Integrating Federal Statistics on Children: Report of a Workshop (1995)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Workshop Summary The Committee on National Statistics and the Board on Children and Families of the National Research Council and the Institute of Medicine held a workshop on March 31 and April 1, 1994, to examine the adequacy of federal statistics on children and families. Concurrent with increased interest in monitoring and understanding the lives of children and families is a recognition of serious shortcomings in federal and other major longitu- dinal data used by policy makers to inform their work. Workshop partici- pants, including staff of the statistical agencies that design and implement surveys, researchers who analyze statistical data, and policy experts who make use of these data to address pressing issues affecting children and families, discussed the following issues: • What are the most pressing information needs of those who formu- late, implement, and analyze policies for children and families? • What are the strengths and shortcomings of existing and proposed federal statistical data sources for addressing these information needs? To what extent do existing sampling strategies follow children across critical transition points? Are there populations of children, such as immigrant children, for whom adequate data do not exist? To what extent is it pos- sible to combine data to examine relations across domains of development, such as school achievement, health, and criminal behavior? • What are the most promising strategies for improving the capacity of the federal statistical system to address these needs? Should the focus be 9 

10 INTEGRATING FEDERAL STATISTICS ON CHILDREN on enhancing the integration of existing data sets or developing a national survey of children? Over the course of the workshop, participants considered these issues against the framework of key developmental transition points in children’s lives from birth through preparation for adulthood. Using the presentations by the authors of five background papers as a basis, the participants dis- cussed: (1) child development in the context of family and community re- sources; (2) children’s transitions into school; (3) federal data on educa- tional attainment and the transition to work; (4) the data needs for monitoring health care reform for children and families; and (5) estimating the inci- dence, causes, and consequences of interpersonal violence for children and families. CHILD DEVELOPMENT IN THE CONTEXT OF FAMILY AND COMMUNITY RESOURCES Jeanne Brooks-Gunn, Brett Brown, Greg Duncan, and Kristin Moore addressed the issue of improving national data to facilitate policy making for children and youth. Datasets must have reliable and, if possible, longi- tudinal assessments of child outcomes and measures should be age- and development-specific, the authors note. Sets of child supplements offer one approach to capturing such information. Assessments should cover as many of the crucial domains of child and adolescent development as possible, given a limited interview period. It may be best to link the timing of assessments to what is known regarding time frames of stability and change, e.g., family income. Since resources, broadly conceived, are instrumental in promoting or retarding development, a high-quality, longitudinal measurement of family income, by source of income as well as amount, is crucial for testing re- source-based theories of child and adolescent development, the authors note. In addition, time resources, especially parents’ time spent with children, are typically neglected or not measured well, although they are crucial resources. Measures of other family process mediators are needed to understand the ways in which resources affect child development. Datasets should include information from as many sources as possible, since resources can come from the neighborhood, the school, and the community, as well as the fam- ily. In addition to resources, it is important that data collection allows for the development of dynamic measures of family structure, especially since the changing structure of American families is likely to have a large impact on children and adolescents. The list of additions proposed by the authors to the many national data collection projects reviewed is long and expensive, and funds are probably 

