, and sexually transmitted diseases to name but a few. And yet the empirical challenge of isolating a cohort effect from constantly evolving period effects may be insurmountable given the data and methods available to researchers.
In this article the focus is on the link between induced abortion and health. Health is broadly viewed to include measures of well-being such as crime and drug use in addition to the more commonly associated measures of health such as infant mortality. Given space limitations, the author concentrates primarily on the US experience with legalized abortion from roughly 1970 to present. The history of abortion in the US is available from a number of sources (Garrow, 1998). The author concentrates instead on two empirical challenges for researchers that have tried to uncover a link between abortion and health. The first is identification. How does one measure the impact of a pregnancy that is never carried to term? The second is data. Unlike births, induced abortions are not part of a national vital registration system. Moreover, abortions are poorly reported in surveys as women are reluctant to admit to them. Finally, the review is selective. The author discusses in detail papers believed to be the most important because of the quality of the research design and their impact on subsequent research. There is more to be learned by careful study of the best papers than a quick pass through the entire literature.
The article is organized as follows. The author first discusses the conceptual mechanisms by which abortion is linked to health. This is followed by a description of data on abortion and the demographics of abortion. The next few sections discuss empirical work supporting possible links. The literature is broadly divided between studies on the determinants of abortion and its impact on fertility and those that estimate either the structural or reduced-form association between abortion and health. There has been relatively little work on the supply side of abortion markets.
Conceptual Link Between Abortion And Health
How does one study the health of a fetus that has never been born? The simple answer is that you cannot, which necessitates indirect approaches. Demographers, for example, consider abortion as an expression of an unwanted pregnancy. They assume that the wantedness of a pregnancy varies along a continuum from those that are aborted to pregnancies that are mistimed but carried to term. Thus, even pregnancies that result in live births may be characterized as unwanted and contrasted with the outcomes of births described as wanted. Data on wantedness in the US come from surveys of new mothers in which they are asked about their pregnancy intention when they first discovered that they were pregnant. Births are classified as wanted, mistimed, or unwanted on the basis of a series of responses by the mother. Mothers whose pregnancies are unwanted at conception are hypothesized to smoke more or receive less prenatal care than mothers whose pregnancies were planned. As a result, births from pregnancies that are unwanted are expected to be less healthy than births from pregnancies that are wanted. Neglect is hypothesized to continue after birth. Children who were unwanted at conception may receive less nurturing than those who are wanted. The result would be lower academic achievement, behavioral problems, and possible delinquency as adolescents (Brown and Eisenberg, 1995).
It is unclear whether unwanted pregnancies based on post hoc surveys of women who gave birth provide insights as to the outcomes of pregnancies that are aborted had they instead been carried to term. Early studies of wantedness in Europe tried to estimate the impact of the latter by analyzing outcomes of women who were denied abortion. The most famous sample is the Prague Cohort of 1961–63. A total of 220 children whose mothers were twice denied an abortion for the same pregnancy were matched to children whose pregnancies had been wanted and followed for 30 years. There were few differences between the unwanted cohort and their wanted controls at birth, but by the age of 20 years, there was evidence of less personal satisfaction and psychological instability.
Economic models that linked abortion and health were first discussed by Grossman and Jacobowitz (1981). The authors argued that abortion as a method of fertility control helps parents to achieve a desired family size. Using models of the family and household production pioneered by Becker and Lewis (1973); Grossman and Jacobowitz (1981) incorporated abortion reform into a model of infant mortality. Parents maximized a utility function that depended on consumption goods, the number of births, and the survival probability of each. Both the number of children and their survival probability were choice variables. The survival probability depended on a set of endogenous inputs. Thus, parents affected the health of an infant by their choice of goods (e.g., cigarettes) and medical care during pregnancy. The model generated a structural and reduced-form production function of child survival. Grossman and Jacobowitz (1981) argued that subsidized family planning services and legalized abortion decreased the cost of fertility control which lowered the optimal number of births but raised the survival probability of each.
This quantity–quality framework became the explicit model in many of the empirical analyses that followed.
Lowering the price of an abortion allowed women and parents greater control over the timing and number of children. This gave parents more control over the quality of each child as parents used time and market goods to enhance a child’s health and human capital. Thus, pregnant teens could delay birth until they were more financially and emotionally prepared for parenthood. Older women could terminate unwanted fetuses that could divert resources from their current children or abort fetuses that were at a greater risk of poor health. With the advent of genetic testing and advanced sonography, abortion as a fetal selection mechanism became even more explicit.
Refinements of the selection mechanism followed. Abortion was characterized as one decision along a sequence that included the decision to get pregnant, the decision to abort or give birth, and the decision to marry or remain single (Grossman and Joyce, 1990; Lundberg and Plotnick, 1990). Increases in the cost of abortion impacted these other decisions. For instance, some women use pregnancy as a way to assess the suitability of a potential father. Increasing the cost of an abortion raises the price of this ‘option,’ resulting in fewer abortions but fewer pregnancies as well.
Abortion as a sorting mechanism is not the only pathway through which women and their potential offspring were affected. Akerlof et al. (1996) developed a model in which women’s bargaining position with men was weakened by the availability of safe, legal abortion. Before abortion, sex was more closely linked to commitment. If an unmarried woman became pregnant, there was pressure on the man to ‘to do the right thing’ by marrying her. Abortion altered that expectation. Women willing to abort could have sex without an implied commitment of marriage in the case of pregnancy. Men could insist that a pregnancy be terminated instead of marriage. This put women opposed to abortion at a disadvantage in attracting men. To compete for men they had to be more willing to have sex without a commitment of marriage. The model predicts that the legalization of abortion will result in a decrease in ‘shotgun’ marriages and an increase in out-of-wedlock childbearing. Both predictions are consistent with the stylized facts in the 1970s. The link to health comes through the immiseration of women and children as the number of female-headed households rise. Economists used the model by Akerlof et al. (1996) to argue that the legalization of abortion could be associated with the rise in crime, in direct contradiction to Donohue and Levitt (2001).
The Akerlof et al. (1996) framework has not been used in the empirical literature on abortion and health. The quantity–quality framework has been the mainstay in the literature. By enabling parents to achieve an optimal number of births, abortion enhances the resources devoted to the children who are born. Thus, any empirical association between abortion and health rests importantly on the association between abortion and fertility. This may seem obvious because an aborted pregnancy is an averted birth. However, other methods of fertility control are substitutes for abortion which implies that a rise in the abortion rates need not be associated with a fall in birth rates. Couples that may have used condoms before the legalization of abortion may be less vigilant about contraception after legalization. A pregnancy that occurs under a regime on legalized abortion may not have occurred under a regime in which abortion is prohibited. Without demonstrating that a change in the birth rate is associated with a decrease in the price of an abortion, it is difficult to establish that parents are trading off quantity for quality.