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.
120 INTEGRATING FEDERAL STATISTICS ON CHILDREN Love, John M., Mary Ellin Logue, James V. Trudeau, and Katharine Thayer 1992 Transitions to Kindergarten in American Schools. Portsmouth, N.H.: RMC Re- search Corporation. McLanahan, S. 1985 The reproduction of poverty. American Journal of Sociology 90:873-901. 1988 Family structure and dependency: early transitions to female household headship. Demography 25:1-16. Moore, Kristin, and Nancy Snyder 1991 Cognitive attainment among firstborn children of adolescent mothers. American Sociological Review 56:612-624. Mott, Frank L., and Elizabeth Menaghan 1993 Linkages betweeen Early Childhood Family Structure, Socioeconomic Well-Being and Middle-Childhood Socio-Emotional Development. Paper presented at the an- nual meeting of the Population Association of America, Cincinnati, April. National Education Goals Panel 1993 Reconsidering Childrenâs Early Development and Learning: Toward Shared Be- liefs and Vocabulary. Draft Report. Washington, D.C.: National Education Goals Panel. National Task Force on School Readiness 1991 Caring Communities: Supporting Young Children and Families. Alexandria, Va.: National Association of State Boards of Education. Nettles, Saundra 1992 Community Structure and Social Support. A paper in support of planning activi- ties for the NCES longitudinal survey of young children. Johns Hopkins Univer- sity, Center for Social Organization of Schools. NICHD Early Child Care Network in press Child care and child development: The NICHD Study of Early Child Care. In S.L. Friedman and H.C. Haywood, eds., Developmental Follow-up: Concepts, Domains and Methods. New York: Academic Press. OâConnell, Martin, and Amara Bachu 1992 Whoâs minding the kids? child care arrangements: fall 1988. Current Population Reports 30:70. Office of the President 1994 Budget of the United States Government. Washington, D.C.: U.S. Government Printing Office. Powell, Douglas R. 1992 Families and Young Childrenâs School Readiness. Paper prepared for the National Center for Education Statistics. Department of Child Development and Family Studies, Purdue University. Puma, Michael, Calvin Jones, Donald Rock, and Roberto Fernandez. 1993 Prospects: The Congressionally Mandated Study of Educational Growth and Op- portunity (The Interim Report). Washington, D.C.: U.S. Department of Educa- tion. Ramey, Craig, and Sharon L. Ramey 1992 Child and Family Transitions to School: Measuring Adaptation throughout the Elementary School Years. Paper prepared for the National Center for Education Statistics concerning the Longitudinal Studies Program of Young Children. Civitan International Research Center, University of Alabama at Birmingham. Ramey, Sharon L., and Craig Ramey in press Early educational intervention with disadvantaged childrenâto what effect? Ap- plied and Preventive Psychology.
CHILDRENâS TRANSITION TO SCHOOL 121 1993 The National Head Start-Public School Early Childhood Transition Study: An Overview. Birmingham, Ala.: Civitan International Research Center. Schweinhart, Lawrence, Helen Barnes, and David Weikart 1993 Significant Benefits: The High/Scope Perry Preschool Study through Age 27. Ypsilanti, Mich: High/Scope Press. Select Committee on Children, Youth, and Families 1985 Opportunities for Success: Cost-Effective Programs for Children. Washington, D.C.: U.S. Government Printing Office. Seppanen, Patricia, John Love, Dianne Kaplan deVries, Lawrence Bernstein, Michelle Seligson, Fern Marx, and Ellen Kisker 1993 National Study of Before- and After-School Programs. Washington, D.C.: U.S. Department of Education. Shonkoff, Jack 1992 The Conceptualization and Measurement of Child and Family Health. Paper pre- pared for the Longitudinal Studies Program of Young Children, University of Massachusetts. Simpson, Gloria, David Keer, and Marcie Cynamon 1992 Plans for the 1993-94 National Health Interview Survey on Disability. Paper presented at the annual meeting of the American Statistical Association, August. Stafford, Frank P. 1987 Womenâs work, sibling competition, and childrenâs school performance. Ameri- can Economic Review December: 972-980. Timmer, Susan Goff, Jacquelynne Eccles, and Keith OâBrien 1985 How children use time. Pp. 353-382 in F. Thomas Juster and Frank Stafford, eds., Time, Goods, and Well-Being. Ann Arbor: University of Michigan. West, Jerry 1992 Early Childhood Longitudinal Study (ECLS). Washington, D.C.: National Center for Education Statistics. West, Jerry, Elvie Hausken, and Mary Collins 1993 Profile of Preschool Childrenâs Child Care and Early Education Program Partici- pation. Washington, D.C.: National Center for Educational Statistics. Wolock, Isabel 1981 Child health and developmental problems and child maltreatment among AFDC families. Journal of Sociology and Social Welfare 8(1):83-96. Woodward, Bob, and Benjamin Weiser 1994 Costs soar for childrenâs disability program. The Washington Post, February 4. Zill, Nicholas, Kristin A. Moore, Ellen Wolpow Smith, Thomas Stief, and Mary Jo Coiro 1995 The life circumstances and development of children in welfare families: a profile based on national data. In L. Chase-Lansdale and J. Brooks-Gunn, eds., Escape from Poverty. New York: Cambridge University Press. Zill, Nicholas, and Margaret Daly, eds. 1993 Researching the Family: A Guide to Survey and Statistical Data on U.S. Families. Washington, D.C.: Child Trends, Inc. Zill, Nicholas, and Charlotte Schoenborn 1990 Developmental, learning, and emotional problems: health of our nationâs children, United States, 1988. Advance Data 190: November 16.
122 INTEGRATING FEDERAL STATISTICS ON CHILDREN Federal Data on Educational Attainment and the Transition to Work Aaron M. Pallas Children and families have historically been viewed in the United States as matters of private, rather than public, concern. It has only been in the last two to three decades that the role of the state has expanded into issues of the health and well-being of families. Even today, however, most ob- servers would be hard-pressed to argue that the United States has a national policy on children and families. Nevertheless, as the interest of the state in the well-being of children and families has grown, the federal government has begun to debate and implement policies designed to promote the status of children and their families. These debates have often taken place in the absence of timely and accurate data on the well-being of children, and without the benefit of careful analysis of the implementation and effectiveness of the programs that have been serving children and families over the past 30 years. While it may be too harsh to claim that the lack of useful data has led to bad policy, there can be little question that timely, accurate, and relevant data on children and families have the capacity to inform public policy, particu- larly federal policies directed at the lives of children. In this paper, I examine federal data on an important dimension of childrenâs well-being: childrenâs progress through school and into the labor force. A key challenge faced by all societies is the task of providing chil- dren with the personal qualities that enable them to become productive Aaron M. Pallas is at the Department of Education, Michigan State University. 122
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 123 adult members of society. In our society, adulthood is often defined in terms of the ability of an individual to be financially and emotionally inde- pendent and self-sustaining, and is operationalized through the accession of adult work and family roles. Thus, moving into the labor force and forming a family through marriage or parenthood are key markers of adulthood. Yet young people do not move into the labor force unassisted by social institutions. In industrial societies, schools are the social institutions charged with instilling the knowledge, skills, and values that will enable young people to become competent adults. Thus, policy concerns involving the productivity of the American labor force and the competitiveness of the United States in the world economy are frequently translated into concerns about the performance of the American education system. Most people still see a good education as the ticket to economic self-sufficiency, although there are competing explanations for why individuals who go farther through school are likely to achieve greater economic success than those who obtain less schooling. I argue that the analysis of the ways in which American youth negotiate the transition to adulthood reflects an important tension between individual trajectories and the role of social institutions. Institutionally based data often are not reflective of the set of pathways that individuals travel as they become adults. Conversely, studies of individuals independent of the orga- nizational and institutional 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. I suggest, therefore, a need for a set of data collection mechanisms that balance data on individuals and on institutions. THE DEMAND FOR INFORMATION ON EDUCATIONAL ATTAINMENT AND THE TRANSITION INTO THE LABOR FORCE In this section I briefly review some of the major policy questions posed at the federal level that concern educational attainment and the transi- tion into the labor force. These questions have taken on a heightened importance in recent years, in the context of rising concerns about the per- formance of the education system and its capacity to prepare young people for productive work roles in society, and about the apparent decline of U.S. competitiveness in the world economy. It is widely believed that a better- educated cadre of youthâboth with respect to the quantity and quality of the education they receiveâis the key to regaining the economic growth and productivity that characterized the post-World War II economy. I further subdivide the relevant policy questions into two categories: descriptive and analytic. Descriptive questions inquire about the nature and
124 INTEGRATING FEDERAL STATISTICS ON CHILDREN scope of a policy domain and are intended to gauge basic information about its current status and trends. Such questions are often addressed with social indicator data that provide coarse but useful information about the status and direction of change in a system. âWhat is the youth unemployment rate this month?â is an example of a descriptive policy question, and the season- ally adjusted monthly unemployment rate for youth ages 16-24 is an ex- ample of the data that might be used to address this question. Analytic policy questions typically inquire about the relations among variables. Such questions are often used to inform decisions about policy implementation and its likely effects. The questions are frequently posed in cause-and-effect terms and typically are narrower in their scope than de- scriptive policy questions. âDoes participation in a training program for out-of-school youth decrease youth unemployment?â is an example of an analytic policy question, and a carefully controlled study comparing the unemployment rates of program participants and similar nonparticipants is an example of the data that might be used to address this question. Educational Attainment I begin by examining some of the key policy questions regarding edu- cational attainment. In the U.S. schooling system, primary schooling is universal, and virtually all young people attend secondary schools, although not all complete their secondary education. Policy makers are frequently concerned with the key branching points of high school completion, postsecondary access, and postsecondary completion. Descriptive Policy Questions Since high school completion is widely regarded as a minimum prereq- uisite for the development of the skills, knowledge, and values needed to succeed in the emerging economy, there is a great deal of interest in under- standing who is graduating from high school and who is not. The overall high school dropout and completion rates are important indicators of the health of the education system, although these rates have been measured in many different ways, often leading to widely divergent estimates of the scope of the problem. Moreover, given the nationâs historic commitment to equality of opportunity, coupled with concerns about an emerging underclass, policy makers wish to know whether dropout and completion rates differ for traditionally disadvantaged populations, such as racial and ethnic minori- ties, women, and the poor, or for residents of central cities or recent immi- grants to the United States. There also is interest in whether dropout and completion rates vary across different regions of the country, and in the