WORKSHOP SUMMARY 11 insufficient to support all of these augmentations and also fund a new na- tional survey of children. Therefore, it is important to consider whether it would be better to allocate resources in a piecemeal fashion across existing surveys or to attempt to pool those resources and spend a substantial frac- tion of them on a new panel study focused exclusively on child and adoles- cent development. No-cost and low-cost additions to existing surveys should be top priori- ties, the authors note. However, in considering the trade-offs between rec- ommending high-cost additions to existing surveys and the fielding of a new survey, a new survey may be the best use of available resources be- cause, although many of the existing datasets provide valuable information on child development, they were either designed for other purposes or are designed too narrowly to serve as a general resource for research on chil- dren. Discussant Gary Sandefur commented that investment in a new longitu- dinal survey of children would be useful in estimating causal models of child development, but it would not fully identify the most pressing infor- mation needs of those who formulate, implement, and analyze policies for children and families. For those using data in policy development, espe- cially those who cannot wait for a new survey, Sandefur called for a focus on such predictor variables as income, family functioning, community re- sources, and instability. Data users need better measurements of what families are doing with their income—how it is utilized and how it affects such child outcomes as high school graduation and teenage pregnancy, he suggested. It is also important to consider what income provides access to, including material resources, opportunities, and perceived security. Family functions, such as time use and how it changes and parent practices in terms of supervision and discipline, are also important, as are the characteristics and availability of community resources and differing patterns of utilization of these re- sources by children and families. Instability, in resources and in where and with whom children live, must also be captured by major data collection efforts. These underlying variables, as well as changes over time, are pos- sible causes of child outcomes, such as health, educational attainment, and employment. A tension exists between short-run scientific and policy needs and long- term needs. Proposals for new surveys would, in the long run, meet the information needs identified, as well as make possible broader and richer policy and scientific research. But quickly addressing gaps in our knowl- edge could be done in the context of existing studies, such as the National Longitudinal Survey of Youth Child-Mother Supplement and the Survey of Income and Program Participation (SIPP) Child Module, Sandefur noted. Given limited national resources for data collection, the choice between a 

12 INTEGRATING FEDERAL STATISTICS ON CHILDREN new survey and supplementation of existing surveys is difficult to make; indeed, they are most appropriately viewed as complementary strategies. Discussant Donald Hernandez responded to several suggestions in the Brooks-Gunn et al. paper by explaining current and proposed changes in data collection by the Census Bureau: • The Census Bureau is planning a Survey of Program Dynamics (SPD), which will follow the entire 1993 SIPP panel for an additional seven years, much of it devoted to measuring child well-being. • The 1996 SIPP panel will identify each parent of a child who lives in the household; this approach will be extended in order to identify biologi- cal, step-, and adoptive parent-child relationships. • The 1990 decennial census effort to explicitly identify biological and stepchildren met with limited success; unless a new approach to data collec- tion is developed, a strong case for the uses of these data would be required. • Likewise, very strong cases would be required in order for the Cen- sus Bureau to consider restoring the marital history question and identify children who are in school. • Speaking to suggestions that the Census Bureau consider producing a matched file with family and neighborhood-level data from the 1990 cen- sus comparable to similar files produced from the 1970 and 1980 census, Hernandez commented that, if the demand for such data were documented and a funding source identified, the Census Bureau would no doubt produce such a data file. • The suggestion that the Census Bureau should consider creating fam- ily-level records in addition to household and person records for their PUMS microdata files is feasible, but the case needs to be made to justify the additional expenditures required. Turning to SIPP, Hernandez noted that specific improvements suggested in the paper—data on child outcomes and on the family processes that translate resources into child outcomes and a broader array of resources, for example—are high on the list of important areas to be considered for inclu- sion. Responding to a criticism that SIPP has a relatively small sample size, limiting efforts to minimize attrition and to follow any children who leave the household, Hernandez said the sample size in the 1996 SIPP will be larger (about 34,000 children). A proposed Survey of Child and Adolescent Development, suggested in the paper as an alternative to high-cost augmentations, warrants thorough evaluation, Hernandez said. He and the authors discussed the design and implementation of such a survey. Since the 1996 SIPP panel will be col- lecting much of the needed data on income, program participation, family change, etc., that will be essential for a national survey of child and adoles- 

