Overall, the results described here indicate that the welfare-to-work programs implemented as part of the JOBS Program did have significant impacts on children's development, but these impacts were not widespread and were generally small. There were not strong indications of unfavorable impacts on children when their mothers were assigned to participate in a JOBS program, and indeed it is important to note that the significant impacts included favorable impacts on children.
In the aggregate, program impacts on children were not found on many outcomes, and those impacts that were found tended to be small. In addition, the direction of impacts differed for the three aspects of development examined: There were favorable impacts particularly in the area of cognitive development, and particularly in the Atlanta site, while there were unfavorable (albeit generally small) impacts in the area of health, specifically in the Riverside site. Findings in the area of children's behavioral and emotional adjustment encompassed both favorable and unfavorable impacts. Only a single finding at the aggregate level (an unfavorable impact on a rating of child health in Riverside's labor force attachment program ) was of sufficient magnitude to be called "policy relevant."
In terms of subgroup impacts, we acknowledged that children from families at higher risk could show a pattern of unfavorable or of favorable child impacts. However, we hypothesized that children from higher-risk families might experience disproportionately unfavorable impacts resulting from their mothers' assignment to a JOBS program.
As for impacts at the aggregate level, findings at the subgroup level were few and tended generally to be small. The subgroup impact findings identified a pattern of unfavorable impacts, with a number of these large enough to be considered policy relevant, for children from lower-risk families. In particular, such a pattern occurred for children from lower-risk families in three programs: Grand Rapids' labor force attachment program (in which unfavorable and policy relevant impacts were found on behavioral and health outcomes); Riverside's human capital development program (again with unfavorable and policy relevant impacts in the areas of child behavior and health); and in Riverside's labor force attachment program (with unfavorable and policy relevant impacts in all three aspects of development). It is also worth noting that some of these unfavorable and policy relevant impacts may be considered "moderate" to "large" in terms of effect sizes. The fact that the unfavorable and policy relevant impacts occurred in a concentrated manner (for lower-risk families in particular programs), and that some of these were "moderate" to "large" in magnitude, suffices to indicate that we cannot entirely dismiss the possibility of unfavorable impacts. It will be important to continue to track findings for children in lower-risk families assigned to these programs in the five-year follow-up of the Child Outcomes Study.
At the same time, there was a pattern of favorable program impacts for children in higher-risk families whose mothers had been assigned to human capital development programs or to Atlanta's labor force attachment program.(19) These favorable impacts were generally small, however, with none of sufficient magnitude to be called policy relevant. Again this pattern needs to be followed over time. Interestingly, all four of these programs also had favorable impacts on mothers' educational attainment (e.g., receipt of a high school diploma or GED; receipt of a trade degree). We need to determine if, in the longer-term, programs that improve mothers' educational outcomes continue to have beneficial implications for children from higher-risk families.
In sum, most children were not adversely affected and in some cases, children were helped by their mothers' participation in a JOBS program. Yet for delimited subgroups (particularly lower-risk families in both of Riverside's programs and in Grand Rapid's labor force attachment program) there is reason for vigilance. Further follow-up of the children (five years after enrollment) will be important.
Given these findings, we must consider it a possibility that the welfare-to-work programs put in place by the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 and implemented in the context of other policy changes, such as increased child care funding, increased earnings disregards, and Medicaid and SCHIP eligibility expansions, can have impacts on children's development. It is indeed possible that the different obligations and opportunities of the new welfare policy will result in stronger or more pervasive impacts on children than we have found here. Yet we should not begin with an assumption that the newer programs will have uniformly unfavorable or favorable impacts on children. Rather, the present findings suggest that outcomes for children are likely to differ by domain of development, by the specific features of local programs, and by characteristics of the families. The match between type of program and whether the family entered the program at higher or lower risk may continue to prove important. Regarding children, then, the present findings underscore the importance of continuing to monitor the magnitude and direction of program impacts on children's developmental outcomes in the new policy context.
We have learned that welfare-to-work programs implemented under JOBS had the potential to affect multiple aspects of family life, including aspects of family life not explicitly targeted by the programs. Non-experimental statistical analyses indicate that impacts on children sometimes reflected the simultaneous influence of both favorable and unfavorable program effects on families. Further, we have learned that child impacts reflect multiple program influences on families. Pathways through which child impacts appear to have come about in the present study included JOBS programs' impacts on outcomes explicitly targeted by the programs (such as employment and receipt of AFDC) and especially, impacts on further, non-targeted aspects of family life (such as parenting behavior and maternal psychological well-being).
Thus far it appears that the new welfare-to-work policies implemented in the states following the passage of PRWORA, combined with a robust economy and other policies, have contributed to recent changes in employment rates and on receipt of public assistance (Moffitt, 1999). The fact that impacts on children in the present study were linked to such targeted outcomes underscores the possibility that current policies may also be having effects on children. The present set of findings point to the importance of examining not only the effects of current policies on employment, receipt of public assistance, earnings and income, but also on such aspects of family life as parenting behavior and mothers' psychological well-being.
In sum, we have learned that welfare-to-work programs do indeed have the potential to affect children, both favorably and unfavorably. It is important to identify whether, to what extent, and how, impacts on children are occurring in the present policy context. Welfare policies originated in a concern for the well-being of children. Effects on children should continue to be monitored and considered in policy decisions.