American families changed dramatically during the last third of the twentieth century. From 1900 until the late 1960s, roughly three-quarters of all American sixteen-yearolds had lived with both of their biological parents. By 2000 only about half of all sixteenyear- olds were living with both biological parents. During the first half of the twentieth century, moreover, most parents who were not living with their children had no choice about the matter: They were dead. By the end of the twentieth century, most of the parents who were not living with their children were alive but living elsewhere (see figure 1.1). The focus of this chapter, however, is not simply the fact that family structures changed, but the fact that they changed very differently depending on parents' education and race. All groups are postponing marriage, but not all are postponing parenthood. As a result, the rise in single-parent families is concentrated among blacks and among the less educated. It hardly occurred at all among women with college degrees. Children's families have changed in broadly similar ways throughout the developed world, but no other nation has changed as much as the United States. The best evidence on this comes from the Fertility and Family Surveys (FFS), which provide data on the stability of both marriages and cohabiting unions in Europe during the late 1980s and early 1990s. Gunnar Andersson (2002) used these data to estimate the fraction of fifteen-year-olds who would live with both of their biological parents if parents continued to split up at the same rate as that found at the time of the FFS.1 Figure 1.2 shows his projections for the United States and six countries in western Europe. If nothing changed, roughly seven-eighths of southern European fifteen-year-olds and two-thirds of northern European fifteen-year-olds would live with both of their biological parents at age fifteen, compared to only half of all American fifteen-year-olds. The transformation of the American family since 1960 has been both an intellectual challenge and a recurrent source of frustration for social scientists. Some of the nation's best-known social theorists, including Gary Becker and William Julius Wilson, have sought to explain the change. Yet no consensus has emerged about why American families changed or why the amount of change varied by race and education. The most widely cited empirical papers seem to be those that disprove a popular explanation, not those that support one. Indeed, it is only a slight exaggeration to say that the main contribution of empirical social science to our understanding of family change has been to show that nothing caused it. Yet despite the absence of an identifiable culprit or a smoking gun, families did change. This report is divided into five sections. The first section discusses which changes in family structure should worry us. We argue that: • Those whose primary goal is to reduce child poverty should mainly worry about the increased proportion of children living with only one adult, especially adults with relatively low potential earnings. • Those whose primary goal is to improve children's overall well-being should also worry about the fact that even children in two-adult households are less likely to be living with both of their biological parents and more likely to be living with a stepparent or their mother's live-in boyfriend. • Those whose primary goal is to bring American behavior into line with traditional moral rules should worry about the rising age of marriage (which has led to a large increase in premarital sex), the decline of shotgun weddings (which has led to a large increase in nonmarital births), and the high rate of divorce. The next section describes the most important changes affecting children's living arrangements over the past generation. • Single-parent families have become more common in all demographic groups, but the increases have been greatest among less-educated women and African Americans. • Women in all demographic groups are marrying later than they did a generation ago. But while highly educated women have postponed both marriage and parenthood, less-educated women have postponed marriage more than parenthood. As a result, nonmarital births have risen dramatically among less-educated women but more modestly among the highly educated. • Compared to white women, African American women have also postponed marriage more than childbearing, so nonmarital births have risen faster among African Americans than among whites. • White women with a college degree seldom have children out of wedlock. Most of those who become single mothers marry before their child is born but then divorce. Divorce rates among the highly educated have not risen since the early 1980s, so the fraction of highly educated mothers raising children on their own has not changed much since 1980. • The increase in nonmarital births, which continued into the 1990s, was confined to the bottom two-thirds of the educational distribution. As a result, the association between a woman's educational attainment and her chances of being a single mother rose after 1980. • Cohabiting couples now account for nearly half of all nonmarital births in the United States. Among whites, cohabiting couples account for almost all of the increase in nonmarital births since 1980. But unlike cohabiting couples in Europe, cohabiting couples in the United States usually split up within a few years of their baby's birth. The fact that "high-risk" couples have been switching from marriage to cohabitation since 1980 probably played a role in stabilizing the divorce rate. It is not clear whether the switch from marriage to cohabitation increased the risk that any given couple would split up. • Attitudes toward premarital sex, gender roles, the value of marriage, and the acceptability of nonmarital childbearing changed rapidly during the late 1960s and 1970s. Such attitudes have changed far less since 1980. Changes in attitudes have not always moved in tandem with changes in behavior. Premarital sex increased sharply in the 1960s and 1970s, when attitudes were becoming more permissive. But premarital sex continued to increase during the 1980s and early 1990s, when attitudes towards premarital sex were not changing much. • The age at which individuals first have sexual intercourse (whether marital or premarital) fell only a little between 1960 and 1990, while the age at which individuals married rose substantially. Taken together, these two facts suggest that premarital sex mainly replaced marital sex, not abstinence. This pattern does not vary much by education. • In the early 1960s, most marriages were closely followed by a first birth. That linkage has now been broken. More women are having their first child well before they marry, and more women are marrying well before they have their first child. The third section discusses the social science literature on the role of wages, welfare benefits, and the sex ratio in explaining family change. The following section discusses the role of non-economic influences, such as changes in women's sense of control and efficacy, changes in attitudes toward nonmarital sex, cohabitation, nonmarital births, and divorce, and improvements in contraception, the legalization of abortion, and no-fault divorce. In the next section, we present some new evidence regarding some of these theories. We then present our conclusions in the final section. We come to several conclusions regarding the existing literature on changes in family composition: • Traditional economic models emphasize the economic advantages that flow from the fact that one partner can specialize in child-rearing while the other specializes in market labor. These models predict that improvements in men's earning power will make marriage more common, while improvements in women's earning power will make marriage less common. These models also predict that lower welfare benefits will make marriage more common, while a lower ratio of men to women will make marriage less common. • The wages available to men, the wages available to women, welfare benefits, and the sex ratio can explain a large fraction of the change in single-parenthood for different racial and educational groups. But the estimated impact of these variables is extremely sensitive to how we specify the statistical model.