WORKSHOP SUMMARY 13 cent development, Hernandez noted that it may be that the most cost-effec- tive approach to fielding the survey will be to use the 1996 SIPP panel as the basic data collection vehicle. The content of new questions and new surveys generated lively discus- sion among the workshop participants. Some agreed on the need for a well- informed, thoughtful process to inform hard choices about investing federal resources for data collection on children, including decisions regarding whether or not to support the costs of a new child-focused survey. Others, who thought it unwise to put all one’s eggs in one basket, called for maintaining a variety of surveys. There was further discussion of the challenges of any effort to design and implement a new, comprehensive national survey on children, including the difficulty of pooling money across federal agencies and time constraints on data collection in homes. The content of new questions and new surveys generated discussion on genetic relatedness of family members, creative uses of sibling data, expanded indices of social and economic status, and time use. A number of workshop participants cited the importance of being able to follow children longitudinally as they move from the households where information is originally obtained to other households and settings over the course of their development. The participants also discussed the value of using an integrated consortium of sampling frames (i.e., households, schools, health and service organizations) for a new survey in order to capture mul- tiple perspectives on children’s development. CHILDREN’S TRANSITION TO SCHOOL Sandra Hofferth outlined the scientific issues involved in evaluating children’s transition to school. Developmental outcomes in this transitional period are typically categorized into three groups: the cognitive domain, including language and achievement; the socioemotional domain, including self-concept, social interaction, and behavioral problems; and the health domain, including physical development and abilities and good health hab- its. The national objectives in the Goals 2000 law add approaches to learn- ing and language usage to the set of categories. Children receive a number of inputs: not only do they receive financial contributions from parents, but they also receive valuable resources such as time, care, and attention from them. Development is also influenced by the environment in which children are raised, including the social and eco- nomic environment and the stability of these circumstances. Mediating factors during early childhood, such as parenting style, communication, atti- tudes, and beliefs, also affect later well-being. Moderating factors such as temperament, parent, family, community, neighborhood, and school charac- teristics are also important. 

14 INTEGRATING FEDERAL STATISTICS ON CHILDREN Research has demonstrated a relationship between early behavior prob- lems and later problems in school and antisocial activities, implying that programs to reduce aggressive behavior should be developed in the early childhood years. However, the effectiveness of such activity has rarely been addressed. A cost-effectiveness analysis of different approaches would be beneficial. Since the only way to address the relevant questions is through scientific analyses that compare the relative effects of these differ- ent factors in one model, it is important that models be as comprehensive as possible, including all potential mediators to the extent possible. Hofferth identified several evaluation studies that currently exist and four studies that will potentially fill some of the gaps. Some gaps, how- ever, will remain, including the lack of a longitudinal study of a national representative sample of children starting prior to school entry and follow- ing children into school, lack of state-level measures, lack of collection of time-use data, and small sample sizes. To fill the remaining gaps, Hofferth recommended supplementing on- going collections, such as the Panel Study of Income Dynamics and the National Longitudinal Survey of Youth Child-Mother Supplement. A coor- dinated data collection effort providing some overlap between surveys would also improve the situation, since it would allow for more cross-study com- parisons and validation than previously possible. Another approach would involve implementing new surveys, such as a national survey of children and a national time-use survey of children and their parents. Discussant John Love addressed one of the central implicit questions of the workshop as it applies to children’s transition into school: What statis- tics are required to design, implement, and evaluate policies aimed at im- proving the health and well-being of children and families? Love noted the importance of studying the transition into school, pointing out that the way in which this transition is handled could make a difference in extending the benefits of children’s preschool experiences. If there are disruptive effects of discontinuity of experience as children enter school, understanding the nature of this transition can help minimize them. Love expanded Hofferth’s three basic scientific questions to account for disparities between age groups, address transition processes in and of themselves, and examine specific inputs that are related either to the transi- tion event or the transition period. He also suggested several additional questions: • What are the important transitions at the time of school entry? Cata- loging these and learning more about how prevalent each is—how many children experience each type of transition—are prerequisites to deciding what policy questions are important to address with federal statistics. • What differences do transitions make in children’s lives? Currently, 

