Abortion and Selection

Abstract

HE legalization of abortion in the United States in the early 1970s represents one of the most important changes in American social policy in the twentieth century. This policy change had obvious implications for the likelihood of giving birth in the case of an unintended pregnancy, and resulted in a drop in birth rates. In addition, we will demonstrate that it led to a rise in pregnancy rates. As a result of both of these changes, abortion legalization may have altered the characteristics of birth cohorts. In particular, children’s outcomes may have improved on average, because they were more likely to be born into a household in which they were wanted. Two earlier papers have investigated the implications of such positive selection through abortion for the quality of cohorts born after abortion legalization, but have used different sources of variation that, implicitly, relied on different aspects of the selection process. Gruber, Levine, and Staiger (GLS, 1999) examined the living circumstances of children, and focused on the reduction in births caused by initial legalization. Donohue and Levitt (DL, 2001) examined crime among youths, 1 focusing on differences in abortion rates between states after legalization. This paper builds on those results by considering the impact of abortion access on a broad array of cohort characteristics in adulthood. Since the relevant cohorts were born in the early 1970s, children born at that time are now in their thirties. Using the 2000 Census, we can examine adult characteristics, such as completed educational attainment, employment, and poverty status. Beyond evaluating outcomes in adulthood, this paper also makes important contributions by addressing the inconsistent methodological approaches and the implicit models of selection used in past research. The approach of GLS used, as motivation, that (i) after five states repealed anti-abortion laws in 1970, relative birth rates fell in those states; (ii) after the 1973 Roe v. Wade decision, birth rates converged between early legalizing states and other states. They treated these two changes as a two-part experiment in reducing the rate at which unintended pregnancies translated into births. DL, on the other hand, noted that (i) abortion rates continued to rise after 1973, and that (ii) the abortion rate remained persistently higher in the early legalizing states even after widespread legalization. They used these diverging trends in the abortion rate as a measure of variation in the overall level of selection into pregnancy and birth. In this paper, we note that the post-1973 combination of (i) converging birth rates and (ii) persistent differences in abortion rates implies that in the latter half of the 1970s increased abortions did not map perfectly into fewer births. Instead, it must be the case that pregnancy rates rose in early legalizing states relative to other states in the late 1970s. Since the approaches used by GLS and DL rely on different changes in reproductive behavior, they are not equivalent. We introduce a comprehensive model of selection that identifies the two distinct mechanisms—increased pregnancies and decreased births—by which abortion access may alter cohort characteristics. We argue that in the first part of the decade, decreased births in early legalizing states, out of a steady number of unwanted pregnancies, were the dominant force in the selection process. In the late 1970s, on the other hand, we argue that a larger pool of pregnancies in repeal states, from which a steady number of births were selected, means that early legalizing states could continue to have more positively selected cohorts of children even after widespread legalization.