6Changing Family Formation Behavior Through Welfare Reform

Family Cap Rules

Generally the family cap provisions under the waivers in effect at the time PRWORA was passed contained the stipulation that an AFDC family would not receive additional cash assistance for those children born while the family was receiving AFDC benefits. Under PRWORA, states have the option of imposing family caps, but are not required to do so.

By the time PRWORA passed, 21 states had obtained federally approved waivers containing family cap provisions (Table 6-3). States were given the flexibility to modify existing waiver terms and conditions under PRWORA, though many chose not to do so, at least in the short run.

The New Jersey family cap, implemented in 1992 as a provision of the state's Family Development Program, has been the most widely publicized such provision and became somewhat of a model for other states. Under it, AFDC families received no assistance for a child born more than 10 months after AFDC application (there also was a 10-month grace period for current recipients), with three exceptions: (1) the first child of a minor already included in an AFDC grant; (2) families who left AFDC due to earnings, remained employed for 90 days, then terminated employment for good cause; and (3) families who left AFDC for any reason and remained off welfare for 12 consecutive months. New Jersey also offered an additional earned income disregard for employed mothers with infants who are subject to the cap, allowing them to keep more of their earnings.

Family cap initiatives were characterized by two features during the pre-PRWORA period. First, as in New Jersey, the "typical" family cap provision was not initiated as a stand-alone provision, but was embedded within larger, more complex welfare reform initiatives.⁶ Seventeen of the twenty-one states with family caps overlap with states that obtained waivers altering their two-parent family provisions (see further below). Second, there existed considerable variation in the design and execution of the provisions. Although most states excluded from the cap children born within 10 months of application, two states (California and Illinois) had 24-month exclusions. Most exempted children who were conceived through incest or rape. Eight states (Arkansas, Connecticut, Delaware, Indiana, Nebraska, New Jersey, Tennessee, and Wisconsin) exempted the first-born children of minors who were already receiving benefits as a child of a welfare recipient. Six states (Arizona, Connecticut, Delaware, Indiana, New Jersey, and Wisconsin) exempted children conceived while the parent was not receiving benefits.

There were also variations by the recipient's age and the birth order of the children. The Kansas waiver permitted a second child to receive benefits; the third child received 50 percent of the benefits normally allotted for an additional dependent; and subsequent children received no benefits. Wisconsin's Parental and Family Responsibility Initiative (which included only recipients under the age of 20 at the beginning of the program) provided 50 percent of the otherwise available benefits for a second child, and no benefits for a third or subsequent child (Center for Law and Social Policy, 1995).

Some states accompanied the family cap provisions with exemptions that changed the treatment of family earnings for women with infants and/or women who received child support. For example, Arizona, Massachusetts, and New Jersey liberalized their treatment of earnings for women with infants. California, Virginia, and Florida exempted child support from countable income, while Nebraska and Delaware discounted child support in computing countable income.

Unemployed Parent Rules

Thirty-three states and the District of Columbia obtained waivers to eliminate some or all of the eligibility restrictions on two-parent families (Table 6-4). Under the AFDC-UP program, three primary provisions restricted the benefit eligibility of two-parent families. Under the 100-hour rule, an AFDC-UP family lost eligibility if the principal wage earner was employed for 100 hours or more a month, even if the family's wages from employment were so low that the family would still be eligible for AFDC. Under the work history requirement, an AFDC-UP family had to demonstrate a "connection to the work force" in one of two ways—the principal wage earner must have received or been eligible for unemployment compensation benefits or must have had $50 or more of earnings in at least 6 of 13 quarters ending within a year before applying for AFDC. Under the 30-day waiting period requirement, and AFDC-UP family could not receive aid for 30 days after the principal wage earner became unemployed (Center for Law and Social Policy, 1995).

All of the states (including the District of Columbia) with waivers related to the unemployed parent provisions eliminated the 100-hour rule, and 25 states requested, and most received, waivers eliminating both the 100-hour rule and the work history requirements. Eighteen states eliminated all three rules. Eliminating these rules allowed working fathers to remain with their families and report their actual hours of work without loss of benefits to the families.

As with the family cap rules, some states obtained multiple waivers with similar provisions. For example, Wisconsin obtained three sets of waivers containing provisions dealing with two-parent families, all with somewhat different combinations of rules. Both its Work Not Welfare and Pay for Performance demonstrations eliminated only the 100-hour rule along with other welfare policy changes (time limits in the former case and changes in work requirements and supports in the latter). The Parental and Family Responsibility Initiative, in contrast, eliminated all three of the unemployed parent rules.

The majority of states simply eliminated the three specific rules, but a few made more complex policy changes. Florida's Family Transition Program eliminated all three rules, but also provided case-by-case exceptions to the 6-month time limit, during which the family's principal wage earner was supposed to complete an employability plan. Missouri's demonstration, the Missouri Families—Mutual Responsibility Plan, Phase I, eliminated the 100-hour rule and work history requirements only for families in which both parents were under 21 years of age. Hawaii's Families Are Better Together based its deprivation criteria on need rather than the unemployment of a parent and eliminated the 100-hour and 30-day rules and the work history requirements.