WORKSHOP SUMMARY 15 we can only conjecture about how children’s experiences of transitions af- fect their later school success. • What is the period of transition into school? Evidence indicates that kindergarten serves as a transition from preschool to first grade. A full understanding of transitions at the time of school entry will not be possible until we develop ways of describing children’s experience across this one- year period. But if we consider the school entry transition as a period of time rather than a point in time, this will affect the statistics that are needed, according to Love. In order to sort out information priorities, a consensus effort is needed aimed at identifying the most important things to know about how well children are doing during this transition period and what inputs are most important to understand. Also, a reasonable number of specific transitions should be decided on; Love sees the transitions from home to school, from Head Start to school, and from child care to school as the most important. Finally, more should be learned to elucidate how the features of transition experiences make a difference in such areas as children’s adjustment to kindergarten and in the dimensions of development that have been identi- fied as important for school readiness. Discussant Jerry West agreed with the need for a careful conceptualization of the transition from preschool to school in order to guide this area of federal statistics on children. A model should be developed that represents a long-term process from birth to the point of entry and to subsequent outcomes. Such a model would require the collection of information on a wide range of outcomes; careful assessments of children’s experiences in classrooms and in the family would also be beneficial. Although there is a changing conceptualization of education and of what is important regarding understanding school performance, there is a paucity of data that allow us to look at this issue. In addition, data that will permit examination of links between inputs and outputs are needed. West called for a new generation of studies that involve a collaborative effort between private and public agencies and between disciplines, prefer- ably through a systematic approach and starting with a birth cohort. Data users should be involved early in the development of such a survey, he advised. Workshop participants agreed that the absence of data following chil- dren from preschool into school constitutes a serious gap in the federal statistical system, particularly given the growth in reliance on child care and the debate over school readiness. In this context, they returned to several themes raised earlier during the workshop: 

16 INTEGRATING FEDERAL STATISTICS ON CHILDREN • The tension between starting anew with a survey of children and supplementing existing datasets, • The difficulty of capturing the process of development using survey methods, and • The challenges associated with integrating federal datasets given survey- specific constructs, definitions, and survey items. Participants also stressed the importance of creating an effective mechanism for interagency communication about data collection, based perhaps on the Federal Forum on Aging. EDUCATIONAL ATTAINMENT AND TRANSITION INTO WORK Aaron Pallas examined institutionally based data that are designed to illuminate the role of schooling and the process of transition into the labor force. Many of the important analytic policy questions regarding the transi- tion to work view the youth labor market in the context of its connections to other social institutions, such as schools, families, and communities. Thus, attention is directed to what might be called the school-to-work transition system, or the linkage system that joins school and the economy. Among the key weaknesses in the available data sources that address education-to-work experiences are the kinds of questions asked. The failure to include measures of theoretically relevant constructs in attempts to model or understand complex social phenomena (such as efforts to measure cogni- tive ability) results in specification error. Also, the relevant theoretical constructs are sometimes measured poorly, resulting in measurement error. The second weakness involves the issue of when are questions asked. In periods of relatively rapid educational reform and economic change, it may be desirable to address questions about high school completion, postsecondary access and persistence, and youth labor market experience much more fre- quently than every 8 or 10 years, as is the current practice. A third weak- ness is that household surveys typically are not designed to ensure a critical mass or cluster of individuals within a particular organizational context, such as a school. Yet studies of individuals independent of the school and workplace contexts in which they are situated may not reveal the important role that schools and employers play in structuring educational attainment and the transition into the labor force. To improve the capacity of the federal statistical system to address information needs, Pallas suggested: • Conducting a needs assessment in order to understand the current data available and how well it meets the information needs of policy mak- ers, 

