Welfare, the Family, and Reproductive Behavior: Report of a Meeting (1998)

tance to allow unmarried mothers to remain outside the (official) workforce has greatly eroded. Another possible answer is the increasing earnings disparity among full-time workers in the United States and attendant anxiety about the economic future. A third possibility, Dickens suggested, is that the narrow focus on AFDC eligibility has resulted from a lack of politically practicable alternatives: AFDC is much more amenable to policy change than are medical expenditures.

Most research on welfare has dealt with effects on work-related behavior, not on fertility-related behavior. But the effects of the welfare system on marriage and fertility have generated a considerable body of scholarly analysis as well. A review of this research provides clues to what to expect from the current round of welfare reform and a new generation of studies evaluating those reforms.

Table 2 summarizes results of 68 studies of effects of welfare on marriage and fertility, by type of study and by the race of the population studied. Across all methods, a majority of the studies find a significant effect of welfare (positive on fertility, negative on marriage). Although many studies also find insignificant effects and many others find a mixed pattern of results (some significant, some insignificant), an equal weighting of studies strongly suggests the existence of some welfare effects on demographic outcomes. When methods are used that many researchers find more credible ("cross-state changes"), the picture is muddied slightly: effects are greater for black women but smaller for white women. Nevertheless, the central tendency of the literature is clear enough. This pattern of results has been recognized among researchers, among whom there is now a rough consensus that welfare does have some effects on these demographic outcomes.

Robert Moffitt noted that the magnitudes of these effects are still quite uncertain. Not only are there still quite a few studies showing no significant effect of the welfare system (see Table 2), but also many of these studies use stronger methodologies and are sounder than the others (as in the results for whites). In addition, of the studies that find significant effects, some find the size of the effect to be very modest in magnitude compared with other influences on fertility and marriage, although some find sizable effects as well.

Nearly all the studies reviewed by Moffitt analyzed effects of changes, or cross-state differences, in the levels of AFDC benefits. However, new changes in state programs also deal with many other aspects of the system: removing limits on outside earnings, requiring work or training, limiting duration of benefits, eliminating benefit increases for later births, requiring that teenage recipients live with their parents, and so forth. The existing studies are nevertheless indirectly

relevant to the current issues, in Moffitt's view, by testing the general proposition that recipients and potential recipients change their marriage and fertility behavior in response to financial incentives in the welfare program. A further problem with the existing fertility studies, according to Howard Rolston, is that they do not usually differentiate first from subsequent births; it is only the latter that are directly affected by the family cap provisions.

The final point emphasized by Moffitt is that, even though there is a rough consensus that welfare has some effect on marriage and childbearing, the welfare system cannot explain the rise in nonmarital childbearing over the 1980s and 1990s because welfare benefits have been falling over that period (see the section above on trends in the welfare system). To explain that rise, some other factor must have been at work. Leading candidates are a rise in the earning power of women, even low-income women, leading them to be able to support themselves and their children without the earnings of a husband; a decline in the incomes of less educated men, which could have decreased their attractiveness as marital partners; and a decline in the numbers of men available, a hypothesis suggested for disadvantaged blacks (Wilson, 1987). There is considerable research on these other factors, but less research that compares welfare benefits to those facts and attempts to parcel out their relative influences (studies that have attempted to control for some of these other factors include Acs, 1995, 1996; Danziger et al., 1982; Darity and Myers, 1993, 1995; Duncan and Hoffman, 1990; Hoffman and Duncan, 1988, 1995; Lichter et al., 1996; Lundberg and Plotnick, 1990; Schultz, 1994). At the workshop, June O'Neill emphasized that the decline in male wages may have been so great than welfare benefits could have increased in relative attractiveness, whereas William Darity believed that it has been the decline in the pool of marriageable men that is the most important.

Disagreement about the results of past studies may also be due to a failure to focus on the populations in which a welfare difference is really to be expected (Rosenzweig, 1995) or a failure to distinguish intended from unintended pregnancies. At the workshop, Larry Bumpass reported that preliminary results of an analysis of data from the National Survey of Families and Households showed an effect of AFDC benefits where one would be expected: among low-income people (since AFDC benefits do not figure prominently in the plans of high-income people) and for intended pregnancies. But there were no effects on fertility of persons well above the poverty level or on unintended pregnancies for anyone (Bumpass and Brandon, 1996). The great majority (88 percent) of pregnancies to unmarried women are subsequently reported as unintended (Brown and Eisenberg, 1995).

Maynard et al. (1997) argue that the effects of AFDC/TANF family cap provisions on nonmarital fertility rates are likely to be small. The cap-induced changes in income associated with fertility decisions of welfare recipients are small in any case, both in proportion to the costs of childbearing and in absolute amounts. (Even before family caps, AFDC programs provided widely varying

Figure 3 Schema of decisions facing unmarried woman. Source: Klerman (1996).

increments for additional children, varying from $24 to $247 per child, depending on the state and the family size.) Additional children still produce additional benefits from the food stamp and housing programs; Medicaid benefits are unaffected by family caps.