Family Caps

The economic arguments that support instituting family caps as a means of reducing fertility among welfare recipients are quite clear. Relative to current policy, welfare recipients who bear additional children will be poorer. In economic terms, recipients will face a higher net economic cost of bearing additional children than under the former welfare policy—a cost equal to the benefit increments for additional children that were paid under the former policy. Economic theory predicts that, ceteris paribus, increasing the net cost of childbearing will encourage some to forgo additional childbearing. An extension of this reasoning would also predict that, at the margin, some women who are deciding whether to have a child at all and become dependent on welfare might postpone childbearing until they are able to support the child themselves. But, insofar as the caps apply to second and later children, childless women may substantially discount the nominal ''cost" of the new policy in their current childbearing decisions.

In practice, it seems unlikely that the change in economic incentives associated with family caps will induce sizable changes in fertility patterns, for two reasons. First, the policy-induced changes in income associated with the future fertility decisions of welfare recipients tend to be relatively small—both in proportion to the cost of childrearing and in absolute terms.⁷ Their small absolute size is partly a function of earlier benefit-increment policies, under which the monthly increase in benefits for additional children most often was quite nominal. Moreover, increases in Food Stamp and housing benefits typically will offset portions of the welfare benefit reduction, and Medicaid benefits are unaffected by family caps.

Second, economic factors are, at best, weakly related to fertility decisions, in large part, we suspect, because of strong mediating influences of noneconomic factors. Nationally, family sizes tend to be inversely correlated with income. Moreover, three-fourths of pregnancies to women who have incomes below the poverty line are unintended, as are 88 percent of the pregnancies to women who have never been married (Brown and Eisenberg, 1995). Twenty-eight percent of births among never-married women, who are heavily represented among welfare recipients, are the result of unintended pregnancies (Brown and Eisenberg, 1995; Kost and Forrest, 1995). This is consistent with the fact that over 70 percent of low-income women who report not wanting to become pregnant also report using some form of contraception.⁸

Advocates of family caps do not rely solely on the expectation of behavioral responses to support their views. The policy discussions about family caps also often invoke principles and values in addition to social and economic theories and research. For example, many of the arguments of legislators in New Jersey and Wisconsin for both the family caps and the suspension of the 100-hour rule were value based. It seemed to matter less that marriage rates rose and divorce and fertility rates fell as a result of the policy changes than it did that everyone understood that society favors marriage and is not willing to reward those having additional children when they cannot adequately support the ones they have already.

Unemployed Parent Provisions

Altering the unemployed parent provisions affects foru groups of welfare recipients: (1) those with spouses working in low-wage jobs; (2) those living with spouses who are not working but who could work in low-wage jobs; (3) those with nonresident partners working in low-wage jobs for more than 100 hours per month; and (4) those at the margin of a marital decision. It also could affect those nonrecipients who, on moral or other grounds, have decided to marry and not participate in welfare, even though their spouses have limited earnings potential.

Eliminating the 100-hour rule means that, at the margin, some wage earners in two-parent families may increase their work effort beyond their current level, thereby increasing family welfare at the same time that reliance on public assistance declines. Some unemployed spouses of welfare recipients may be enticed to seek employment for similar reasons and with similar results. Some couples who previously bypassed marriage or divorced to preserve their welfare eligibility may elect to get or stay married, because the financial loss from doing so is diminished. On the other hand, some nonrecipients could be attracted to welfare as a result of the policy change, because welfare would offer an income supplement without jeopardizing the marriage option.

Suspension of the 100-hour rule sends two quite clear messages to recipients and nonrecipients: (1) work is valued and (2) so is marriage. Eliminating the "work test" and "waiting periods" is expected to have similar, though somewhat more complicated, effects.

Fertility Decisions

The decision making that leads to family formation and expansion is very complex. Major differences in fertility rates are associated with income class, cultural identity, age, and marital status. These relationships have been changing rapidly over time. Fertility rates are higher among low-income families than among middle- and higher-income families; they are higher among blacks and Hispanics than among whites and Asians; the mean age of first birth has risen to about 23 and the variance has been increasing; and roughly 30 percent of all births are out-of-wedlock, but vastly higher proportions of poor women and teenagers are unmarried when they give birth (U.S. Department of Health and Human Services, 1995).

Several studies that have examined the welfare incentives for childbearing have found a slight association between AFDC and increases in childbearing, but this association is not characteristic of all recipients (Plotnick, 1990; Lundberg and Plotnick, 1990; Gottschalk, 1990, 1992; Moffitt, 1992; Haveman and Wolfe, 1994; Plotnick and Lundberg, 1995). In fact, in 1995 Moffitt argued that we do not know enough to characterize these findings as evidence of an actual effect of AFDC on childbearing (Moffitt, 1995; but see Moffitt, Chapter 4, in this volume). Moreover, the research suggests that impacts on fertility decisions may have been mediated by funding for family planning.

Federal support for family planning and reproductive health issues offers real returns to taxpayers from a variety of sources, most notably lower rates of infant and neonatal death, lower fertility rates among low-income women who are at high risk of tightly spaced births, and fewer low-birthweight babies (Meier and McFarlane, 1994). Although there is limited evidence that these funds reduce teenage pregnancy rates (Haveman et al. 1997), they clearly reduce teenage birth rates through increasing abortions (Olsen and Weed, 1986; Moore et al., 1994).