WORKSHOP SUMMARY 17 • Establishing an interagency working group, composed of the agen- cies that work with data that shed light on the well-being of children, to examine the gaps, overlaps, and redundancies in the current federal statisti- cal system, and • Establishing an oversight committee with the responsibility of arbi- trating among different agency interests. Pallas also cautioned that broad federal statistical data collections that function as systems are not a substitute for program evaluation. Targeted studies of particular programs and policies and specific target populations are a more fruitful source of information about how such programs and policies work. Discussant Russell Rumberger acknowledged that a comprehensive and integrated statistical system is needed to understand and address the com- plex problems confronting American youth today. Although federal data collection efforts have recognized this need over time, there is still room for improvement. Rumberger discussed four areas in which improvements can be made to provide better and more comprehensive information on children. The first improvement involves the type of questions asked. To understand and address the problems of children and youth, it is important to obtain simul- taneous information about the characteristics and conditions of the children themselves as well as the characteristics and conditions of the settings in which they live, work, and go to school. Little is known about community and work settings and how they interact with children’s home environ- ments, even though experiences in those settings and interactions among them can significantly influence cognitive and social development. Another improvement involves information on processes. A number of domains shape the development of children, including the cognitive, psychosocial, and health domains, as revealed in their knowledge and skills, attitudes and values, behaviors, and status. Typically much more information is collected on status variables (i.e., scores on achievement tests) than on process vari- ables (i.e., whether children know how to acquire new information to in- form new questions). Although enrollment and dropout status are com- monly measured, for example, these status measures are necessarily arbitrary and reveal very little about the extent to which students are actually en- gaged in the schooling process and learning. The collection of more longitudinal data rather than cross-sectional data would also be useful. In addition, more information should be collected on high-risk or vulnerable populations, such as institutional or minority groups, that are often undersampled or completely ignored in major federal data collection efforts. Rumberger also called for better use of administrative records, more 

18 INTEGRATING FEDERAL STATISTICS ON CHILDREN multimethod studies, more geographic detail, and use of common defini- tions and unique identifiers. He also suggested that more attention be paid to the use of federal data, rather than simply the generation of it. An information “cross-walk” could serve to index all current data available across agencies on related topics. Discussant Suzanne Bianchi pointed out that, given the trends of in- creased inequality of earnings by educational attainment, restructuring of manufacturing, and the downsizing that seems to be affecting white-collar workers, the interrelationship of schooling and work is of increased impor- tance to the well-being of not only the individuals involved, but also the dependents their work supports, especially dependent children. Current data collection efforts should adequately analyze this phenomenon. In order to study adequately the transition from school to work and the connectedness of learning in school and performing on the job, it would be best to capture the student in school (taking into account aspects of the school context) and follow him or her to a job (taking into account work setting). Broad-based household surveys do not allow researchers to under- stand the aspects of schools that help or hinder new labor force entrants’ productivity on the job or how work settings utilize, add to, and detract from the cognitive abilities and training that workers bring to the labor market. Most analysis begins with an implicit assumption that schooling translates into productivity in the labor market; however, this is a hypoth- esis that should be fully investigated by those studying the school-to-work transition. Although the lack of data on individuals embedded in organizational settings is a limitation, the underutilization of existing data, particularly longitudinal data from the National Center for Education Statistics (NCES), is also a problem. Such data may shed light on school-to-work transition. NCES data could be improved by shortening the 10-year gap that separates longitudinal cohorts to allow for some short-term 2- to 3-year studies. Bianchi also noted the potential of SIPP to address school-to-work transitions. A pairing of the Census Bureau with NCES could help in carrying out this work, she said, though she noted that team approaches are costly and hard to execute. Workshop participants touched on the importance of considering sam- pling frames and approaches that extend beyond the household unit and of obtaining data from multiple informants (e.g., employee and employer, stu- dents, teachers, and school administrators). They also addressed the need to standardize measures of educational status across surveys. Participants de- plored the dearth of information about unconventional forms of employ- ment, including underground employment, and the related lack of informa- tion regarding paths to conventional work choices. 

