Econometric analyses of U.S. abortion policy: a critical review.

• Author: Klick, Jonathan

INTRODUCTION

Few social issues in the United States are as contentious as the legal status of induced abortion. Thirty years after the Supreme Court declared state laws restricting abortion unconstitutional in Roe v. Wade, (1) poll results suggest that the United States population is almost evenly split over whether the next nominee to the Supreme Court should support or oppose legal abortion in most or all contexts. (2) The salience of the issue is arguably among the highest of all national issues. (3)

The motivation behind many individuals' positions regarding abortion policy hinges on normative judgments. That is, many people arrive at their abortion position based on some moral decision about the relative rights of the mother and the fetus. (4) Presumably there is a consequentialist component to the abortion question that is largely ignored in discussions of public opinion about abortion. (5) This consequentialist component, however, has received the bulk of attention in the social science literature regarding abortion policy.

The social science literature on the effects of abortion policy has grown tremendously during the last decade. While public health and demographic scholars had consistently examined the effects of changes in abortion policy even before Roe, (6) there has been an explosion of research on the subject in economics journals in recent years. While a few of these articles are primarily theory based, (7) the lion's share of the increase in attention devoted to this subject has come in the form of econometric or statistical analyses of changes in abortion policy.

In many ways, changes in abortion policy, especially the issue of legalization, present empirical researchers with an ideal mechanism through which to identify causal relationships between changes in incentives relating to sexual behavior and a host of demographic and public health outcomes. The costs and benefits of sexual behavior are subjective and are likely to be correlated with many observable and unobservable individual characteristics. This makes it difficult to draw causal inferences about the sensitivity of sexual behavior to incentives. Even if data are available on the costs and benefits of sexual behavior accruing to an individual, a fairly heroic assumption in most contexts, any observed correlations between behavior, or outcomes and changes in those costs and benefits might very well be artifacts of omitted variables biases. (8) Changes in abortion policy, however, generally affect costs and benefits of sexual behavior in known and unambiguous ways. Further, since the changes are generally the result of legislative action or judicial fiat, they are likely to be exogenous. That is, the changes will be orthogonal or unrelated to an individual's characteristics, obviating the concern that the changes in costs and benefits will be correlated with important variables that are omitted in the econometric analysis. (9)

Econometric researchers have exploited this attractive aspect of changes in abortion policy to examine numerous important social and demographic relationships. These include the effect of incentives on sexual activity, and the effect of limiting unwanted births on welfare payments, crime rates, and women's educational attainment, as well as the direct effect of changes in abortion policy on abortion and fertility decisions.

Setting aside the rights-based arguments for and against abortion availability, these positive analyses have large implications for the evaluation of abortion policy. A sophisticated understanding of what econometric research tells us about the effects of abortion policy could greatly improve social welfare by informing policymakers about the consequences of their actions. Unfortunately, the econometric literature on this topic is technical, placing it beyond the ken of most policymakers. Further, some media popularizations of research in this field have been influenced by the various groups who have a vested interest in seeing abortion policy swayed in one direction or another. Both of these concerns create a troubling disconnect between scholarly research and policy in this area.

In this article, I will attempt to describe the econometric research on abortion policy in a manner that is accessible to someone without formal training in econometrics, highlighting the policy implications of the research as well as any serious methodological shortcomings which might limit the value of any particular piece of scholarship. In many ways, out of necessity, this review will not be exhaustive. I will focus primarily on research using U.S. policy changes and data. This is not meant to suggest that there is no quality research examining the experiences of other countries. There is, in fact, a large literature looking at foreign experiences. (10) If the primary goal of this article, however, is to inform U.S. policy, it will generally be most useful to examine U.S. sources. I have also largely ignored the interesting empirical literature examining the determinants of abortion positions held by voters and politicians. (11) These studies are not directly relevant to evaluating U.S. abortion policy. (12)

In terms of organization, I have divided this review into categories based upon the dependent variables examined in each of the econometric studies. Broadly speaking, the majority of studies can be categorized as looking at outcomes in the following areas: sexual behavior, crime, opportunities for women, and public finance effects. I have chosen to exclude the very large literature examining the effects of abortion policy on demographic patterns. This omission does not reflect any lack of interest in this literature. Covering the research in this area even superficially would more than double the length of this article. Further, a book covering this material in great detail has been written by one of the foremost researchers in this area, Phillip Levine and will be published later in 2004. (13) Given that Levine discusses the existing research in the non-technical tone, to which I aspire in this article, anything I would add by way of comment on this research would be largely superfluous.

Another organizational device that I chose not to employ is differentiating between studies focusing on abortion legalization and studies focusing on subsequent policy changes. While there is a good deal of heterogeneity in the policy variation that econometric studies exploit, as a theoretical matter, all of the changes boil down to changes in access to abortion. Put a different way, policy changes can be seen as changes in the effective cost or price of abortion. For example, legalization simply lowers the cost of obtaining an abortion, since it removes any penalties that the state can impose on those seeking or providing abortions; it also reduces the search costs entailed in finding an abortion provider; (14) and, more than likely, it makes abortion procedures safer. (15) Similarly, increased public funding for abortions through such programs as Medicaid, lowers the cost of abortion. Parental notice requirements, mandatory waiting periods, requirements that information about abortion alternatives be provided to individuals seeking an abortion, and the like, all serve to increase the effective cost of obtaining an abortion. Restrictions on the activities of anti-abortion protestors however, lower the cost of abortion. (16)

1. THE EFFECT OF ABORTION ACCESS ON SEXUAL BEHAVIOR

   The link between abortion access and sexual behavior is fairly straightforward. The risk of an unwanted pregnancy represents a cost of risky sexual activity. If individuals make their choices about sex rationally, as the marginal cost of engaging in sex decreases, we would expect individuals to have more sex. As access to abortion becomes cheaper, the expected cost of an unwanted pregnancy decreases, since abortion can be viewed as ex post birth control or insurance against an unwanted pregnancy, this implies that individuals will engage in more risky sex. (17)

   The rational basis for decisions about sex goes largely unquestioned in the economics literature, though that is not true of other social sciences. (18) Posner provides a good overview of the costs and benefits considered by individuals when making decisions about sexual activities. (19) Phillip Levine reviews some of the evidence for the rational choice model of teenage sexual behavior, contrasting it with the more spontaneous or irrational conceptions of sex that are prominent in other fields. (20) Using state level data, he shows that variations in cost significantly affect the decision to engage in sexual activities, as well as the decision to use preventive birth control measures.

   Formally speaking, if an individual's utility is a positive function of the quantity of sex consumed, and the cost of sex is the increasing likelihood that it will result in an unwanted pregnancy, individuals will consume sex up to the point where the marginal benefit of sex equals the marginal cost of sex. As long as there is decreasing marginal utility of sex, a decrease in the likelihood of an unwanted birth will induce individuals to have more sex. (21)

   Testing the empirical validity of this hypothesis, however, is not easy. Data gathered by researchers on sexual activity suffer from many problems. The earliest large-scale dataset on sexual activity, at least for the United States, was collected by Alfred Kinsey through his Institute for Sex Research, which was founded in 1947. (22) Though the publication of Kinsey's Sexual Behavior in the Human Male (23) in 1948 and Sexual Behavior in the Human Female (24) in 1953 can rightly be called the beginning of systematic empirical research in the field of sexuality, the data are far from inclusive in terms of years and areas covered. (25) Therefore, their usefulness is limited for researchers intending to use them to study reactions to state level...