Tracking Studies. The issues identified here make clear the need to continue to track the experience of children and families under the new policy. This should be done first through administrative data systems, which in most states are now improving to the point that long-term follow-up of cases will be possible. But administrative data systems are limited in the information they contain, so they must be supplemented by more detailed studies of the course of cases. How many cases are being declared exceptions to the time limits? In how many other cases do the rules not fit? What is the trend in the rate of maltreatment of children returned home? What is the safety record of alternative placements? We know that decisions on permanency are being made more quickly, will that trend be sustained? Do quicker decisions in fact result in real permanency, or is there an increase in the disruption of placements intended to be permanent? Is the trend toward increased adoptions sustained or will the number of adoptions begin to decline once backlogs of children in care for a long time are reduced?
Decision Making Studies. Discussions of the operation of the child welfare system inevitably lead to questions about decision making. The questions are both descriptive and prescriptive: how are child welfare workers, supervisors, administrators, and judges making decisions in the changed policy environment and how should they be making them? Prediction and risk assessment are central here. Most, if not all, child welfare departments make use of risk assessment tools. While there has been a considerable amount of research conducted in connection with these instruments, we need more. The effectiveness of risk assessment procedures is still not known with certainty and undoubtedly can be improved.4 Any decision as to a course of action implies predictions of the likely outcomes of both the action to be taken as well as alternative actions. We need to understand better how good those predictions are.
Beyond this, we need greater understanding about how workers make decisions about cases. Again, there is a fair body of literature on the subject, but we need to understand how workers are deciding in the changed policy context. For each of the decisions they face, determinations of services to be delivered to families, deciding to return children home, deciding on the acceptability of a relative's home, moving to terminate parental rights, deciding on an adoptive family, we need to know the factors that enter into the decision, how workers combine those factors in their minds, and how the agency, community, and policy context affects decisions.
Implementation Studies. Besides leading to better decisions, decision making studies probe the ways in which policies are working or not working. Laws and regulations governing the actions of officials are rarely implemented in exactly the ways intended. Street-level bureaucrats shape, and sometimes even subvert, the intentions of policy makers in myriad ways. Sometimes this is a matter of resistance to change, the desire to continue to operate in familiar ways, using skills that are well ingrained. Sometimes it is a matter of inadequate training on new requirements and new approaches. Sometimes it is a matter of adapting general rules to local situations. And sometimes subversion is used to gain more flexibility in order to better serve clients. We need studies of how workers are implementing the requirements of new laws and regulations.
Implementation of Concurrent Planning. One of the central features of the new approaches is the concept of concurrent planning. Although the initial case goal may be return home, workers are required to develop alternatives to that goal from the beginning, usually by exploring the possibility of adoption. This requirement is intended to save time later on, in case the original plan does not work out (the threat of an alternative to return home may also put beneficial pressure on parents to change), but it creates conflicts for workers as well. Workers must be working on two, conflicting plans. Sometimes, as in the case of an older adolescent, an alternative plan such as adoption is just not realistic. Our exploration of current reunification practice suggests that the concurrent plan requirement is interpreted by workers as requiring that adoption or guardianship by a relative or foster parent be explored. "Stranger adoption" (i.e., adoption by an unknown adult) is often not included in concurrent planning. Some workers, contrary to policy, do not begin to work seriously on alternatives until close to the time of administrative or judicial review, when it appears the family is not ready for the return of the child. We need greater understanding of how the concurrent planning requirement is being implemented on the ground and the ways in which it may be avoided or subverted.
The Time Dilemma. The time dilemma facing workers is often put in terms of the conflict between the child's sense of time and the parent's sense of time. For children, a year in placement is a long time, a time in which new relationships may be formed and old ones fade. It is also a time of uncertainty as to the future and it may be a time in which stable, loving relationships are not there to further growth. But parental change does not happen over night, it takes time to overcome old habits, to learn new ways of functioning and particularly of ways to relate to children. The dilemma is particularly acute in cases of parental substance abuse, the treatment of which is characterized by ups and downs, sometimes repeatedly. How many relapses should be allowed? How much time should be allowed for parents to overcome their addictions? Although there apparently has been some waning of the drug problem in our central cities, it remains a central problem facing families in the child welfare system. The field would benefit from greater understanding of the course of treatment of addictions and from better capacity to predict the likely outcome of drug and alcohol treatment.
4. Determining the effectiveness of risk assessment instruments in a definitive way is quite difficult. Ideally, one would want a design in which cases with similar levels of risk (or predictions) are randomly assigned to alternative courses of action.