WORKSHOP SUMMARY 19 MONITORING CHANGES IN HEALTH CARE FOR CHILDREN AND FAMILIES In order to monitor significant changes in health care for children and families, information is needed to evaluate the state of health of the popula- tion, the adequacy of access to health care services, and the equity of use of services across different groups of individuals, according to authors Paul Newacheck and Barbara Starfield. Other pressing data needs include: infor- mation on how services are targeted to the needs of the population served; the extent to which services are accessible, comprehensive, and coordi- nated; and the extent to which provided services are justified by evidence of effectiveness and appropriateness. Policy makers also need information about the organization and distribution of resources within the health care system. Newacheck and Starfield listed the five types of sources that are cur- rently available for obtaining such information—vital statistics and surveil- lance systems, population surveys, provider surveys, administrative records, and health systems data—noting that only a few of the existing databases contain data across the multiple domains relevant to health care. As a result, they suggested, it may be desirable to integrate data at several levels. Linking data can yield significant benefits but presents additional chal- lenges, including the need for clarity in concept development, standardiza- tion of measurement, assignment of personal identifiers that maintain confi- dentiality, and accountability mechanisms. Given that limited funds are likely to be available for the monitoring system, the authors see a need for analysis of the marginal costs and mar- ginal benefits of adding components to existing or newly developed moni- toring systems. As much as possible, the system should rely on existing databases, modifying them as needed while keeping in mind their original purposes. Discussant Nicholas Zill pointed to the rapid pace of health care reorga- nization and the important role assigned to health care data as a source of quality monitoring as critical contexts for considering future data needs in this area. In light of budgetary constraints, Zill called for a clear-eyed look at what is vital in current systems and what is expendable and could, as a result, provide opportunities for reallocating funding for more useful data collection. Although there is usefulness in preserving certain elements of the current health care data system, it is also important to recognize the changing health care needs of children (e.g., health problems that arise from social circumstances, such as exposure to violence) and the lack of data to inform decision makers about whether these needs are being met. Zill noted that the Early Childhood Longitudinal Survey, the State and Local Area Immunization Coverage and Health Survey, and the SIPP Child 

20 INTEGRATING FEDERAL STATISTICS ON CHILDREN Module are under way. However, in part because different agencies are funding different pieces, they are proceeding without coordination. Additional data are needed about patient-provider relationships, health plans, and local communities and populations in order to draw an accurate picture of how health care reorganization affects children and families, ac- cording to discussant Robert Valdez. Valdez commented that many of the surveys that can be used to address health care issues are inadequate to the task of monitoring health care reform; they should be complemented with qualitative and local data that can be used to interpret local health care trends, processes, and effects on children and families. In a discussion of this issue, workshop participants cited the critical need for cross-agency planning and coordination in monitoring health care, from the development of common constructs (i.e., concepts of children’s health status) to standardized questions and response choices across sur- veys. They also touched on the tensions surrounding the establishment of unique identifiers in health care data, notably the value of identifiers for data integration that stands alongside concerns about confidentiality. Par- ticipants talked about the importance of carefully considering the appropri- ate sampling frame (i.e., local, state, national) when collecting health data for various purposes. Several participants commented on the challenges associated with including subgroups of children that are often neglected in national data collection efforts, including institutionalized children and those in multicustody arrangements. INTERPERSONAL VIOLENCE INVOLVING CHILDREN AND FAMILIES The definition of violence as any use of force or threat of force, regard- less of intent or magnitude, raises fundamental questions regarding the na- ture of interpersonal violence and the difficulties involved in measuring it. Colin Loftin and James Mercy addressed data collection activities that bear on serious assaultive violence in primary relationships, such as families, and violence involving children. Three household surveys and four organization-based surveys are cur- rently used to collect data on serious interpersonal assaultive violence. Among the pressing needs in the area of information about violence are the devel- opment of improved methods for eliciting valid responses about complex, sensitive, and traumatic violent incidents. It is often difficult to obtain information on violence to children, especially because the abuser is often present in the setting in which the interview is conducted. There must also be improved coverage of persons who are marginally attached to house- holds, and methodological research on coverage must be undertaken. More information is needed about risk factors, social context, consequences, and 