The behavioral research on the effects of welfare on fertility is complemented by a limited body of demonstration research that tests various changes in welfare and social service policies designed to reduce fertility among teenage parents already on welfare or at high risk of going onto welfare. The eight such programs with strong evaluations represent four different strategies for addressing the needs of teenage parents, albeit with overlapping features: (1) employment and training programs serving significant numbers of teenage parents (Job Corps and Job Start); (2) comprehensive education and training programs targeted specifically at teen parents (New Chance and Project Redirection); (3) welfare-based education and employment programs that mandate education and job preparation services for teenage parent welfare recipients (Ohio Learnfare and the Teenage Parent Welfare Demonstration); and (4) health-focused programs targeted at first-time parents, many or all of whom are teenagers (the Teenage Parent Health Care Program and the Elmira Nurse Home Visiting program). The table in the appendix to this chapter provides references for the evaluations of these programs.

These demonstrations are relevant insofar as roughly half of all welfare recipients are current or former teenage parents, and the rates of repeat pregnancies and births are especially high among this group (Jacobson and Maynard, 1995). Recognizing that subsequent pregnancies and births would interfere with the central education and employment goals of these programs, all of the demonstration programs except the two youth employment programs devoted considerable resources to helping the young mothers control their fertility, offering family planning services and counseling as part of the basic intervention.

Still, repeat pregnancy rates were high—above 50 percent—in all programs except the teenage parent health care program (28 percent over 18 months). Only the two health-focused programs were successful in reducing the repeat pregnancy rates (Table 6-5). The Teenage Parent Health Care Demonstration, which involved home visitor services by trained social workers, reduced the repeat pregnancy rate by an estimated 57 percent, and the Elmira Home Visiting Demonstration reduced the rate by 43 percent.

Pregnancy rates increased significantly among participants in both the Job Start and the New Chance demonstrations (by 13 percent and 8 percent respectively). One can only speculate as to why these programs would increase pregnancy rates. However, one possible explanation includes their role in increasing opportunities for the young women to meet men, their promotion of higher self-esteem, and their facilitation of independent living, while failing to improve contraceptive practices. Among New Chance participants, the abortion rate also increased sufficiently to offset the higher pregnancy rate, whereas in Project Redirection and the Teenage Parent Welfare Demonstrations, the abortion rate declined sufficiently to show increases in the birth rates among program participants, even though the repeat pregnancy rates had not increased significantly. These differential patterns of abortion likely reflect differences in the philosophies of program staff and in access to contraceptive services. The Teenage Parent Welfare Demonstration, for example, had an explicit policy of not making referrals for abortion.

The upcoming evaluation of the Job Opportunities and Basic Skills Training program, which is the employment support program introduced with the Family Support Act of 1988, will include measures of fertility effects. However, none of the intervening evaluations of welfare reforms for adult recipients has measured such effects, in part because fertility control was not a major goal of the interventions. Most of these evaluations relied on limited data collected from state welfare and Unemployment Insurance data systems.

Family Formation, Dissolution, and Single Parenthood

The overwhelming message from this research is that the various welfare reforms are likely to have modest impacts on family formation and fertility (see Chapter 4). In several studies using national databases, researchers have found weak or no relationships between the generosity of welfare benefits and marriage, divorce, and out-of-wedlock childbearing (Ellwood and Bane, 1985; Lundberg and Plotnick, 1990; Haveman and Wolfe, 1994; McLanahan and Sandefur, 1994; Moffitt, 1995; Plotnick and Lundberg, 1995). The effects that were observed are small in relation to the quite dramatic decreases in marriage, increases in divorce, and skyrocketing rates of out-of-wedlock childbearing observed in society (Moffitt, 1992; Brown and Eisenberg, 1995; Lichter, 1995). Given that the research has relied on analysis of time trends and cross-site variation in state-level data, we are cautioned that a "major ambiguity in the conclusion … is whether any observed relationship between welfare and nonmarital childbearing is real … or reflects the effects of other unmeasured cross-state differences in … environments. … [W]e are left with only the suggestion of an effect" (Moffitt, 1995:172).

Some researchers have begun to look at these issues with national longitudinal datasets, from different aspects. One study examines whether patterns of childbearing and marriage affect the earnings potential and general economic well-being of families (Brien and Willis, 1997). Finding relatively modest effects, the researchers speculate that one reason is that the current welfare system discourages men from marrying the mothers of their children.

An emerging literature suggests that welfare-dependent mothers, particularly young black mothers, are not especially interested in marrying the fathers of their children or, in most cases, even in pursuing a formal child support order (Polit, 1992; Maynard, 1993; Quint et al., 1994b). Historically, there has been little or no financial gain to formal child support for recipients of public assistance, which accounts for the father's diminished willingness to contribute formal support when under order to make regular payments. Edin and Lein (1997) report that many young unwed mothers receive informal child support from the fathers of their children. In these studies, young women gave various reasons for their lack of interest in marriage, which can be summarized as follows: Men are not reliable sources of economic and emotional support; welfare can at least be relied upon for economic support.