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 125 timing of dropping out in the secondary school career: Do dropout rates differ by studentsâ age or grade level? Several of these questions have been codified in the National Education Goals agreed to by President Bush and the 50 governors in 1989. Goal 2 states: âBy the year 2000, the high school graduation rate will increase to at least 90 percent.â The two additional objectives associated with Goal 2 state: âThe nation must dramatically reduce its dropout rate, and 75 percent of those students who do drop out will successfully complete a high school degree or its equivalentâ and âThe gap in high school graduation rates between American students from minority backgrounds and their non-mi- nority counterparts will be eliminated.â Although the National Education Goals are surprisingly silent on par- ticipation in postsecondary schooling, there nevertheless are a host of long- standing descriptive policy concerns about access to and completion of postsecondary schooling. This is particularly true in light of the longstanding federal role in financing higher education. It is estimated that more than a third of all undergraduates in recent years have received some form of federal financial aid for their postsecondary schooling. Many of the same descriptive questions about high school completion are replicated at the postsecondary level. Policy makers are interested in knowing whether en- rollment in and completion of postsecondary schooling varies according to the social characteristics of youth, including their gender, racial/ethnic identity, family income level, and age. Moreover, in light of the differentiation of the American higher education system, there is great interest in variability in the type of institutions and programs in which students enroll, particu- larly the contrasts between two-year and four-year postsecondary institu- tions and between academic and nonacademic programs of study. In some instances, in which a particular postsecondary program is strongly linked to a particular career trajectory, policy makers may wish to know the number and types of students enrolling in particular programs, such as those leading to careers in science and engineering, or mathematics and science teaching, for example. Some of these descriptive questions at the secondary and postsecondary levels may be addressed through both surveys of individuals and surveys of institutions. Basic enrollment data, for example, are produced by the Na- tional Center for Education Statisticsâ annual Common Core of Data (CCD) program at the elementary-secondary level and the Integrated Postsecondary Education Data System (IPEDS) program at the postsecondary level. How- ever, such institutional data collections have limited utility for many of the analytic issues considered below.
126 INTEGRATING FEDERAL STATISTICS ON CHILDREN Analytic Policy Questions At the secondary level, the key analytic policy questions take two forms. The first might be called âbasicâ policy research, in the sense that the goal is to provide a greater understanding of the school dropout phenomenon that might be used to generate potential policy instruments designed to promote high school completion. The second might be labeled âappliedâ policy research, in the sense that it pertains to understanding the impact of already formulated programs and policies on school completion. The basic analytic policy concerns at the secondary level consist of a set of questions about the causes and consequences of dropping out of high school, including the dynamics of school enrollment. We still know rather little about why young people leave school before completion, and what happens to them after they do so. We also have relatively little insight into why individuals travel different pathways to high school completion (i.e., regular day high school diplomas, General Educational Development or GED credentials, and other high school completion credentials), why they travel these pathways on different timetables (i.e., moving in and out of the education system), or what the socioeconomic consequences of these differ- ent pathways might be. The concerns I have identified here parallel those raised in a recent report from the Office of Educational Research and Improvement on reach- ing the high school completion goal of the National Education Goals (OERI Goal 2 Work Group, 1994). That report suggested four questions that the authors believed warranted further research that might inform efforts to make progress toward the goal: (1) What do we know about mainstream dropouts? How can we explain the large numbers of youngsters who, with- out seeming disadvantaged, still fail to complete high school? Conversely, why do so many more of their peers succeed in completing high school? (2) What are the factors that lead Hispanics, American Indians, and students with disabilities to leave school at greater rates than those in the main- stream? (3) What are the consequences of completing a GED rather than a regular high school diploma? (4) To what extent does the lure of adolescent employment and the challenge of teenage parenting influence the prospects for higher graduation rates? The applied analytic policy concerns at the secondary level pertain pri- marily to the evaluation of specific policies and programs designed to influ- ence the high school graduation rate, and in addition to understanding the unanticipated consequences of other educational and social programs for the high school graduation rate. For example, the focus of the school reform movement of the 1980s was the attempt to raise standards for stu- dent performanceâby increasing high school graduation requirements, imple- menting high-stakes exit tests, and increasing student workloads. Although
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 127 policy makers may not have intended such reforms to have an adverse effect on school dropout rates, in the absence of new resources targeted to help marginal students achieve the new standards, it seems likely that these reforms might have had an unintended effect on school completion. Yet we have little concrete evidence of the impact of the school reform movement on the likelihood of school completion. Nor do we have definitive evidence on the consequences of incentive programs that rely on rewards or sanctions to spur students to stay in school, such as those that promise college tuition money to youth who complete high school, those that deny driverâs licenses to individuals who leave school before graduation, and those that lower welfare benefits to the families of youth who stop going to school. There also are naturally occurring variations in the school experiences and environments of children and youth that may be consequential for their chances of staying in school. Classroom reward and evaluation systems, the organization of instruction into homogeneous ability groups and age- graded classrooms, grade retention policies, school size, and curricula all represent factors at least partly amenable to policy manipulation that might have consequences for whether young people complete their high school educations. At the postsecondary level, the central analytic questions concern the intertwining of families, work, and postsecondary schooling. There is con- siderable interest in the ways in which families contribute to the financing of the postsecondary schooling of dependent children, and in the ways in which independent youth finance their college educations. Since the federal government is a significant source of financial aid for many young college attendees, policy makers wish to know the consequences for postsecondary enrollment and persistence of variations in the levels and sources of federal and other financial aid to students, and of changes in the eligibility criteria for aid programs. It is, moreover, important to understand postsecondary schooling in the context of other social institutions in which youth partici- pate, such as work, family, and the military, to name some of the most important ones. Increasing numbers of youth are combining work and postsecondary schooling, and many choose between military service and postsecondary schooling upon graduating from high school. The social and economic consequences of variations in the timing of postsecondary school- ing, and the ways in which it is sequenced with family, work, and military experiences, are not well understood. Transition into the Labor Force Unlike that for educational attainment, our knowledge about school-to- work transitions is extremely limited. We lack powerful theories of how young people negotiate the transition from school to work and what social
128 INTEGRATING FEDERAL STATISTICS ON CHILDREN forces are important in that transition. In the absence of strong theories, it is unlikely that existing federal data sources will be especially helpful in illuminating the most pressing policy concerns. Such data are grounded in an understanding of the youth labor market and of learning in school and on the job that is rapidly being superseded by more complex views. For example, human capital theories have viewed skill or ability as the centerpiece of labor markets, with individuals selling their skills and em- ployers purchasing them. In this view, skill supply and demand is among the most important policy considerations in the transition into the labor force. But this view takes skill or ability as a relatively fixed feature of individuals independent of their learning contexts, and thus assumes that the skill, ability, and learning demonstrated in school can readily transfer to the workplace. The cognitive demands of the workplace, and the ways in which it facilitates or inhibits ongoing learning, are left largely unexamined. If, however, one adopts a constructivist view of learning that sees the con- struction of knowledge as inextricably linked to the learning environment, then it would not make sense to view the school-to-work transition as a simple matter of transferring those cognitive skills acquired or developed in school to the workplace. Rather, the transition is much more complicated, as each setting for learning reshapes what individuals know and are able to apply in the next setting. Moreover, as individuals change and develop over time, so too do workplace learning environments, which implies a need to build a dynamic conception of work settings into research designs. Even those data collections that do in fact assess aspects of the workplace typi- cally conceive of it as static, rather than changing. This is especially prob- lematic under conditions of rapid technological change, which clearly char- acterizes a number of industries in the United States. Such technological change can result in either increased or lessened cognitive demands on young workers, depending on the industry or technology in question. Our data collection mechanisms need to take account of the impact of such change on the nature of youth employment. These are not the only such alternative perspectives on the transition to work, and it would be taking this paper far afield to explore even these approaches in greater detail. But the existence of such competing ways of conceptualizing the problemâor, more precisely, the lack of any adequate way of conceptualizing the problem at this timeâpoints out the limits of available data collections to inform our understanding of the transition to work. This understanding seems more likely to emerge from rich field studies than from federal statistics-gathering efforts.
