Based on the ecological model and a lifespan perspective, a number of critical constructs can be identified and assessed using this longitudinal perspective. In some cases, as described above, previous research has established an association between a longitudinal measure and family or child outcomes. In other cases, it would be valuable to conduct such research.
Surveys that could provide longitudinal indicators
A number of longitudinal surveys are being conducted. Most focus on individuals rather than families; however, many of these contain considerable information about the family. For example, the National Longitudinal Survey of Youth, 1997 Cohort includes information about the parent's religiosity and marital quality, and the Early Childhood Longitudinal Study cohorts provide information on family involvement in cognitively stimulating activities. In addition, several surveys have specifically collected some kinds of information for all family members over time, for example, the Panel Study of Income Dynamics.
- The National Longitudinal Survey of Youth, 1997 Cohort, for adolescents 12-16 in 1997 (NLSY97)
- The Early Childhood Longitudinal Study, Kindergarten Cohort (ECLS-K) and Birth Cohort (ECLS-B)
- The Panel Study of Income Dynamics (PSID) and the embedded Child Development Study conducted in 1995 and 2002 (PSID-CDS)
- The Survey of Income and Program Participation (SIPP) and the Survey of Program Dynamics (SPD)
- The National Educational Longitudinal Survey for students who were in eighth grade in 1988 (NELS)
- The National Survey of Families and Households (NSFH)
- The National Longitudinal Survey of Adolescent Health (Add Health)
- The Education Longitudinal Study of 2002 (ELS:2002)
Proportion of a child's life spent with both biological parents. Data from the PSID are available since the 1960s, enabling researchers to develop a truly long-term measure of family composition.
Proportion of an adult's life since first marriage spent in that first marriage. Data from the PSID could be analyzed to create this longitudinal measure. The Chartbook includes a point-in-time indicator based on data from the PSID that also taps the concept of stability in family structure: the percentage of families that experienced a change in family structure during the past two years. This indicator could be assessed over a longer period of time, as well, either by measuring the length of time in which families do not experience a change in family structure or by assessing the number of family structure changes over a longer period of time than two years. In addition, retrospective life history information from the National Survey of Family Growth could also be used to construct such a history for women in the 1995 survey.
Father involvement over time. The Chartbook includes a point-in-time indicator assessing children's contact with non-residential parents based on data from the April Supplement of the Current Population Survey (specifically, among children who have an absent parent, the percentage with any contact with nonresident parent in the previous year). This indicator could be expanded with a focus on fathers so that, from the time a child is born, the number of months or years that the biological father lives with his child could be calculated. A broader measure of involvement could include both co-residence and regular contact with the child (e.g., at least weekly). Both longitudinal measures could be created for short periods of time using SIPP, and a longer-term measure could be created using the PSID.
A richer measure of father involvement would describe the extent to which fathers are engaged in activities with their child and help care for their child. Data on fathering activities for representative national samples of fathers are very scarce at present; however, a set of variables has been included in the NLSY97. As the young men in that study have children, it will be possible to develop a longitudinal measure of father interaction and engagement with his child. In addition, involvement of fathers who are residential can be assessed in the Early Childhood Longitudinal Study, Birth Cohort (ECLS-B), which will be available for analysis in early 2004. (Data are also available for non-residential fathers, with the caveat that the response rate for non-residential fathers was lower than for residential fathers.)
Non-residential father involvement could also be examined using data from the Education Longitudinal Study of 2002 (ELS:2002), which includes a brief question on contact with non-residential parents, as well as a series of questions on the non-residential parent's involvement in the child's school. One drawback of the ELS:2002 is that, although it is a longitudinal survey, data collection begins when adolescents are in tenth grade. However, questions in the ELS:2002 could be added to other surveys that track families with children over longer periods of time.
Turbulence. Multiple changes across domains of life can result in substantial turbulence for families. These kinds of changes can include family composition changes, residential mobility, changes in school or child care arrangements for the children, and periods of unemployment for the parent(s). (Turbulence can affect any area of the social context of families, depending on what type of turbulence is assessed.) A retrospective history was obtained from the parent in the first year of the NLSY97, from which a cumulative measure could be created, and updates can be created using the annual survey data. A more short-term measure can be created using SIPP data. A measure focusing on school changes could be created based on the ELS:2002, which asks retrospectively about school changes since the first grade.
