Purpose of paper. The purpose of this paper is to describe the concept of a longitudinal indicator and to present a conceptual framework for developing such indicators. The paper provides examples of existing longitudinal indicators, such as long-term poverty and welfare dependency. In addition, opportunities for creating new longitudinal measures for families are described. In some cases, it would be possible to develop new longitudinal indicators if just a few variables were added to existing data bases, and several opportunities for making such additions are presented as well.
Definition and importance. A longitudinal indicator is a measure of a family or individual behavior, interaction, attitude, or value that is measured consistently or comparably across multiple points in time and cumulated to provide a portrait over time of an important aspect of family life. Thus, a longitudinal indicator is one type of cumulative indicator. However, unlike cumulative indicators that tally experiences or characteristics at a single point in time, longitudinal indicators cumulate the presence (or the degree of the presence) of behaviors, interactions, attitudes, or values across time.
The key to a longitudinal indicator is not the way in which the data are collected; the data for such an indicator can be collected either retrospectively from a single survey or by combining data from longitudinal surveys across waves. Rather, the key feature is that it assesses a longitudinal construct. For example, instead of, or in addition to, assessing current family structure, a longitudinal indicator might assess the percentage of a child's life that the child has lived with both of his or her biological parents. As noted by Moore (1997), researchers can develop longitudinal indicators that reflect the duration of various family circumstances that children experience, including poverty or family structure, or the presence of chronic health conditions.
Longitudinal indicators can also sum non-consecutive spells in a particular status, for example, the amount of time that a family spends on welfare over a period of years. In addition, longitudinal indicators can be directional. For example, relationships between family members might be tracked over time and found to deteriorate or to improve in quality over time.
Because, by definition, longitudinal indicators assess aspects of families over time rather than at single points in time, they can be used to gain a more comprehensive understanding of how families are faring. Although we have not identified many studies that systematically compare longitudinal and cross-sectional indicators, we hypothesize that longitudinal indicators are likely to be more strongly related to family outcomes than are cross-sectional, or point-in-time, indicators. The reason for anticipating a stronger association for longitudinal measures is the greater exposure of children and families to a given influence over time, as well as the greater reliability of an indicator based on repeated measurements. Because family behaviors, interactions, attitudes, and values change over time, or may be present or absent for only brief periods of time, a cross-sectional indicator may "miss" detecting the characteristics that are most typical for a particular family by assessing that family at only one point in time. For example, a family may normally attend religious services, but a cross-sectional measure assessing attendance in the last month may miss their long-term attendance pattern, if the reference period happened to be an unusual month in which the family was not able to attend services. Furthermore, since family outcomes develop over time, characteristics that are consistent across the stages of the family's life cycle can be expected to have stronger effects on outcomes than would characteristics that are short-term or anomalous for a family.
Longitudinal measures can be particularly useful for learning about the strengths that families maintain across developmental periods of family life, because strengths tend to be enduring characteristics. For example, involvement in school activities tends to vary in frequency and type from elementary school through high school; but involvement over the years may be particularly supportive of children's school success. Additionally, they can assess over time both the qualities of relationships among family members (such as warmth), as well as family behaviors (such as religious attendance), both of which are aspects of family strengths.
On a different note, another advantage of longitudinal indicators is that, under some circumstances, they may be more cost-effective than cross-sectional indicators (Moore et al., 2002). Specifically, by spreading out the cost of assessing a family characteristic over time, rather than investing the resources necessary to gather detailed data for a cross-sectional indicator in each administration of a survey, costs may be minimized. For example, five or six items can be repeated over time and cumulated, in lieu of a 20-item scale assessed at multiple points in time. Indeed, Moore and her colleagues (2002) found that the reliability of varied shortened versions of the NLSY79 HOME-Short Form was comparable across a scale comprised of 6 items assessed 3 times and a scale comprised of 24 or 27 items assessed 3 times. Also, the strength of the association of the shorter scale with adolescent outcomes measured two years later was generally equivalent to the strength of the association of the long scale, even after controlling for social, demographic, and economic factors. Additionally, if longitudinal indicators have stronger predictive validity than cross-sectional indicators, the overall number of questions needed to assess a construct can be reduced. Thus, Moore and her colleagues found that the scale comprised of 6 items assessed at 3 points in time (cumulated for a total of 18 items) predicted delinquency and PIAT test scores (though not parent-child activities) better than a 27-item scale measured at one point in time.
Of course, there are some topics for which longitudinal indicators would not be useful. For example, we would not consider major dramatic events such as a hurricane or serious car accident to be in the same conceptual category as a longitudinal indicator. Such events represent important occurrences that may have significant consequences for children and families, so assessing them in other cumulative indicators, such as a measure of turbulence, would be informative. However, longitudinal indicators represent a particular sub-group of cumulative indicators that track ongoing involvement in a line of behavior, interactions, values, or attitudes to ascertain whether it represents a consistent aspect of family life over time.