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 129 Descriptive Policy Questions Perhaps the most basic descriptive policy questions regarding the tran- sition into the labor force pertain to the quantity and quality of youth em- ployment. Policy makers are interested in the incidence and intensity of employment and unemployment among youth, both during and after periods in which they are enrolled in school. Among those youth who are em- ployed, the number of hours worked, type of work, and wages all are funda- mental features of the youth employment picture. Moreover, policy makers want to know how youth employment and unemployment vary by the social background of youth, including their race/ethnicity, gender, household com- position, family income level, geographic location, age, and educational attainment, and whether these patterns are stable or changing over time. All of these factors are measurable. Unfortunately, perhaps the most important descriptive policy questions are not as easily operationalized. For example, from the standpoint of human capital theory, the key to under- standing the youth labor market and the policy issues surrounding it is the supply of and demand for skill. Youth bring with them to a job a set of skills derived from schooling, previous work experience, and family and community influences. Conversely, employers seek employees with the requisite skills to carry out the tasks required of workers in their firms. The matchâor mismatchâbetween the supply of and demand for skill has more profound implications for youth labor market policy than virtually any other feature. Yet the little direct evidence on skill supply and demand among youth does not derive from large-scale federal statistical surveys, but rather from small, industry-specific or firm-specific case studies. Even individual industries or firms, however, are not homogeneous or highly stable over time, so that generalizations are extremely difficult. Analytic Policy Questions Many of the important analytic policy questions regarding the transition to work view the youth labor market in the context of its connections to other social institutions, particularly schooling, families, and communities. Thus, attention is directed to what might be called the school-to-work tran- sition system, or the âlinkageâ system joining schooling and the economy. Whereas the match or mismatch in skill supply and demand is the most important feature of the linkage system, there are other important policy issues to be explored as well. For example, how do schools help young people make the transition from school to work, and how might they be more successful? How do partnerships or contracts between schools and employers affect the ways in which youth negotiate the transition from school to work? What kinds of signals do schools provide to employers
130 INTEGRATING FEDERAL STATISTICS ON CHILDREN about the qualities of youthful workers, and how can we improve the qual- ity of information available to both youth and employers about each other (see Rosenbaum et al., 1990)? Another set of policy questions is concerned with the linkages between work and families and communities. Mizell (1988) identifies a pressing need to understand how families and communities influence adolescent work behavior. The transition to work often is viewed in the context of specific communities, because the youth labor market is not a national market, but rather a local one. Thus, local community conditionsâe.g., the kinds of social dislocations defined by weakened social institutions that William Julius Wilson has written about in The Truly Disadvantaged (1987)âframe youth attitudes toward work, the reasons why youth work, and the opportu- nities youth have for productive work. Other policy questions focus on the characteristics of the jobs in which youth work more so than on the individuals working in those jobs. For example, what kinds of opportunities for advancement, training and learn- ing are offered in the workplaces where youth are found? In what ways do workplaces function as learning environments for youth, and how is this learning connected to future success in adult work and social life (Rosenbaum et al., 1990)? The kinds of analytic policy questions I have described here are diffi- cult ones to address, and we may not currently have the tools to shed light on all or even many of them. In some cases, we may be able to learn a great deal simply by piggybacking some new survey items on existing data collections, thereby preserving existing data while forging a new link to additional questions of interest. But not all instances will be this easy. Some questions are not easily approached via sample surveys. Questions about learning in the workplace, for example, may be as resistant to paper- and-pencil measures as questions about learning in school. Still other ques- tions may require detailed observation or understanding of work sites or data from employers. In either event, samples of individuals separated from the organizational context of their work experience may not be the best means for understanding some of these complex questions. There are, to be sure, other kinds of questions for which surveys of individuals may be appropriate. For example, although earlier studies have addressed issues such as the consequences of adolescent work experience, the connections between school performance and getting a good job, and the ways in which adolescent work before school-leaving is connected to work after leaving school, such studies need to be updated. In some cases, the data on which they are based are now quite old, and there is no guaran- tee that they apply equally well to the tighter labor market conditions that confront the current generation of youth. General-purpose longitudinal sur- veys of individuals have been the primary means for addressing the kinds of
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 131 issues I am describing, and they will probably continue to be so in the future. Charner and Fraser (1988), in reflecting on the kinds of policy issues noted above, call for an integrated database focused on youth employment (p.55): Currently there does not exist a single data base which examines the de- tailed patterns of student participation in work experiences; the multiple dimensions of the nature of these work experiences; the roles and responsi- bilities of students as workers; and the effects of differential work experi- ences on different educational, family, social, and personal short and long- term outcomes. Such a dataset on a representative sample of students from different subgroups of the population is clearly needed. They go on to attempt to specify some of the requisite data elements in such a data base (p.56): A set of common questions needs to be developed which can be used by future researchers. These questions need to cover, at a minimum, the following: (a) work histories including: type of job, duration, hours worked, wages, length of employment, employer, and benefits; (b) multi- ple dimensions of work including: occupational self direction, position in the organizational structure, job pressures, and extrinsic risks and rewards; (c) reasons for working including: financial, experiential, learning, and social-psychological reasons; and (d) outcomes including: attitudinal and behavioral outcomes, both short and long-term. Charner and Fraserâs (1988) data collection agenda is ambitious, but it does point out some of the gaps in the existing data on the transition to work. It is to these data, and the statistical programs that produce them, that I now turn. THE AVAILABLE FEDERAL STATISTICAL DATA SOURCES In this section, I describe the key federal sources of data on educational attainment and the transition into the labor force. I include in this descrip- tion major data collection activities carried out by or sponsored by federal government agencies, such as the Census Bureauâs Current Population Sur- vey and the Longitudinal Studies Program of the National Center for Educa- tion Statistics (NCES). I also consider the Panel Study of Income Dynam- ics (PSID), an important ongoing study that originated in the Office of Economic Opportunity and is now funded through the National Science Foundation and various private sources. Because it is no longer sponsored by the federal government, perhaps it is inaccurate to describe the PSID as a federal data source. Nevertheless, I include it because of its unique proper- ties and the capacity of the study to address important questions pertaining
132 INTEGRATING FEDERAL STATISTICS ON CHILDREN to schooling, work, and the life course. A brief summary of these datasets and their pertinent features is displayed in Table 1. Current Population Survey The Current Population Survey (CPS) is a sample survey conducted by the Bureau of the Census in the Department of Commerce, with the collabo- ration of the Bureau of Labor Statistics in the Department of Labor. Over its 53-year history it has been the primary source for perhaps the most prominent indicator of the health of the American economy, the seasonally adjusted monthly unemployment rate. The CPS also produces data on the characteristics of the labor force, including the number and characteristics of individuals who are employed, unemployed, or not in the labor force. The survey provides basic data on the labor force participation of individu- als, including their earnings, numbers of hours worked, occupations and industrial classifications, and job search strategies (among the unemployed), as well as basic demographic data on households. In addition, there are periodic supplements sponsored by the Bureau of the Census or other fed- eral agencies that provide additional information on such topics as school enrollment and child care. The CPS samples approximately 60,000 households drawn from over 700 areas (typically counties and independent cities) located across the 50 states and the District of Columbia. The sampled households are rotated, with each household in the sample for four consecutive months, then out of the sample for eight additional months, and then back in the sample for four consecutive months. This design is intended to reduce sampling errors in month-to-month or year-to-year comparisons by ensuring a certain amount of overlap in the sample composition over short time periods (Bureau of Labor Statistics, 1988). Although the CPS data are longitudinal, the 16- month span of a householdâs participation in the sample is much more like a snapshot than a moving picture of the school-to-work transition. A house- hold informant provides data on the civilian noninstitutionalized household members age 16 or over, totaling approximately 113,000 individuals each month (Bureau of Labor Statistics, 1988). The CPS sample thus generalizes to the civilian noninstitutionalized population of the United States age 16 years of age or over. The CPS has recently undergone extensive revisions to its survey in- strument and data collection procedures. The questionnaire has been re- vised to clarify existing definitions, incorporate changes in definitions rec- ommended by several commissions and advisory groups, and improve the wording and sequencing of questions. In addition, all interviews are now carried out with the aid of computers, enabling the use of dependent inter- viewing (using information from the previous monthâs interview in the cur-