Marital quality. Identifying couples who not only remain together but who avoid high levels of conflict and sustain high levels of satisfaction would represent a valuable indicator. Research indicates that outcomes are better for children raised by couples in a low-conflict marriage (Seltzer, 2000). Research indicates that this association would be even stronger if it were measured over time (Peterson and Zill, 1986). At present, this construct is not assessed regularly in an ongoing longitudinal survey.
Consistently positive parent-child relationships. Strong parent-child relationships are regularly found to predict positive outcomes for children (Resnick et al., 1997), and this association holds even when numerous background factors are statistically controlled (Hair et al., forthcoming 2004). The Chartbook includes two point-in-time indicators assessing positive parent-child relationships: the percentage of adolescents with positive relationships with their parent, and the percentage of parents of children under age 13 who expressed various forms of warmth and affection to their child every day in the past month. These indicators could be extended so that the consistency of such positive relationships over a number of years is assessed. However, one challenge with assessing these indicators is that, at present, there is no single dataset that assesses positive parent-child relationships for children of all ages. Rather, the Chartbook used two data sources: the PSID for younger children, and the NLSY97 for older children. In the NLSY97, data provided by adolescents are available in every wave on their relationship with their residential mother, residential father, and their non-residential mother and father (if any), which could be cumulated to produce a measure of a consistently positive relationship, e.g., a relationship that is consistently in the top third or that is consistently rated as a 3 or a 4 on a scale of 0 to 4.
Consistent parental monitoring. In addition to supportive parent-child relationships, another parenting behavior that has been linked with positive outcomes for children is parental awareness of adolescents' friends and activities (Child Trends and the Ohio State University Center for Human Resource Research, 1999). This measure is included as point-in-time indicator in the Chartbook, and could be assessed longitudinally as well.
Consistent authoritative parenting. Authoritative parenting is measured by a compound variable that includes warm and supportive parenting combined with firm and consistent discipline (Baumrind, 1966). A number of studies have found that children whose parents engage in authoritative parenting develop better (Baumrind, 1966; Maccoby and Martin, 1983), though a few authors have cautioned that more controlling parenting, combined with warmth, is more common and could be more beneficial in some low-income families (Brody and Flor, 1998). A brief measure of authoritative parenting was included in the NLSY97 and repeated each year for adolescents who were 12 to 14 in 1997. These items could be cumulated to produce a measure of consistent authoritative parenting. (Analyses could also explore the possibility that consistent firm but not necessarily warm parenting is associated with positive outcomes for adolescents in lower-income families.) Authoritative parenting could be measured longitudinally, assessing the consistency of parents' supportiveness and strictness over time. Such a measure could be created from the NLSY97.
Communication and family routines. Communication and family routines are two additional aspects of family functioning that could be assessed longitudinally, using questions from the National Longitudinal Survey of Adolescent Health (Add Health) and ELS:2002.
Family, Work and Child Care
Child Care. Child care arrangements vary over the life of a child and family, so a longitudinal measure in this case would involve stable arrangements over a period of several years. Accordingly, data from SIPP (or ECLS-B, in several years) could be explored as a source of such information.
Consistent employment and a steady income. Downward dips in income and periods of unemployment pose challenges to any family. Examining whether and how many months such negative experiences occur could be explored with SIPP or PSID data.
School Involvement and Civic Engagement
Parental involvement in their child's school. Data from NELS and ELS:2002 provide information about parental attendance at meetings and other forms of involvement over several years and could be cumulated to indicate the proportion of parents who remain involved in their child's education over time. For younger children, data from the ECLS-B Kindergarten Cohort could be analyzed to create a similar type of measure for elementary school children.
Attendance at religious services. Annual data from the NLSY97 can provide insight into consistent attendance for adolescents aged 12 to 16 in 1997. Patterns of involvement during the teen years have never been examined to our knowledge, despite strong correlations between religious involvement and positive outcomes for children (Bridges and Moore, 2002) and adults (Sherkat and Ellison, 1999). The ELS:2002 also includes a question on the frequency of religious attendance of a parent together with his or her child.