The best available evidence from the research suggests that rule changes may only minimally affect the likelihood that mothers in AFDC-UP families will end up as single parents. The experimental research on family formation and dissolution is limited to that from the negative income tax experiments of the late 1960s and 1970s. The somewhat controversial results from these demonstrations suggest that more generous income support offered under the experiments had, at most, modest impacts on family formation and stability (Cain, 1987; Hannan and Tuma, 1990; Cain and Wissoker, 1990). Using longitudinal state-level data, another study found no evidence that the AFDC-UP program destabilized two-parent families, and found some evidence suggesting that the program may have encouraged family stability (U.S. General Accounting Office, 1992). None of the welfare reform demonstration evaluations of the 1980s focusing specifically on the AFDC population has examined the effects on this set of outcomes, even though many of them included program provisions that, theoretically, could have affected family formation, stability, and fertility decisions. Moreover, none of the early evaluations of demonstrations coming out of the Family Support Act of 1988 focused on the family formation outcomes.⁹

We do know from the many welfare reform demonstrations of the 1980s, as well as from evaluations of many other employment and training programs targeted at disadvantaged young men, that they have modest or no effects on work behavior and earnings. The Job Training Partnership Act programs led to modest, but statistically significant, gains in employment and earnings of disadvantaged men (Bloom et al., 1994). The subsidized employment demonstrations of the 1980s were only modestly successful in getting male AFDC recipients into low-wage jobs, but at high cost (Brock et al., 1993). And the mandatory welfare-to-work programs of the past decade, for example California's welfare reform demonstrations, have led to relatively modest changes in the employment behavior of men (Gueron and Pauly, 1991; Friedlander and Burtless, 1994).

We might infer from the modest changes in employment behavior that these policies have had little to no effect on decisions about family formation and dissolution. However, few of these studies have looked specifically at such outcomes. Two of the evaluations that have addressed them focused on teenage parents and service-oriented interventions rather than financial incentives for working—the Teenage Parent Welfare Demonstration and the New Chance demonstration. In both cases, the programs had no significant impact on family structure over a 2to 3-year follow-up period (Maynard, 1993; Quint et al., 1994a, 1994b).

Evaluation of New Jersey's Family Development Program

The preliminary results of the New Jersey Family Development Program illustrate both the complexities in carrying out meaningful evaluations of welfare reform and the limitations of findings produced even under the best of circumstances. Beginning in August 1993, most of the state's welfare applicants and recipients were converted to the new welfare policies and services, including mandatory work requirements, expanded job training, liberalized eligibility for two-parent families, the family cap, and decreased taxes on earnings for those having additional children while on welfare. For evaluation purposes, several thousand applicants and recipients in selected counties were randomly assigned to remain on the old welfare policies and thereby serve as a control group or counterfactual. This control group is being compared with about 6,000 randomly selected welfare applicants and recipients participating in the Family Development Program.

For several reasons, however, the emerging results of the evaluations are not especially useful for guiding either state or national policy (O'Neill, 1994; Camasso, 1995; Myers, 1995; Donovan, 1995). First, they reflect the effects of a complex set of reforms, only one piece of which pertains to the treatment of additional children born to welfare recipients. Economic and social theories predict that changes in economic opportunities and net wages might affect fertility behavior as much as the reform-induced changes in welfare benefits and economic needs resulting from the birth of an additional child. Neither of the two evaluations conducted was designed to disentangle the impacts of these various elements.

Second, there was tremendous publicity concerning the development and introduction of the family cap. Not only was New Jersey the first state to implement the policy, but it also was the first state to take a public stand on the issue of whether welfare ought to hold families financially harmless for their decisions to bear additional children. As a result, many control group members believe they have been subject to the family cap provision, even though they have not. In the parlance of the evaluation literature, the control group may have been subjected to contamination bias.¹¹

Third, the actual financial consequence of the policy change for families having additional children was quite small. This is because Food Stamp benefits were increased to account for the family size change. Because there was no increase in cash benefits for families having another child under the cap, their Food Stamp benefits increased sufficiently to fill half to two-thirds of the lower cash assistance resulting from the new family cap policy relative to the old AFDC benefit rules.

Fourth, there are problems with the measurement of outcomes in the New Jersey demonstration. The primary source of data on births has been the welfare records. Yet, those whose benefits are not increased as a result of a birth (those subject to the cap) are less likely to report the birth to welfare officials than are those in the control group, for whom the reporting of a birth will trigger a benefit increase.¹²

Finally, the evaluation design does not lend itself to the measurement of any entry or exit effects. That is, it will not measure any deterrent effect, in the form of either delayed parenting or increased abortion rates. Nor will it measure any possible program entry effects due, for example, to the more generous education and training services.

Similar Complexities in Other State Demonstration Evaluations

The prospects that state evaluations will sort out empirically any causal links between selected welfare reforms and family outcomes of interest are not favorable, for a variety of reasons:

• There are very few tests of single policy changes or even of a limited number of changes. As is clear from Table 6-6 and would be equally evident from a similar table for the unemployed parent reforms, there are very few pure tests of specific policy changes. For example, there was initial hope that the unemployed parent reforms in several states (California, Utah, and Wisconsin) would provide clear tests of the underlying economic theories. But closer examination of the project and evaluation designs tempered this optimism: ''All three experiments with the 100-hour rule are flawed, and problems with both design and implementation reduce the usefulness of the experiment" (Birnbaum and Wiseman, 1996). Utah's 100-hour rule demonstration, implemented in 1992, initially eliminated only the 100-hour rule. In 1995, however, the demonstration was incorporated into the state's Single Parent Employment Demonstration Project (SPEDP), which contains at least 11 other major provisions, making it increasingly difficult to segregate the individual component effects.
• There is increased churning and volatility within states. For example, family caps have been introduced in Wisconsin in three different waivers. At least six states with waivers affecting two-parent provisions have modified their state plans. With the passage of PRWORA, we can expect another round of major reforms and policy changes. These planned variations in policy are sometimes in response to and other times overlaid on changes in the political, social, and economic climate in the states.
• Future support for strong evaluations appears highly uncertain. Under PRWORA, the federal government will not mandate evaluations of new reforms. Rather, it will encourage states to continue expanding their knowledge by providing financial support and sponsoring selected demonstrations. Serious empirical work likely will require investments of state resources or significant contributions from private sources.
• There are very difficult technical and methodological challenges to estimating the behavioral responses to many of the new reforms . For example, the new generation of reforms intends to change community and agency cultures. Classical experiments with random assignment of subjects is inappropriate when the purpose of the reform is to effect community change. But without an experimental design, how does one develop an appropriate counterfactual? How do we design an evaluation whose results can be generalized to other settings with different economic conditions, different complements of employment services, and different ground rules governing welfare participation?

The evaluations of the welfare reform demonstrations can best be thought of as laboratory studies of changes in a particular set of policies, being tested in particular economic, social, and political contexts. As such, they can be enormously valuable to the states sponsoring them as well as to states that share similar social and economic contexts. Moreover, they contribute to the general knowledge base by providing additional data regarding behavioral responses to particular "bundles" of policy changes.

States should be encouraged to continue their evaluation efforts, particularly if they have strong experimental designs in place to capture important aspects of their reforms. At the same time, it is important to capitalize on the great social experiment that is about to unfold as authority and responsibility for welfare policy devolve to the states. States may want to be "free riders," letting others do the empirical work upon which they can build future policy choices. But they must realize that they will bear the full fiscal risk of bad policy decisions and reap the benefits of wise ones, thus raising the stakes associated with formulating policy on the basis of either no research or flawed research.

Predicting Behavioral Responses to Welfare Policies

In the context of the broader mission of supporting extensive simulation results by states, research needs to focus more on the behavioral responses of individuals to specific welfare parameters. We need to know the expected change in the out-of-wedlock birth rate associated with particular income changes and needs resulting from an increase in family size. It also is important to know how marriage rates change as we vary the probability that individuals will be eligible for welfare and, conditional on their being eligible, vary their expected benefits. In both of these examples, we need to take into account the initial characteristics of the individual and other contextual factors. The change in the out-of-wedlock birth rate associated with various benefit increments likely will differ by the number of children one has already, the baseline benefit level (including whether the person is on welfare at all), and other social circumstances.

The particulars of the policy parameters that states have already changed or will modify under TANF are less important in themselves than are the consequences of these changes in terms of welfare eligibility and expected benefit levels for current and prospective welfare recipients. In general, the decisions of current and prospective recipients regarding future employment, family formation, child support, and childbearing are going to be determined less by the policies and policy language than by what actually happens to their family's welfare eligibility status and benefit levels if they make particular choices such as the choice to have another child, to marry the father of their child, to get a job, to cooperate with child support, or to get divorced.

Still, the specific formulation of certain policies itself may send important messages to the public regarding expectations and values, which could affect behavior. A prime example in which the message may be as powerful, or more powerful, than the financial consequences of the policy is the family cap. Family caps send a very clear message that taxpayers look down upon out-of-wedlock childbearing, especially if parents are unable to support their children.

The wave of welfare reforms offers an unparalleled opportunity to learn about behavioral responses to income incentives, economic uncertainly, and social support opportunities. We also can estimate the independent effect, if any, of the manner in which the economic consequences and support opportunities are packaged. For example, imagine that we collect and analyze longitudinal data on decisions that poor or near-poor people make and their consequences in terms of program eligibility and benefits for each welfare jurisdiction and for each year from 1992, when the flurry of welfare reform activity began, through the next 5 years. We could gather such information from a micro-level review of welfare applications and case records.¹³

That is, we could sample records from the welfare files for all those years in all 50 states and in some counties and construct a database that includes, for each case: (1) basic information, such as number of children, age of youngest child marital status, employment status, time on welfare, child support status, area of residence, age, and race/ethnicity; (2) status changes, if any, reflecting major behavioral decisions (for example, the birth of another child, a marriage or divorce, the acquisition or loss of a job, and increasing or decreasing hours of employment); and (3) for those with a status change, the impact of this on their welfare eligibility and benefit levels. The resulting database will allow us to create measures of the expected welfare response a person in situation "X" might face if he or she makes particular decisions regarding family formation and status.

Once we have the hypothetical consequences for welfare eligibility and benefits of various behavioral choices that people make, we can use these data in statistical models designed to measure the strength of the behavioral response by current and prospective welfare recipients to particular welfare policy environments. For example, we can estimate the impact of a particular benefit change (such as the introduction of a family cap) on fertility decisions of individuals with particular baseline characteristics and facing a particular policy context prior to the change. In essence, however, we would translate the family cap into particular eligibility and benefit changes associated with fertility decisions that would result in the particular site and for particular individuals. Similarly, we could estimate the impact of a particular marriage penalty or reward on the likelihood of marriage, given the prevailing baseline context.