TABLE 1 Summary of Federal Statistical Data on Educational Attainment and the Transition to Work Study Title Sample Periodicity Data Sources Key Measures Current Population Rotating nationally Monthly â households Household informant Labor force participation; school Survey (CPS) representative sample are in sample for 4 interview provides data enrollment; educational attainment; of 60,000 households months, then out of on civilian income and earnings; basic sample for 8 months, noninstitutionalized demographic data then back in sample for household members 4 months age 16 orover National Longitudinal Nationally representative Base-year survey in Student questionnaires; Family background; educational and Study of the High sample of high school 1972, with follow-up base year standardized occupational plans; postsecondary School Class of 1972 seniors in 1972 â sample surveys in 1973, 1974, test battery; school schooling, work and military (NLS-72) ranges from 17,000 1976, 1979 and 1986 records from respondentsâ experience; attitudes and values students in base year to high schools; postsecondary 12,000 students in fifth transcripts follow-up High School and Nationally representative Base-year survey in Student questionnaires; Family background; school Beyond (HS&B) sample of 30,000 high 1980; follow-up cognitive test battery in experiences; educational and school sophomores and surveys for both cohorts base year and first occupational plans; work 28,000 high school seniors in 1982, 1984, and follow-up; base-year experience; postsecondary EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK enrolled in more than 1986; follow-up survey parent and teacher schooling; attitudes and values; 1,000 public and private for sophomore cohort questionnaires; high family formation schools in 1980 in 1992 school administrator and teacher survey; postsecondary education transcripts continued on next page 133
TABLE 1 Continued 134 Study Title Sample Periodicity Data Sources Key Measures National Education Nationally representative Base-year survey in Student questionnaires; Family background; language use; Longitudinal Study of sample of 25,000 eighth 1988; follow-up cognitive test battery in school experiences; work 1988 (NELS:88) graders enrolled in more surveys in 1990, 1992, base year and first and experiences; plans for the future; than 1,000 public and and 1994; scheduled second follow-ups; parental involvement; teachersâ private schools in 1988 follow-up) in 1996, with parent questionnaire perceptions; school policies and possibility of one or in base year and practices; curriculum content two additional follow- second follow-up; ups at two-year school questionnaires; intervals teacher questionnaires; dropout questionnaires Beginning Nationally representative Base-year survey in Student questionnaires; Family background and financial Postsecondary sample of 7,000 students 1989-90; follow-up National Postsecondary data; postsecondary schooling and Study (BPS) entering postsecondary survey in 1992; Student Aid Study employment experiences, including schooling in 1989-90 scheduled follow-ups at (NPSAS:90) parent academic progress, field of study, two-year intervals questionnaires; persistence, and future plans; postsecondary transcripts family formation; financial aid and expenses associated with postsecondary schooling Baccalaureate and Nationally representative Base-year survey in National Postsecondary Family background; financing of Beyond Study sample of 16,000 students 1992-93; scheduled Student Aid Study (NPSAS) undergraduate and graduate (B&B) completing their follow-ups at 1, 3, student and parent education; educational experiences baccalaureate degrees in 6, 9, and 12-year questionnaires; institutionally in undergraduate and graduate 1992-93 intervals generated student records school; employment experience; family formation; future expectations INTEGRATING FEDERAL STATISTICS ON CHILDREN
National Nationally representative Base-year survey in Respondent interviews; high Sources of income; training; Longitudinal sample of 12,000 1979; annual school transcripts; Armed employment and unemployment Study of Youth noninstitutionalized civilian reinterviews Services Vocational Aptitude spells; occupational mobility; (NLSY) and military youth ages 14 Battery scores; assessments attitudes about and knowledge of to 21 in 1979 of children of mothers in the work; education; family sample . formation; child care; child cognitive and socioemotional development Panel Study of Income National sample of Base-year interview in Respondent interviews Sources of income; employment Dynamics (PSID) approximately 8,000 1968; annual and earnings histories; family households, beginning in reinterviews composition; household 1968 expenditures; housing; child care information; disability and illness; job search strategies; retirement plans and experience; savings patterns; standardized test performance EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 135
136 INTEGRATING FEDERAL STATISTICS ON CHILDREN rent interview) and reducing errors produced by complicated skip patterns (Polivka and Rothgeb, 1993). Of special interest here is the revision of the question eliciting educa- tional attainment in the CPS. Beginning in January 1992, the CPS item asks, âWhat is the highest level of school . . . has completed or the highest degree . . . has received?â The response options include: âless than 1st grade,â â1st, 2nd, 3rd, or 4th grade,â â5th or 6th grade,â â7th or 8th grade,â â9th grade,â â10th grade,â â11th grade,â â12th grade NO DIPLOMA,â âHIGH SCHOOL GRADUATEâhigh school DIPLOMA, or the equivalent (for example: GED),â âSome college but no degree,â âAssociate degree in collegeâOccupational/vocational program,â âAssociate degree in collegeâacademic program,â âBachelorâs degree (For example: BA, AB, BS),â âMasterâs degree (For example: MA, MS, MEng, MEd, MSW, MBA),â âProfessional School Degree (For example: MD, DDS, DVM, LLB, JD),â and âDoctorate degree (For example: PhD, EdD).â I belabor this because this new item is a substantial improvement over the old item, which for more than 40 years consisted of two parts: âWhat is the highest grade or year of regular school . . . has ever attended?â and âDid . . . complete the grade?â As Kominski and Siegel (1993) note, the old item did not identify specific degrees, so that analysts were obliged to infer attainment of a degree from years of schooling. This was a particular problem in the classification of high school graduates, as some individuals have completed 12 years of school without obtaining a high school diploma, while others have completed fewer than 12 years of school but obtained a credential equivalent to a high school diploma such as a GED or an alterna- tive credential. Although the change in this item does interrupt a time series spanning 50 years, Kominski and Siegel (1993) argue persuasively that the meaning of educational attainment had changed sufficiently over this period to make the old item misleading. The CPS data also are used to provide estimates of the national high school dropout rate. Because the October school enrollment supplement asks about both current enrollment status and enrollment status one year
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 137 prior to the survey date, as well as current attainment level, it is possible to identify those individuals who were enrolled in grades 10-12 a year ago, are not currently enrolled in grades 10-12, and have not completed high school (Kominski, 1990). Such individuals may be said to have dropped out of school in the past year. NCES, using the October CPS data, reports the proportion of such individuals in a particular population group as the event dropout rate (see, e.g., Kaufman et al., 1992). NCES also relies on the October CPS to estimate the proportion of a particular population group that has not completed high school and is not currently enrolled in school. This proportion is referred to as the status dropout rate, because it estimates the proportion of a population group that has the status of dropout at a particular point in time. The primary strengths of the CPS for policy purposes are its periodicity and ability to generate national estimates. The data can be used to generate annual trends in the educational and occupational attainments of the U.S. young adult population, including rates of secondary and postsecondary access and degree completion and of employment and unemployment. Sub- ject to variations in question wording and survey procedures, the CPS data provide a consistent data series for broad trends in the schooling and work accomplishments of youth. The CPS data are not as useful for addressing policy concerns targeted at specific subpopulations, such as minority youth. Because the numbers of minorities enrolled in grades 10-12 in the CPS sample in a given year is not large, estimates of statistics pertaining to them (e.g., the dropout rate for Hispanic males) are subject to substantial sampling errors. This problem is, of course, heightened if the subpopulation is further subdivided by level of last grade attended or income level. Kominski (1990) used 3-year moving averages to estimate such rates. Such moving averages increase the sample size for a particular statistic, thereby reducing the sampling variability, but simultaneously increase the nonsampling errors by relying on data from different years. Hauser (1991:9) notes several other difficulties in using the CPS data as a backdrop to policy formulation: There are some problems in using the CPS data to measure adolescent educational transitions. The samples become excessively small and statis- tically unreliable when we try to focus on key transitions, especially among minority groups. Family income is not measured well, and academic abil- ity is not measured at all. . . . We lose the link with parents when children leave their parentsâ household and do not live in group quarters at school. The CPS does not cover persons in the military or in institutions, like prisons and jails, that now house a substantial minority of young adults. The CPS tells us little about the schools or colleges in which students are enrolled; we learn only whether enrollment is at a 2-year or 4-year public