Participation in activities. As noted above, a measure of consistent participation in activities during the high school years has been created using data from NELS (Zaff et al., 2003). While it might be desirable to track consistent exercise and healthy habits (Harris et al., forthcoming, 2004), data availability poses a barrier. Some information has been collected in the National Longitudinal Survey of Adolescent Health, but the first two waves are just a year apart, and the third wave is about five years later.
Additionally, the ELS:2002 includes questions on involvement in a variety of activities, such as sports, hobbies, clubs, and band or orchestra. Questions from the ELS:2002 could also be used to assess civic engagement and volunteering.
Disconnected youth. For a longitudinal measure of disconnected youth, an indicator could asses the proportion of youth who, over time, are neither working nor in school, nor married to someone who is. The NLSY97 could serve as a data source for such an indicator.
While it would be possible to create a number of longitudinal measures, several methodological issues need to be acknowledged. One important issue is the periodicity of data collection. Many longitudinal surveys are not conducted annually, and even when they are, an annual survey still does not provide a continuous record of family experience. For example, family activity patterns may vary across the seasons or across the school year, making it hard for respondents to count up or provide an overall average. Also, respondents may have trouble remembering the dates of events over the course of a year, or, they may recall only the major events or changes. In addition, attitudes and values may not be recalled accurately. Thus, the quality of relationships between family members is one example of a construct that cannot be recalled with precision and that might benefit from being assessed quarterly or even more often.
On the other hand, respondents can probably report changes in many constructs, such as employment, fairly accurately, particularly when respondents are provided with a calendar to aid their reporting. Unfortunately, while respondent recall has been studied extensively for income, leading to the fielding of the Survey of Income and Program Participation every four months, no knowledge base exists that enables us to identify the degree to which the quality of non-economic data is affected by using longer versus shorter recall periods, so it is impossible to recommend minimum acceptable periodicities for assessing various constructs. We note, however, that the use of calendars may aid respondents in reporting on constructs that are liable to change throughout the course of a year. Additionally, while more frequent assessment can certainly improve data accuracy, we acknowledge that the huge expense of increasing the periodicity of surveys will typically outweigh the benefits of marginal improvements in data quality.
Also, some constructs are difficult to measure across the life cycle. In particular, constructs that need to change as children become older are difficult to measure as longitudinal indicators. For example, a measure of parental awareness and monitoring has been included in the NLSY97; however, at present we lack the knowledge base to develop an age-adjusted indicator of appropriate monitoring across ages 12 through 17. A cross sectional indicator, or a short-term indicator, could be developed for such constructs; but truly longitudinal indicators would require some development.
Another concern is posed by attrition from longitudinal surveys. If surveys experience high levels of attrition, then indicators developed from such surveys may not be representative of the society. Fortunately, most federally sponsored surveys have fairly high response rates; but it is nevertheless necessary to be alert to this possibility.
Assessing the way longitudinal indicators change in the population over time also represents a challenge. Other population changes (for example, immigration), in addition to sample attrition, render the remaining sample less representative of the general population with similar characteristics. While longitudinal surveys are the ideal source for longitudinal indicators, repeated cross-sectional surveys are the most appropriate for trend analyses. This is because longitudinal surveys typically follow one cohort over time (an exception is the Panel Study of Income Dynamics), but trend analyses compare different cohorts across time. For example, longitudinal measures could be created using consecutive waves of the National Longitudinal Survey of Youth, 1997, which collects data every year on adolescents who were ages 12 to 16 in 1997. However, analogous data sources do not exist to compare longitudinal indicators for 12- to 16-year-olds in 1997 to longitudinal indicators for 12- to 16-year-olds in 2000 or 2003. For trend analyses, one possible solution could be to collect retrospective data in repeated cross-sectional surveys. The quality of data from repeated cross-sectional surveys using retrospective measures would not be as high as with longitudinal data, for the same reason that data quality is not as high as would be ideal even with longitudinal surveys that are conducted annually or less frequently. Specifically, the longer the recall period, the greater the opportunity for recall bias or simply for respondent errors in reporting.