The debate at the workshop surrounding the probable effect of family cap provisions on abortions served as a reminder that abortion decisions can be affected by a large number of considerations and conflicting pressures. Jacob Klerman proposed a schema for categorizing and analyzing the sequential and interlinked choices facing an unmarried woman (Figure 3). By closing off the possibility of outcome E (having a baby, staying single, and receiving higher welfare payments), welfare reformers usually hope that they are increasing the probabilities that women will follow path A (never getting pregnant in the first place, either through abstinence or effective contraception) or outcome C (getting married, before or after getting pregnant). They are willing to accept the likelihood that some women and their children will end up on path D (still single, with no increased payment or, under some welfare reform models, with no payment at all). Pro-life opponents of welfare reform fear that the effect of closing option E will be to increase the number taking option B, abortion.

At first glance, according to Klerman, closing off one of the options has to make each of the others more likely to be chosen. In practice, there are likely to be interactions among the choices. Much depends, Klerman argues, on the consistency over time and foresight that women have (and on the degree to which conceptions can be planned). For example, women considering having sex might be less likely to do so if welfare benefits are reduced, leading to fewer pregnancies and therefore fewer abortions. In general, welfare reforms need not lead to an increase in abortions if women can take steps before conception to avoid the abortion/welfare choice that would face them during pregnancy.

Since more than 20 states have implemented family caps under AFDC waivers and more will do so under TANF, the issue could theoretically be decided by observing what happens to the number of births and the number of abortions in states varying in their eligibility rules. But valid data on abortion are hard to come by. Klerman reviewed the major sources of data on abortion and their limitations for research on welfare effects:

• provider surveys (conducted biennially by the Alan Guttmacher Institute),

• household surveys,

• registration systems (in 14 states), and

• surveillance data reported by states to the U.S. Centers for Disease Control and Prevention.

Self-report data from household surveys would be most useful for most analyses (because they are linked to data on individual characteristics), but abortions are severely underreported in such data. The provider surveys show less underreporting but are usable only at the state level, and there are further problems in distinguishing the state of provision from the state of residence.

Klerman reviewed six studies that used state-level data on welfare benefits and abortions to examine their relationship (with various methods for correcting for unobserved factors potentially affecting both abortion rates and AFDC generosity). The studies showed mixed results. Two found that higher AFDC benefits are associated with higher abortion rates, not lower, implying that welfare generosity indeed figures in the decisions (Matthews et al., 1995; Blank et al., 1994). But the size and significance of effects are dependent on model specification, and other well-designed studies have found no effect of AFDC generosity on abortions (e.g., Argus et al., 1997).

Another source of evidence, in addition to the mainly observational studies reviewed by Moffitt and Klerman, is evaluations of demonstration programs for teenage parents already on welfare or at high risk of going on welfare, which were intended in part to affect their fertility behavior. About half of all welfare recipients are current or former teenage parents, and the rates of repeat pregnancies are particularly high for this group (Maynard et al., 1997). Maynard and her colleagues (1997) reviewed the results of seven demonstration projects aimed primarily or exclusively at teenage parents (Table 3). These included employment and training programs (Job Corps and Job Start); comprehensive education and training programs (New Chance and Project Redirection); education and employment programs mandating education and job preparation services for teenage parents on welfare (Ohio Learnfare and the Teenage Welfare Parent Demonstration); and two health-care-focused programs for first-time parents. The only two that had a substantial effect on repeat pregnancy rates were the two health care programs: the Teenage Parent Health Care Demonstration and the Elmira

Home Visiting Demonstration, which showed reductions of 57 and 43 percent, respectively. Pregnancy rates increased among participants in the Job Start and New Chance demonstrations. For New Chance participants, abortion rates rose sufficiently to offset the effect of the higher pregnancy rate, leading to lower birth rates, whereas in Project Redirection and the Teen Welfare Parent Demonstration, the abortion rate declined to such an extent that program rates had higher birth rates, even though repeat pregnancy rates had not increased (Maynard et al., 1997).

The conclusion from evaluations of the education, training, and "workfare" demonstration projects is again ambiguous: they seem to affect pregnancy, abortion, and fertility in different ways, for reasons that we don't fully understand. The apparent success of the health-care-related projects in reducing subsequent pregnancy rates suggests that fertility is most affected by health care interventions, not by labor force interventions.

Jacqueline Darroch pointed out that, in recent years, abortion rates have been declining for all groups. She noted as well that short-term impacts of welfare reform may differ from long-term impacts; for example, marriage and contracep