138 INTEGRATING FEDERAL STATISTICS ON CHILDREN or private institution. Other recent content changes have reduced the use- fulness of the October data. . . . At the same time, unlike the institutional or longitudinal surveys of NCES, the October CPS does provide annual data on college entry and enrollment. In sum, then, the CPS functions best as a relatively crude indicator of trends in the educational attainment and employment experiences of youth. It is of less value for policy issues that demand greater detail in the descrip- tions of the schooling or employment experiences of youth, and of limited use for examining the dynamics of schooling and the transition to work. It is hard to imagine major changes to the structure of the CPS. Its primary purpose as a barometer of the health of the economy is so impor- tant that any change that might jeopardize the reliability and validity of the data series is likely to meet with stiff resistance. Even seemingly small changes to the structure of the CPSâfor example, oversampling minorities or the poorâcould have adverse effects on estimates of key educational and economic trends. More fundamental changes in the study designâmore detailed questions regarding schooling and work, or an extended follow-up, to cite but a few examplesâseem out of the question, due to the technical problems and costs that might ensue. There are some small changes, com- ing on the heels of the recent CPS redesign, that might be undertaken, but the CPS is unlikely to be transformed in ways that would fundamentally alter its use for monitoring educational attainment and the transition to adulthood. The next set of federal data collections complement the strengths and weaknesses of the CPS. Whereas the periodicity and stability of the CPS allows for a timely, consistent measure of trends in the educational attain- ment and employment of youth, the National Longitudinal Studies Program of NCES examines cohorts spaced 8 to 10 years apart. However, this latter program provides much greater analytic detail on the dynamics of educa- tional attainment and employment than is provided by the CPS. The NCES National Longitudinal Studies Program For more than two decades, NCES has been conducting longitudinal studies of large, nationally representative samples of school-age youth. The National Longitudinal Studies Program has been an important resource for the analysis of studentsâ experiences as they move through secondary school and into early adulthood. The program has become even more valuable in recent years, through NCESâ outreach and dissemination efforts. Although the datasets were once extremely cumbersome to work with, they now are much more accessible and user-friendly, and NCES is more directly in- volved in the training of new users. The datasets are so rich that the hundreds of doctoral dissertations, journal articles and books based on them
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 139 merely scratch the surface of their potential contributions to our under- standing of schooling and the transition to adulthood. When Ellis Page, describing the initial analyses of the base year of High School and Beyond in 1981, said, âWelcome to the data feast!â he was not exaggerating. NCES has fielded three national longitudinal studies of secondary stu- dents that have traced their educational and occupational careers into young adulthood: the National Longitudinal Study of the High School Class of 1972 (1972), High School and Beyond (HS&B), and the National Education Longitudinal Study of 1988 (NELS:88). In addition, NCES is currently planning a new study in this series, the National Education Longitudinal Study of 1998 (NELS:98), that will begin with a new cohort of middle or high-school-age students in 1998 (Davis and Sonnenberg, 1993). NCES has also begun two longitudinal studies focusing on participation in postsecondary education: the Beginning Postsecondary Student Longitudinal Study (BPS), and the Baccalaureate and Beyond Longitudinal Study (B&B). Each of these studies is discussed below. The National Longitudinal Study of the High School Class of 1972 The National Longitudinal Study of the High School Class of 1972 (NLS-72) is a stratified probability sample of public and private high school students enrolled in the 12th grade in spring 1972. The sample design called for surveying 18 high school seniors in more than 1,000 high schools across the country. Schools in low-income areas and with high minority enrollments were oversampled. There have been five follow-up surveys, in 1973, 1974, 1976, 1979, and 1986. The NLS-72 study thus follows a single cohort of youth from (approximately) age 18 to age 32. The sample size increased from approximately 17,000 students in the base year to approxi- mately 21,000 students in the first follow-up, due to the addition of base- year nonrespondents. Response rates for the follow-up surveys are uni- formly high. The 1986 follow-up, a subsample of the original sample, included more than 12,000 individuals. In each wave of the NLS-72 study, respondents completed an approxi- mately hour-long survey questionnaire tapping aspects of family background, educational and occupational plans and experiences, and attitudes and val- ues. The follow-up questionnaires, collected primarily through the mail, typically focused on respondentsâ experiences since the preceding follow- up. In the base year, respondents also completed a standardized test bat- tery. A subset of the sample completed a standardized test in the 1979 follow-up. In addition, NCES gathered school record information from the respondentsâ high schools and conducted a Postsecondary Education Tran- script Study in 1984. The NLS-72 study is limited in its capacity to inform our understanding
140 INTEGRATING FEDERAL STATISTICS ON CHILDREN of the contribution of secondary schools to the educational and occupational careers of youth. Since the survey began with students enrolled in the 12th grade in spring 1972, there are only a small number of students who did not complete high school, even though about one-sixth of all 18- to 19-year- olds (and a substantially higher number of minority youth) in 1972 were neither enrolled in school nor high school graduates. The survey provides no information whatsoever on the experiences of this group, which might have experienced the most difficulty in the transition to adulthood. More- over, the formative educational and social experiences of these youth are addressed either retrospectively or, more commonly, not at all. Because there are no data on studentsâ achievements or attitudes that predate the 1972 base-year data collection, it is virtually impossible to examine the contribution of schools and schooling to the educational and occupational trajectories these youth traveled. The High School and Beyond Study NCES began the High School and Beyond (HS&B) study in 1980. Once again, the design called for a two-stage stratified national probability sample of high school students, with schools the first-stage sampling unit and stu- dents within schools the second-stage unit. The study randomly selected approximately 1,100 public and private high schools across the country, oversampling high-achieving private schools, public schools with concen- trations of Hispanic students, Catholic schools with large proportions of minority students, and alternative public schools. Up to 36 sophomores and 36 seniors were sampled within each of these schools, yielding a base-year sample size of approximately 30,000 high school seniors (the HS&B senior cohort) and 28,000 high school sophomores (the HS&B sophomore cohort). A total of 12,000 members of the senior cohort were followed up in 1982, 1984, and 1986; roughly 27,000 members of the sophomore cohort were followed up in 1982, and about 15,000 members were followed up in 1984, 1986, and 1992. The senior cohort thus has been followed from approxi- mately age 18 to age 24, whereas the sophomore cohort has been followed from approximately age 16 to age 28. The HS&B study was designed to facilitate comparisons with NLS-72, and in fact there have been a series of analytic studies comparing the experiences of both the sophomore cohort and the senior cohort with the cohort of high school seniors in NLS-72 (see, e.g., Ekstrom et al., 1988; Alexander et al., 1987). In the base year of the study, members of both cohorts completed an hour-long questionnaire and a cognitive test battery of similar length. In addition, a school administrator filled out a school questionnaire, and a subsample of parents of sample members completed a parent questionnaire. A teacher of the sampled student filled out a brief comment checklist on the