Assuming we have sufficient variation from our natural experimental database, we should be able to generalize the results of particular policy shifts to other settings.¹⁴ We could also incorporate into these behavioral models measures of possible "message" effects (for example, the effects of a family cap policy beyond its effect on the marginal benefit level associated with a decision to have another child). Doing this would require gathering data for analysis on the understanding and beliefs of poor and near-poor families as well as service deliverers.

Contextual and Ethnographic Analysis

Now more than ever, it is critical that trained social scientists conduct systematic, in-depth evaluations to further our understanding of the economic and social welfare of highly at-risk families; of the behavioral choices these families face and the decisions they make; and of the family, community, and social services they draw on to meet the challenges faced by those living near or in poverty. The most optimistic world under welfare devolution is likely to be one in which state and local welfare reforms succeed in trimming the welfare rolls by providing greater incentives for those with the greatest social capital and familial support to eschew welfare. It also will undoubtedly fail to meet the income security needs of a portion of the current and prospective recipient pool who simply enter parenthood and/or adulthood without the social capital to escape poverty through their own labor or that of their partners. Even with strong employment support interventions, some poor families will hit the time limits and will risk taking their families to homeless shelters or the streets.

The unfolding social experiment in welfare policy will enable us to identify and learn from these two very important subgroups of the poverty population. If we plan carefully, we can estimate the number and characteristics of those falling into each category. But, more important, we also could learn an enormous amount about the factors that contribute to family resilience under a "tougher" world vis à vis social welfare policy, and why some families need much more than offered by either the past welfare system or that which emerges under TANF. The results of such in-depth studies could provide the foundation for designing preventive and ameliorative policies directed at this the most needy group and at moving other families more quickly through transition dependency.

One subset of such research could build on more routine longitudinal surveys of near-poor and poor families to track their movements on and off welfare; in and out of the labor market; through family formations and dissolutions; and through fertility decisions. Researchers could then carefully select families fitting various profiles of special interest—for example, seemingly low-resource families who succeed under various welfare support settings, their counterparts who "fail," and high-resource families who fail under various welfare support options—and conduct an in-depth study of a sample of such families over a period of time to reconstruct the factors that contributed to varying degrees and in varying combinations to the successes or failures of the families.

Another subset of this strand of research would entail intensive ethnographic research with at-risk families.¹⁵ The focus of these studies will be on understanding the interactions among a broad set of contextual factors in determining behavioral choices of families and their short- and long-run implications. Such research would, for example, contribute to our understanding of the issues that poor and near-poor families consider in deciding whether to subsist on a low-wage job rather than apply for welfare or whether to place their child in poor-quality care in order to avoid welfare time limits. This research could then also follow the families to inform us of how the various decisions these families make play out for themselves and their children.

Targeted Experimental Evaluations

Some areas of reform are so different from past policy and practice that prudence argues for conducting targeted experimental evaluations as part of a phased implementation plan. Despite their limitations in terms of generalizability of findings and their inability to address systemic reform, we should not abandon experimental evaluations. They are unquestionably the best means of judging the efficacy of targeted interventions, particularly in cases where there is not yet sufficient evidence to support universal implementation of the policy.

The strategies for conducting these types of evaluations are well documented, and there are dozens of models to draw on (bell et al., 1995; Gordon et al., 1997; Boruch, 1997). These types of evaluations are especially valuable in cases where policy makers are interested in comparing the status quo with proposed new policies or interventions designed to achieve specific improvements in outcomes. The researchers then design an efficient experimental evaluation with the primary goal of testing the hypothesis that the new policy or intervention does have the intended consequences. To the extent feasible, such studies should also be accompanied by qualitative evaluations of the intervention or policy change and in-depth studies of purposefully selected subsets of the demonstration participants for the purpose of understanding the contents of the so-called black box and, further, of helping to understand and interpret the impact analysis findings and translate them into more general conclusions for the benefit of the broader research and policy community.