WORKSHOP SUMMARY 21 sequences of violent episodes. Also, more and better information is needed about precursors and long-term consequences of severe assaultive violence. Finally, estimates of the incidence of serious assaultive violence in state and local areas are needed. Loftin and Mercy called for a coordinated federal focus on data related to violence within and across agencies. They also recommended a coordi- nated methodological research program on issues that cut across data sys- tems; the exploration of more efficient ways to identify cases; and an explo- ration of the feasibility of collecting data on violence in key heath care settings—and the training of health care providers to recognize cases. Discussant David Cantor noted that surveys offer several important ad- vantages as a source of information. Surveys can fill in the gaps left by administrative systems. For example, the police collect information only on the incidents that are reported to them, which, at best, constitute only 70 to 75 percent of all crimes; crimes are even less well reported by youths ages 12 to 19. In addition, surveys help ensure centralization of quality as well as standardization, whereas administrative records are subject to unknown quality controls and decentralized data collection. Finally, surveys are con- sistent in their use of terms, although there is not necessarily uniform ad- herence to definitions at the administrative level, especially across local and state jurisdictions. Cantor noted, however, that surveys also have disadvantages involving the validity of measures of violence as reported by either victims or offend- ers and including the confidentiality/sensitivity of information provided, errors in survey design, and coverage problems. The Child Abuse Prevention, Adoption, and Family Services Act of 1988 created an interagency Task Force on Child Abuse and Neglect, as noted by discussant Michael Rand. However, estimating the incidence, causes, and consequences of interpersonal violence of children and families involves more than better cooperation or improved coordination. Child maltreatment encompasses not only physical abuse, but also psychological abuse, neglect, and sexual abuse; the agencies that handle each of these problems work from different perspectives. Their different missions and perspectives result in different data needs, emphases, and even definitions of basic constructs. Differences and conflicts need to be identified and resolved if agencies with different mandates and perspectives are to coordi- nate efforts to improve estimates of the incidence and prevalence of the problem, he suggested. There is a clear need to improve the ability to quantify child abuse and neglect. Although partial information is available through several sources, efforts to develop a composite are difficult; one major problem is that the definition of what constitutes child abuse is not agreed on. In addition, 

22 INTEGRATING FEDERAL STATISTICS ON CHILDREN abuse occurs in a range of settings, and no database currently assesses incidence across settings. A broad approach to data collection is one possible solution to the problem of differential definitions, needs, and perspectives, Rand posited. Information could be collected on all forms of violence and abusive behav- ior, and those deemed not sufficiently serious or not of specific interest to the agency could be discarded. However this approach may not solve all of the problems of a lack of consensus given that any data collection effort is necessarily constrainted by the underlying definitions and principles that guide it. Furthermore, definitional inconsistencies are not the only problem of measuring child abuse; because of the private nature of most violent acts, many subjects—especially children—are very sensitive to providing any information about the incident. Rand agreed with Loftin and Mercy that federal agencies with interests in child and family violence and abuse should coordinate efforts to design an improved mechanism to produce statistics on child and family violence. The interagency Task Force on Child Abuse and Neglect can be an integral part of such a coordination effort. Rand suggested that, although efforts to coordinate federal child abuse statistics may be extremely difficult given the different perspectives and objectives of each of the federal agencies involved, the potential benefits of cooperative efforts make it imperative to undertake the effort. In the discussion, workshop participants noted that a great deal of re- search is exploring more effective data collection methods in the area of interpersonal violence. For example, the National Institute of Justice is experimenting with questions that tap personal perceptions of safety. A youth risk supplement has been added to the National Health Interview Survey, using a portable tape recorder and a personal computer to ensure respondents’ confidentiality (technology that is being considered for the National Crime Survey). In this case, respondents do not give their re- sponses to the interviewer, but rather enter them directly as anonymous data. The child supplement to the Health Interview Survey may also add questions on discipline practices, which some feel step over the line be- tween nonviolent and violent behavior. Participants cited additional prob- lems confronting data collection on violence: • Definitions of neglect vary by state, making it difficult to under- stand, let alone integrate, the information that is collected on this topic. • Although the co-occurrence of child abuse and domestic violence in the same households is a pressing policy concern, no datasets currently exist that allow for the examination of this issue. • Similarly, data do not exist to allow for integrating information about child victimization at home, in school, and in other settings. 

WORKSHOP SUMMARY 23 • Institutional review boards can provide barriers to collecting data on violent behavior. • Although violence has been identified as a major public health prob- lem, it is not clear if data collection efforts in the context of health care reform will include this type of outcome.