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 141 studentâs habits and performance. As with the NLS-72 base-year student questionnaire, the HS&B base-year student questionnaire emphasized fam- ily background, school experiences, and future educational and occupational plans, including financial planning for college. Subsequent follow-up ques- tionnaires elicited detailed information on the characteristics of jobs held and postsecondary schooling, as well as attitudes, values, and family forma- tion. Spells of schooling, work, marriage, and childbearing are noted via start and stop dates for each event, thus providing relatively complete event histories over the period of time covered by the follow-ups. The first follow-up also administered the cognitive test battery to the sophomore cohort. Of special interest is the first follow-up of the sophomore cohort, due to the window it provided on dropping out of high school. Because the HS&B study surveyed (and even administered a cognitive test battery to) youth identified as out of school at the time of the first follow-up, it has provided very useful information on the causes and consequences of dropping out of high school and on the dynamics of high school completion. There have been a number of supplemental data collections carried out under the auspices of HS&B, including the Postsecondary Education Tran- script Study of the senior cohort in 1984 and a similar study of the sopho- more cohort scheduled for 1993 (Davis and Sonnenberg, 1993). Perhaps the most prominent is the Administrator and Teacher Survey carried out in 1984. Historically, NCES surveys have been more attuned to individual attainments and career trajectories than to the influence of organizational environments on those attainments. Hence, the questionnaires are rarely designed to assess the influence of educational programs and policies on student outcomes. In light of this, a consortium of educational research and development centers returned to about 500 of the original 1,100 high schools in the base year of the HS&B study in 1984 to gather additional information on topics such as school climate, governance, and school improvement pro- grams. Surveying one administrator and up to 30 teachers in each of these schools, the consortium generated a dataset that could be linked to the HS&B student data to provide reliable information on school organization. Unfortunately, the assumptions needed to use 1984 data on school organiza- tion to inform an understanding of studentsâ experiences between 1980 and 1982 are not always plausible (see, e.g., Chubb and Moe, 1990). The National Education Longitudinal Study of 1988 The National Education Longitudinal Study of 1988 (NELS:88) was the third NCES longitudinal study, and its design retained many of the features of NLS-72 and HS&B and added to them. The most important design change was a shift in the grade cohort followed over time. Whereas NLS-
142 INTEGRATING FEDERAL STATISTICS ON CHILDREN 72 traced high school seniors, and HS&B had both a sophomore cohort and a senior cohort, the NELS:88 study began with a cohort of students enrolled in the 8th grade in spring 1988. As with NLS-72 and HS&B, the design specified a two-stage probability sample with schools as the first-stage sam- pling unit and students within schools as the second-stage unit. An average of 24 students in more than 1,000 public and private schools containing 8th grade students were sampled, yielding a sample size of approximately 25,000 8th graders. To date this sample has been resurveyed twice, in 1990 and 1992, when many of these youth were in the 10th and 12th grades, respec- tively. The study design calls for two more follow-ups, in 1994 and 1996, with the possibility of one or two additional follow-ups at two-year inter- vals (Davis and Sonnenberg, 1993). Thus, the current plans for the study involve following the NELS:88 8th grade cohort from approximately age 13 to at least age 21, and perhaps age 25. In each of the three waves of NELS:88, sample respondents completed a questionnaire and a cognitive test battery. The student questionnaire gathered information regarding family background, language use, school experiences, work experiences, and plans for the future. During the base year of the study, a parent or guardian of the sampled student was asked to complete a questionnaire, as were two of the sampled studentsâ teachers (in predetermined subject areas), as well as a school administrator. The parent questionnaire emphasized family background information, including paren- tal status measures and parental involvement in and knowledge about their childâs schooling; the teacher questionnaire examined teachersâ perceptions of the sampled studentâs academic performance and personal qualities, the curriculum content of the courses they taught, and their background and beliefs about their work environment. In addition to the student questionnaire and cognitive test battery, the first follow-up also included a school questionnaire (as many students changed schools in moving from the 8th grade to the 10th grade) and a teacher questionnaire, as well as a special questionnaire for school dropouts. The second follow-up included each of these kinds of instruments, as well as a parent questionnaire. Since recent developments in the analysis of school effects have docu- mented the importance of a âcritical massâ of students within schools for reliable estimation, the NELS:88 sampling design presented potential ana- lytic problems. In the NLS-72 and HS&B studies, the number of sampled students within high schools was fairly large, since high schools were the first-stage sampling units. But in NELS:88, the first-stage sampling unit was schools containing an 8th grade, many of which were middle or junior high schools. Consequently, many of the students in the study not only changed schools between the 8th and 10th grades, but also more impor- tantly were spread more thinly among a large number of high schools.
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 143 Thus, many of the high schools attended by NELS:88 sample members contain only a handful of respondents. Recognizing this problem, NCES agreed to sponsor a School Effects Supplement, adding additional students and teachers to 250 of the high schools in the study. Although there are no base-year data (e.g., cognitive test scores) on these additional students, they do provide additional information for estimating school effects on second- ary school experiences and outcomes. The National Education Longitudinal Study of 1998 NCES is currently beginning the planning of the National Education Longitudinal Study of 1998 (NELS:98). Although the study design has yet to be determined, including the starting grade for the cohort of youth to be followed over time, it is expected that NELS:98 will resemble NELS:88 in its design and substantive foci. As with NELS:88, and HS&B, the NELS:98 study will examine the process by which young people move through the schooling system and into adult work and family roles. The most likely sources of data will once again be students themselves, their parents and teachers, and school administrators. The Beginning Postsecondary Longitudinal Study As noted above, when students cross school boundaries they frequently are more dispersed, so that a sampling design involving sizeable clusters of students at one level of schooling almost invariably results in small clusters of students at a higher level of schooling. Nowhere is this clearer than with the transition from secondary school to postsecondary education. As a consequence, NCES has initiated the Beginning Postsecondary Longitudinal Study (BPS), a study of a large, nationally representative cohort of students beginning postsecondary education. The BPS sample is drawn from the National Postsecondary Student Aid Study (NPSAS), a triennial cross-sec- tional study of students enrolled in a range of postsecondary institutions, including less-than-2-year institutions, community and junior colleges, 4- year colleges, and universities located across the country. The sample in- cludes undergraduate, graduate, and professional school students, some of whom are receiving federally funded financial aid and others of whom are not. The purpose of BPS is to examine studentsâ progress, persistence, and attainment of postsecondary education and their transitions between postsecondary education and work. By relying on a beginning cohort of postsecondary students, BPS will capture the experience of ânontraditionalâ (e.g., older) students as well as those who enter postsecondary schooling immediately or shortly after completing high school. The first BPS cohort consists of about 7,000 students who began
144 INTEGRATING FEDERAL STATISTICS ON CHILDREN postsecondary schooling for the first time in the 1989-90 academic year and responded to the NPSAS:90 survey, which also gathered background and financial data from the parents of over 6,000 of these students. These students were resurveyed in 1992 and are scheduled to be followed up every two years. The base-year and follow-up surveys address studentsâ postsecondary schooling and employment experiences, including academic progress, field of study, persistence, and future plans; family formation; and financial aid and expenses associated with postsecondary schooling. In addition, BPS will gather postsecondary transcripts. NCES anticipates drawing new BPS cohorts from every other NPSAS cohort, i.e., on a 6-year cycle (Davis and Sonnenberg, 1993). NCES planning documents indicate that BPS is intended to focus on issues of progress and persistence in undergraduate education. Among the policy questions that BPS is intended to address are: How does the progress and persistence of nontraditional students differ from that of recent high school graduates? How does part-time or discontinuous attendance affect progress and persistence in postsecondary education? What is the modal rate of academic progress, and what educational experiences are related to making average progress toward completion? What is the match between postsecondary goals and attainments? What is the nature of postsecondary studentsâ access to graduate and professional programs? Although the planning documents are not clear on how long each BPS cohort will be followed, I suspect that the follow-ups are intended primarily to trace studentsâ progress through postsecondary schooling and not much beyond that. Thus, the BPS study is not likely to be an important resource for understanding the labor force careers of current enrollees in postsecondary schooling. However, because sample members will be followed even if they leave postsecondary schooling early in the life of the study, BPS may be especially useful in monitoring the early labor market experiences of those postsecondary students who enter the labor force upon completion of 2-year postsecondary programs or other programs of limited duration. The Baccalaureate and Beyond Longitudinal Study Paralleling the BPS effort is another new NCES survey tied to NPSAS, called the Baccalaureate and Beyond Longitudinal Study (B&B). The first B&B cohort consists of 16,000 students completing their baccalaureate de- grees in the 1992-93 academic year, a subsample of students surveyed in the cross-sectional 1993 NPSAS survey. The B&B study will resurvey this cohort at 1, 3, 6, 9, and 12-year intervals. For baccalaureate degree completers who are graduating at age 22, the follow-up period will span ages 23-34. Of course, some of the baccalaureate recipients in the B&B cohort will be older.