References

• Beebout, H., R. Cohen, J. Czajka, J. DiCarlo, M. Dynarski, J. Jacobson, and R. Moffitt 1995. Simulation of Trends in Employment, Welfare, and Related Dynamics (STEWARD). Report submitted to the U.S. Department of Health and Human Services, Assistant Secretary for Planning and Evaluation. Washington, D.C.: Mathematica Policy Research, Inc.
• Bell, S., L. Orr, J. Blomquist, and G. Cain 1995. Program Applicants as a Comparison Group in Evaluating Training Programs: Theory and a Test. Kalamazoo, Mich.: The Upjohn Institute.
• Birnbaum, M., and M. Wiseman 1996. Extending assistance to intact families: State experiments with the 100 hour rule. FOCUS 18(1, Special Issue):38–41.
• Bloom, D., V. Fellerath, D. Long, and R. Wood 1993. LEAP: Interim Findings on a Welfare Initiative to Improve School Attendance Among Teenage Parents. New York: Manpower Demonstration Research Corporation, May.
• Bloom, H., L. Orr, G. Cave, S. Bell, F. Doolittle, and W. Lin 1994. The National JTPA Study: Overview of the Impacts, Benefits, and Cost of Title IIA. Bethesda, Md.: Abt Associates, January.
• Boruch, R. 1997. Randomized Experiments for Planning and Evaluation: A Practical Guide. Thousand Oaks, Calif.: Sage Publications.
• Brien, M., and R. Willis 1997. Costs and consequences for the fathers. In R. Maynard, editor. , ed., Kids Having Kids: Costs and Social Consequences of Teenage Pregnancy. Washington, D.C.: Urban Institute Press.
• Brock, T., D. Butler, and D. Long 1993. Unpaid Work Experience for Welfare Recipients: Findings and Lessons from MDRC Research. New York: Manpower Demonstration Research Corporation, September.
• Brown, S., editor; , and L. Eisenberg, editor. , eds. 1995. The Best Intentions: Unintended Pregnancy and the Well-Being of Children and Families. Washington, D.C.: National Academy Press. [PubMed: 25121228]
• Cain, G. 1987. Negative income tax experiments and the issues of marital stability and family composition. In Alicia Munnell, editor. , ed., Lessons from the Negative Income Tax Experiments. Boston, Mass.: Federal Reserve Bank of Boston.
• Cain, G., and D. Wissoker 1990. A reanalysis of marital stability in the Seattle-Denver income-maintenance experiment. American Journal of Sociology 95(5):1235–1269.
• Camasso, M.J. 1995. New Jersey's evaluation. In Addressing Illegitimacy: Welfare Reform Options for Congress. Washington, D.C.: American Enterprise Institute, September.
• Cave, G., H. Bos, F. Doolittle, and C. Toussaint 1993. JOB START: Final Report on a Program for School Dropouts. New York: Manpower Demonstration Research Corporation, October.
• Center for Law and Social Policy 1995. CLASP Guide to Welfare Waivers: 1992–1995. Washington, D.C.: Center for Law and Social Policy, May.
• Donovan, P. 1995. The 'family cap': A popular but unproven method of welfare reform. Family Planning Perspectives 27(4):166–171. [PubMed: 7589359]
• Edin, K., and L. Lein 1997. Making Ends Meet: How Single Mothers Survive Welfare and Low-Wage Work. New York: Russell Sage Foundation.
• Ellwood, D., and M. Bane 1985. The impact of AFDC on family structure and living arrangements. In R. Ehrenberg, editor. , ed. Research in Labor Economics. Greenwich, Conn.: JAI Press.
• Friedlander, D., and G. Burtless 1994. Five Years After: The Long-Term Effects of Welfare-to-Work Programs. New York: Russell Sage Foundation.
• Gordon, A., J. Jacobson, and T. Fraker 1997. How to Evaluate Welfare Reform: Guidance for States. Princeton, N.J.: Mathematica Policy Research, Inc., March.
• Gottschalk, P. 1990. AFDC participation across generations. American Economic Review 2(May):367–371.
• 1992. Is the correlation between AFDC participation across generations spurious? Department of Economics, Boston College.
• Gueron, J.M., and E. Pauly 1991. From Welfare to Work. New York: Russell Sage Foundation.
• Hannan, M., and N. Tuma 1990. A reassessment of the effect of income maintenance on marital dissolution in the Seattle and Denver experiment. American Journal of Sociology 95(5):1270–1298.
• Harwood, J. 1997. Wall Street Journal (January 30):1.
• Haveman, R., and B. Wolfe 1994. Succeeding Generations: On the Effects of Investments in Children. New York: Russell Sage Foundation.
• Haveman, R., B. Wolfe, and E. Peterson 1997. Children of early childbearers as young adults. Pp. 257–284 in R. Maynard, editor. , ed., Kids Having Kids: Economic Costs and Social Consequences of Teen Pregnancy. Washington, D.C.: Urban Institute Press.
• Herr, T., S. Wagner, and R. Halpern 1996. Making the Shoe Fit: Creating a Work-Prep System for a Large and Diverse Welfare Population. Chicago, Ill.: Project Match, Erikson Institute.
• Hershey, Alan, and M. Silverberg 1993. Program Cost of the Teenage Parent Demonstration. Princeton, N.J.: Mathematica Policy Research, Inc.
• Jacobson, J. and R. Maynard 1995. Unwed mothers and long-term welfare dependency. In Addressing Illegitimacy: Welfare Reform Options for Congress. Washington, D.C.: American Enterprise Institute, September.
• Kost, K., and J. Forrest 1995. Intention status of U.S. births in 1988: Differences by mothers' socioeconomic and demographic characteristics. Family Planning Perspectives 27(1):23–27. [PubMed: 7720847]
• Lichter, D. 1995. The retreat from marriage and the rise in nonmarital fertility. In Report to Congress on Out-of-Wedlock Childbearing. U.S. Department of Health and Human Services, National Center for Health Statistics. Washington, D.C.: U.S. Government Printing Office.