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 145 Because B&B will trace studentsâ educational and occupational attain- ments for 12 years past the receipt of the baccalaureate degree, the study will contribute to our knowledge about postbaccalaureate schooling, includ- ing professional education, and the early labor force experience of highly educated youth, including estimates of the economic returns to advanced schooling. Since relatively few of the young people sampled in the other NCES longitudinal studies ever attend postbaccalaureate schooling, B&B will provide new information about the dynamics of progress and persis- tence in graduate and professional education. As with BPS, the B&B study will rely on base-year NPSAS student and parent surveys, institutionally generated student records, and follow-up sur- veys of students. The data gathered will describe studentsâ family back- grounds, including the financing of undergraduate and graduate education, educational experiences in undergraduate and graduate school, employment experience and family formation, and future expectations. National Longitudinal Survey of Youth (NLSY) and Children of the NLSY An important limitation of the studies noted above is their inability to link data on change and continuity in the family status of youth with their educational and occupational careers. Such studies typically provide little or no information on household composition or economic status prior to the base year of the survey, at which time respondents were in their early or late teens. They cannot therefore inform questions about the influence of persistent poverty or marital disruption on the educational attainment of youth. A study that holds the promise of such analyses is the National Longi- tudinal Survey of Youth (NLSY). The NLSY is the fifth in a series of National Longitudinal Surveys of Labor Market Experience, currently ad- ministered by the Department of Laborâs Bureau of Labor Statistics. As with its predecessors, it is a longitudinal study of the labor market experi- ences of a nationally representative cohort of individuals, in this case more than 12,000 noninstitutionalized civilian and military youth ages 14 to 21 in 1979. The survey oversampled black, Hispanic, and low-income white youth and has had extremely high response rates in the base year and subsequent follow-ups. The sample has been reinterviewed annually since the base year, although the military sample was discontinued in 1985 (Manser et al., 1990). To date, then, the NLSY cohort has been followed from age 14-21 to age 28-35. As befitting a survey sponsored by the Bureau of Labor Statistics, the focus of the NLSY interviews is the employment experiences of respon- dents, including their sources of income, training, employment and unem-
146 INTEGRATING FEDERAL STATISTICS ON CHILDREN ployment spells, occupational mobility, and attitudes about and knowledge of work. Data are gathered on the dates of key events, a strategy that enables analysts to examine event histories. In addition, the interviews routinely ask about education and family formation and occasionally have included questions about child care, substance abuse, personal values, and career plans. The NLSY interviews have been supplemented by the High School Transcript Survey, sponsored by the National Center for Research in Vocational Education, and the Profiles of American Youth, sponsored by the Department of Defense, which administered the Armed Services Voca- tional Aptitude Battery to 94 percent of the sample (Manser et al., 1990). Perhaps the most remarkable feature of the NLSY is the recent expan- sion of the study to include the children of the NLSY. In 1986, nearly 5,000 children of the civilian women in the NLSY sample were assessed, and most of these were reassessed in 1988 and 1990 (Chase-Lansdale et al., 1991). Although these children are not a nationally representative sample, since their mothers were on average younger when they were born than mothers in general, as the cohort of mothers ages and more children are added to the sample with succeeding births, the sample of children will eventually be representative of all children born to women ages 14-21 in 1979. The sample size of older children remains rather small, but data re- ported in Chase-Lansdale et al. (1991) indicate that it now includes 300 youth age 17 or older, and in succeeding years this number will expand dramatically. Most of these youth will have completed a supplemental questionnaire intended to tap aspects of adolescent development, and younger children in the sample have already been assessed for their cognitive and socioemotional development with subscales of familiar psychological in- struments. Panel Study of Income Dynamics The Panel Study of Income Dynamics (PSID) is a longitudinal study of families begun in 1968. The study has followed both the original families interviewed in 1968 and âsplit-offâ families formed by individuals in those original families who subsequently left home and formed new families. These latter families include families formed when marriages dissolve and one or both partners form new households, and families formed when chil- dren leave home. Because the sample adds new families formed by chil- dren leaving home and discards families when all family members die, the resulting sample of families is unbiased (Duncan and Morgan, 1985). The PSID has conducted annual interviews with the head of the house- hold of all of the families in the sample, which in 1985 consisted of 6,500 families and 16,000 individuals (Duncan and Morgan, 1985). As might be
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 147 surmised from the studyâs title, much of the data gathered pertain to sources of income in the calendar year preceding each interview, as well as detailed employment histories for the head of the household and partner and less- detailed data on the employment and earnings of other family members. The core data also include information on family composition, household expenditures, and housing (Duncan and Morgan, 1985). Other data have been gathered on a one-time or intermittent basis. Topics that have been covered at least once in the PSID interviews include child care information, disability and illness, job search strategies, unemployment, retirement plans and experience, savings patterns, and standardized test performance. Because the core data are gathered annually, it is possible to form event histories for families and link them to the experiences of family members. The chapters in Elder (1985) provide examples of creative analytic efforts to model the consequences of variations in family experience (e.g., poverty status, marital stability, disability status) for families and their members. Of particular interest is the capacity to examine the effects of family back- ground on children in longitudinal perspective. There are, however, limits to the ways in which the PSID can inform our understanding of educational attainment and the transition into the labor force. Relatively few sample members are undergoing the transition to adulthood in a particular year, so that cross-sectional comparisons of the educational and occupational experiences of youth are not highly reliable. In addition, the data on family members who are not heads of households or their spouses are rather sketchy, so that data on adult children living at home are much thinner than data on similar youth who have formed inde- pendent households. Moreover, the data on children are reported by the household head rather than the children themselves, and thus are subject to some potential distortion. SUMMARY STRENGTHS AND WEAKNESSES OF THE AVAILABLE DATA Perhaps the most important point that one can make regarding the avail- able sources of data on educational attainment and the transition into the labor force is that it is hard to characterize these data sources as a system. The data collections are administered by different agencies, each with its own substantive foci, and little thought has been given to how these differ- ing data collections fit together or where the key redundancies and gaps might lie. No single organization has had the authority or the responsibility to manage the collection of information on educational attainment and the transition into the labor force. Consequently, the comparability of data gathered by different agencies, or even different branches within an agency, has been largely a hit-or-miss proposition. To be sure, communication has
148 INTEGRATING FEDERAL STATISTICS ON CHILDREN improved both across and within agencies, so that, for example, item word- ing is much less variable across surveys than it once was. Nevertheless, it is clear that the existing data collection mechanisms fail to inform many important policy concerns about educational attainment and the transition into the labor force. I believe there are three major sources of slippage between the data that are gathered by the federal system and the data that might be most useful to policy makers: what questions are asked, when they are asked, and of whom they are asked. What questions are asked? The foregoing discussion has suggested that a number of aspects of educational attainment and the transition to work are not well represented in the ongoing data collection system. Al- though the failure to ask the right questions may be seen as the source of the problem, it is important to understand that there are two very different reasons for the failure to ask the right questions. The first reason is what social scientists sometimes refer to specification error in models of social processes: the failure to include measures of theoretically relevant con- structs in attempts to model or understand complex social phenomena. Perhaps the most important model specification error in the studies reviewed above occurs when studentsâ cognitive ability or knowledge is not taken into account. Because student learning is frequently correlated with studentsâ schooling experiences and also is related to educational outcomes, the failure to take measured cognitive performance into account can lead to upwardly biased estimates of the effects of schooling experiences on such outcomes such as educational attainment. For example, the fact that youth who have been retained in grade in their primary or secondary school ca- reers are more likely to drop out of school than those who have not does not necessarily imply that grade retention âcausesâ dropping out. Rather, it may simply be that those students who are struggling academically early in their schooling careers are both more likely to be held back and more likely to leave school before completion. It would, however, be extremely diffi- cult to distinguish between these two possibilities without early data on studentsâ cognitive performance. The second kind of failure to ask the right questions stems from what social scientists refer to as measurement error. In this case, the relevant theoretical constructs are in fact represented in the data that are collected, but they are simply measured poorly. There is an ample number of ex- amples of poorly measured variables in the datasets described earlier. Edu- cational attainment, including when students were enrolled in school, and the kind of schooling in which they were enrolled, is measured unevenly in the studies I have reviewed, as is the quality of youth work experience. Even contextual factors such as family income or household configuration are subject to this criticism. When are the questions asked? It is axiomatic that the periodicity of
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 149 data collection needs to be calibrated with a theory of the stability or vola- tility of the phenomenon of interest. Phenomena that are known or pre- sumed to be stable do not need to be measured frequently, whereas phenom- ena that change quickly over fairly short periods of time may need to be monitored repeatedly over that burst of time. We measure the unemploy- ment rate monthly, because we know that those forces in the economy that affect employment and unemployment can change very rapidly, and in fact we do observe month-to-month variation in the unemployment rate that may exceed sampling variability. Conversely, a decennial youth unemployment measure would be of little use because it would likely miss most of the action from month to month and year to year. Yet it may not be meaningful to estimate a monthly dropout rate, because the forces that govern the drop- out rate are not believed to shift from month to month. Annual or biennial data might be more appropriate, as a year or two is often the length of time before a new program or policy initiative starts to affect students in the way in which it was intended. While I am not inclined to argue that any of the datasets reviewed earlier gather data too frequently, I do believe that there are instances in which the data collections are not frequent enough to allow for the mean- ingful monitoring of trends in educational attainment and the transition to work. In a context of relatively rapid educational reform and economic change, it may be desirable to address questions about high school comple- tion, postsecondary access and persistence, and youth labor market experi- ence as frequently as every two to three years. Those analytic questions depending on longitudinal data have typically relied on data from the NCES longitudinal studies, which have been spaced 8 to 10 years apart. This may simply be too long a time span to monitor changes in educational attainment and the transition to work. To cite but one instance of this problem, one of the objectives associ- ated with the Goal 2 of the National Education Goals pertains to the propor- tion of school dropouts who will return to complete a high school degree or its equivalent. Each of the three National Education Goals Reportsâthose issued in 1991, 1992, and 1993âhas reported data on the percentage of High School and Beyond sophomores in 1980 who dropped out but then returned and completed high school by 1986. These data were already old when they were initially reported, but comparable data from the NELS:88 study, the subsequent NCES longitudinal study, will not be available until 1996. Whereas the NCES longitudinal studies are exceedingly valuable, per- haps they need to be supplemented by more frequent short-term longitudi- nal studies focused on specific policy concerns, such as postsecondary ac- cess or the transition from high school to work. It may be advisable to assess trends in postsecondary access or persistence and their dependence
150 INTEGRATING FEDERAL STATISTICS ON CHILDREN on social background factors on a two- to three-year cycle. Such a hybrid strategy would retain the strengths of the NCES longitudinal studies pro- gramâespecially the extraordinary diversity in the range of data gatheredâ while still providing more timely information on a smaller range of policy issues. Since these âgapâ studies could not reflect the complexity of studies like HS&B and NELS:88, careful thought would have to be given to just what information is essential to gather to inform the most pressing policy concerns on a two- to three-year cycle. Of whom are the questions asked? Implicit in the purpose of a work- shop entitled âIntegrating Federal Statistics on Childrenâ is the presumption that such statistics can or should be integrated. Many of the important sources of data on childrenâs well-being are sample surveys of households and their members who are followed through time. Of the data sources I have reviewed, only the NCES longitudinal studies are sensitive to the organizational context of schooling and work. If, as I have argued, organi- zational and institutional contexts are important in understanding educa- tional attainment and the transition to work, then survey designs that are indifferent to these contexts may result in asking questions of the wrong people. For example, questions about the effects of schools and their policies and programs on educational and labor market outcomes may require a sufficient number of students within schools to estimate those effects reli- ably. Household surveys typically are not designed to ensure a critical mass or cluster of individuals within a particular organizational context, such as a school or a firm. Even the NCES longitudinal studies, which by design sample students within schools, run into difficulties on this account, as students disperse when crossing grade levels or move from one school building to the next. It may be that we will need to depend on a synthetic cohort approach drawing on different cohorts at different levels of the edu- cation system (for example, NELS:88 coupled with BPS or B&B) to under- stand organizational and institutional influences on the educational and work careers of youth. The important point here is, I believe, that the policy questions of interest should dictate study designs. Not all questions will be best ad- dressed by household surveys, and some may warrant relatively unconven- tional sampling plans. We might imagine, for example, sampling youth working in specific firms that hire young people (e.g., fast food restaurants) or working in different sectors of the economy, thereby building important contextual comparisons into the study design. There is an additional concern about who gets asked questions. Many of the important policy questions in the areas of educational attainment and youth employment pertain to the effectiveness of specific policies and pro- grams. For example, we want to know what works in dropout prevention
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 151 programs, or whether Upward Bound is successful in helping disadvantaged youth make the transition to postsecondary education. To what extent can the federal data system inform these questions? The answer is, only a little. The broad federal statistical data collec- tions that function as indicator systems are not a substitute for program evaluation. Evaluations of specific programs and policies typically require data that are tailored to understanding those programs and policies. Routine data collections are unlikely to ask the right questions of the right respon- dents. Targeted studies of particular programs and policies are a much more fruitful source of information about how such programs and policies work. For example, the NELS:88 study asked school administrators to de- scribe the dropout prevention programs in their schools and asked students both in and out of school about their experiences with such programs. But the questions are necessarily broad and vague and unlikely to delve suffi- ciently deeply into the characteristics of specific programs or into studentsâ experiences with them to have much policy relevance. We can learn little about how programs are implemented, who they serve, and how effective they are from the broad social surveys that make up the current federal data system. Conversely, there are other mechanisms for gathering information that might address such policy concerns. For example, the Office of Planning and Policy in the U.S. Department of Education is sponsoring an evaluation of the School Dropout Demonstration Assistance Program, a federally funded demonstration program involving 65 sites around the country, conducted by Mathematica Policy Research. The data gathered through this evaluation will not be linked to any of the data sources I have described, but they will bear directly on issues in the implementation and impact of model dropout prevention programs. COVERAGE OF SUBPOPULATIONS OF POLICY RELEVANCE In light of historic concerns about equality of opportunity, both educa- tional and otherwise, the datasets described previously are often used to examine the status of traditionally disadvantaged populationsâracial/ethnic minorities, the poor, women, and, more recently, language minority and disabled children and youth. Although the issue of the adequacy of the coverage of such subpopulations in these datasets is necessarily subjective, there are two general points to be made. First, some studies exclude policy-relevant subpopulations by design. The CPS, for example, excludes institutionalized youth, who are likely to be disproportionately poor and minority. Similarly, the NCES longitudinal studies have omitted early school-leavers by design, as the NLS-72 began
152 INTEGRATING FEDERAL STATISTICS ON CHILDREN with a cohort of high school seniors, and HS&B with cohorts of seniors and sophomores. These studies provided limited knowledge about school drop- outs in general and early school dropouts in particular. Second, given the designs and sample sizes of most of the data collec- tions reviewed here, subpopulations that are relatively rare in the population are unlikely to be studied reliably without substantial oversampling. Mi- nority racial or ethnic group members, language minority youth, and dis- abled youth represent relatively small shares of the youth population and thus may not be represented in study samples in sufficient numbers to esti- mate their experiences reliably. Some studies have oversampled several small groups of policy interest. For example, the NELS:88 study oversampled Hispanic and Asian and Pacific Islander students, as well as mainstreamed hearing-impaired students enrolled in Individualized Education Programs. Often, though, such sample augmentations are dependent on an outside agency providing supplemental funds. In the case of NELS:88, the Hispanic and Asian and Pacific Islander sample augmentation was funded by the Office of Bilingual Education and Minority Language Affairs in the Department of Education, and the sample of hearing-impaired students by Gallaudet Uni- versity. STRATEGIES FOR IMPROVING THE CAPACITY OF THE FEDERAL STATISTICAL SYSTEM TO ADDRESS INFORMATION NEEDS I conclude this brief review by asking how we might increase the ca- pacity of the federal statistical system to address the needs of policy makers for data on the well-being of children and youth. This is no mean task, given the historic fragmentation and politicization of the federal statistical system. I have no sure-fire strategies to suggestâsimply some provisional suggestions. 1. Conduct a needs assessment. A necessary first step is to understand the information needs of policy makers, the terrain of the available data, and the quality of the fit between the two. The workshop for which this paper has been prepared is an excellent first step in mapping out what is known and what is needed, and the selection of agency professionals, con- tractors, and academics as discussants broadens the base of the discussion. I believe the discourse over what is needed and what is available should be broadened even further to include agency executives in key policy-making roles (e.g., the assistant secretary for postsecondary education), legislative committee members and/or their staffs, the major advocacy groups (e.g., the Childrenâs Defense Fund), professional associations (e.g., the American Edu- cational Research Association), and the think tanks that represent a key
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 153 consumer group for federal data relevant to policy (e.g., the Urban Insti- tute). 2. Convene an interagency working group composed of the agencies responsible for the production, dissemination and analysis of federal statis- tics bearing on the well-being of children and youth to examine gaps, over- laps, and redundancies in the federal statistical system. Such a working group could take the proceedings of this workshop as a point of departure. In my view, it would be essential for the working group to have the neces- sary technical expertise to examine the plausible budgetary implications of alternatives to existing data collections. It also would be crucial to have the Office of Management and Budget represented in this working group, in light of its oversight role for the collection of federal data. 3. Lodge oversight responsibility for the federal data system as it bears on the well-being of children and youth in an individual or group that has the capacity to hold the agenciesâ feet to the fire and arbitrate among differ- ent agency interests. As I have noted earlier in this paper, the distinctive histories and constituencies of the various federal agencies that gather sta- tistics on children and youth must be taken into account in trying to under- stand why the system looks the way it does. Agencies occasionally are in competition for the responsibility to gather certain kinds of data, and perenially they are in competition, whether direct or indirect, for the scarce federal dollars devoted to statistics. There seems to be a need for an oversight body that can rise above the individual and sometimes competing interests of particular agencies and shape a federal data system that is responsive to the full spectrum of policy concerns about children and youth, particularly those concerns that cross traditional agency boundaries. These are some possible action steps, but I am uneasy about concluding this paper without questioning one of the fundamental assumptions that undergirds it and the workshop for which it was prepared. I believe that the federal statistics system can be a useful tool for policy makers concerned with educational attainment and the transition to work in particular, and the well-being of children and youth in general. But statistics, and the surveys and censuses that generate them, are just one part of a portfolio of data sources that can be useful. It is important to keep in mind that there are many important policy issues concerning the well-being of children and youth that are better addressed through delimited program evaluations, qualita- tive case studies, and other modes of social science analysis that do not produce statistical estimates with known precision. In the absence of pow- erful theories of how the education system and the labor market work, we need to draw on as many defensible sources of knowledge as we can. It would be a shame if attention to the statistics system were to diminish the