• Long, D., J.M. Gueron, R.G. Wood, R. Fisher, and V. Fellerath 1996. LEAP: Three-Year Impacts of Ohio's Welfare Initiative to Improve School Attendance Among Teenage Parents. New York: Manpower Demonstration Research Corporation, April.
• Luker, K. 1996. Dubious Conceptions: The Politics of Teenage Pregnancy. Cambridge, Mass.: Harvard University Press.
• Lundberg, S., and R. Plotnick 1990. Effects of state welfare, abortion and family planning policies on premarital childbearing among white adolescents. Family Planning Perspectives 22:246–251. [PubMed: 2289541]
• Maynard, R., editor. , ed. 1993. Building Self-Sufficiency Among Welfare-Dependent Teenage Parents. Princeton, N.J.: Mathematica Policy Research, Inc., June.
• Maynard, R. 1997. The role for paternalism in teen pregnancy prevention and teen parent services. In L. Mead, editor. , ed., The New Paternalism. Washington, D.C.: Brookings Institution.
• Maynard, R., and A. Rangarajan 1994. Contraceptive use and repeat pregnancies among welfare-dependent teenage mothers. Family Planning Perspectives 26(5):198–205.
• Maynard, R., W. Nicholson, and A. Rangarajan 1993. Breaking the Cycle of Poverty: The Effectiveness of Mandatory Services for Welfare-Dependent Teenage Parents. Princeton, N.J.: Mathematica Policy Research, Inc., December.
• McLanahan, S., and G. Sandefur 1994. Growing Up with a Single Parent. Cambridge, Mass.: Harvard University Press.
• Meier, K.J., and D.R. McFarlane 1994. State family planning and abortion expenditures: Their effect on public health. American Journal of Public Health 84:1468–1472. [PMC free article: PMC1615165] [PubMed: 8092373]
• Moffitt, R. 1992. Incentive effects of the U.S. welfare system: A review. Journal of Economic Literature 30:1–61.
• 1995. The effect of the welfare system on nonmarital childbearing. In Report to Congress on Out-of-Wedlock Childbearing. U.S. Department of Health and Human Services, National Center for Health Statistics. Washington, D.C.: U.S. Government Printing Office.
• Moore, K., C. Blumenthal, B. Sugland, B. Hyatt, N. Snyder, and D. Morrison 1994. State Variation in Rates of Adolescent Pregnancy and Childbearing. Washington, D.C.: Child Trends, Inc.
• Myers, R. 1995. New Jersey's 'family cap.' In Addressing Illegitimacy: Welfare Reform Options for Congress. Washington, D.C.: American Enterprise Institute, September.
• New Jersey Department of Human Services, Office of Public Affairs 1995. News release, May 16.
• Olds, D., C. Henderson, Jr., and R. Tatelbaum 1988. Improving the life-course development of socially disadvantaged mothers: A randomized trial of nurse home visitation. American Journal of Public Health 78(11):1436–1145. [PMC free article: PMC1350235] [PubMed: 3052116]
• Olsen, J., and S. Weed 1986. Effects of family-planning programs for teenagers on adolescent birth and pregnancy rates. Family Planning Perspectives 20:153–170.
• O'Neill, J. 1994. Expert testimony in C.K. vs. Shalala, USDC, D. NJ, Civil Action No. 93–5354.
• O'Sullivan, A., and B. Jacobsen 1992. A randomized trial of a health care program for first-time adolescent mothers and their infants. Nursing Research 41(4):210–215. [PubMed: 1408861]
• Plotnick, R. 1990. Welfare and out-of-wedlock childbearing: Evidence from the 1980s. Journal of Marriage and the Family 52:735–746.
• Plotnick, R., and S. Lundberg 1995. Adolescent and premarital childbearing: Do economic incentives Matter? Journal of Labor Economics 13:177–200. [PubMed: 12291238]
• Polit, D. 1992. Barriers to Self-Sufficiency and Avenues to Success Among Teenage Mothers. Princeton, N.J.: Mathematica Policy Research, Inc.
• Quint, J., D. Polit, H. Bos, and G. Cave 1994. a. New Chance: Interim Findings on a Comprehensive Program for Disadvantaged Young Mothers and Their Children. New York: Manpower Demonstration Research Corporation, September.
• Quint, J., J. Musick, and J. Ladner 1994. b. Lives of Promise, Lives of Pain: Young Mothers After New Chance. New York: Manpower Demonstration Research Corporation, January.
• U.S. Department of Health and Human Services, Administration for Children and Families Various Terms and conditions of waiver provisions for [state]. years.
• U.S. Department of Health and Human Services 1995. Out-of-Wedlock Childbearing in the U.S. Washington, D.C.: U.S. Government Printing Office.
• 1997. Setting the Baseline: A Report on State Welfare Waivers. June. Washington, D.C.: U.S. Government Printing Office. Http://aspe​.os.dhhs.gov/hsp/isp/waiver2.
• U.S. General Accounting Office 1992. Unemployed Parents: An Evaluation of the Effects of Welfare Benefits on Family Stability. GAO/PRMD-92-19BR. Washington, D.C.: U.S. Government Printing Office.
• U.S. House of Representatives, Committee on Ways and Means 1993. 1993 Green Book: Background Material and Data on Programs Within the Jurisdiction of the Committee on Ways and Means. WMPC-103-18. Washington, D.C.: U.S. Government Printing Office.
• 1996. a. 1996 Green Book: Background Material and Data on Programs Within the Jurisdiction of the Committee on Ways and Means. Washington, D.C.: U.S. Government Printing Office.
• 1996. b. Personal responsibility and Work Opportunity Reconciliation Act of 1996: Conference Report to accompany H.R. 3734. Section 401, Purpose (p. 9), July 30.