Directions for Future Research
In his concluding remarks, Stephen Raudenbush highlighted several important messages from the data presented, as well as from the administrators of programs that target the needs of highly mobile students. Although the data are neither complete nor conclusive, he identified these propositions as emerging strongly from the workshop discussions:
Mobility is highest and likely to be most harmful among particular subgroups. Poor families move more than nonpoor families. Hispanic and particularly African American families move most frequently of all. There are consistent negative associations between moving and achievement and other outcomes for disadvantaged children, which are most pronounced for the children who move most and for special education students and English language learners.
Some kinds of mobility are more harmful than others. Moves made within districts are most likely to be harmful, as are moves made during the school year, rather than between grades. However, the reasons people move vary, as do their destinations. Mobility could have positive effects in some situations and negative ones in others. For this reason, the effects tend to average out in the context of large data sets, suggesting that mobility has little effect when averaged over heterogeneous populations. However, the impact may be quite significant for subgroups, even though these effects can be difficult to capture.
The greatest harm is associated with multiple moves. Children who moved three or more times in the first few years of school show the most negative associations. However, there is good reason to regard multiple moves over time not only as a clearly defined variable but also as a marker for a cluster of developmental problems and other risk factors. A high rate of mobility could be a contributing factor on its own, but it is consistently accompanied by other risks, such as family disruption, homelessness and economic disruption. It is difficult to disentangle the factors; for example, if a highly mobile child is also frequently absent from school, is that because of the mobility, or is it a sign of other underlying problems in the family?
High mobility in schools affects everyone. The best available evidence suggests that all children in highly mobile schools experience negative effects, even if they do not move themselves. The churning of students is likely to make instruction more difficult, to interfere with the continuity of programming, to necessitate more review, and to disrupt social networks.
Given this picture, Raudenbush said, “it is time to get past the question of whether moving by itself has an average effect in the total population of U.S. families.” Reasons for moving and circumstances are so heterogeneous that a fresh research agenda is needed to focus on the subgroups most likely to make the sorts of moves that have negative effects. The research priorities he listed are rigorous evaluation of interventions designed to: stabilize housing and therefore to prevent excessive residential mobility, to support school stability when many short-distance moves affect students, and to protect children against the negative impacts of residential and school mobility. Promising interventions need to be evaluated to make sure they can be faithfully implemented and their effects should be assessed, preferably using randomized experiments. Rent subsidies, adjusting school policies to help children stay in a school even when they move a short distance, coordination of instruction across local schools, coordination of family services, perhaps using the school as a central source—all look like promising approaches that merit detailed evaluation, he observed.
With that overview on the table, several presenters offered their perspectives on the major research questions, the methodological approaches they would use to answer those questions, and the strengths and limitations of those approaches.
Longitudinal and Early Childhood Data Sets
Donald Hernandez focused on what could be learned from new research with three large, nationally representative longitudinal data sets on early childhood—the Early Childhood Longitudinal Study-Kindergarten (ECLS-K), the National Longitudinal Survey of Youth (NLSY), and the Panel Study of Income Dynamics (PSID)—although he suggested that his observations could apply to other data sets as well. As mentioned in Chapter 2, the ECLS-K data cover the kindergarten class of 1998-1999, and a second round will sample the kindergarten class of 2010-2011. The NLSY sampled women ages 14 to 21 in 1979 and, later, the children born to those women. It collected information about variables including employment, education, training, fertility, health, attitudes, marriage and cohabitation, mobility, and crime. The PSID sample began with a sample of families in 1968 and has followed their children and grandchildren since then. The data collection focuses on income, employment, expenditures, housing, and program participation.
Each of these surveys allows for measures of mobility on a wave-to-wave basis, based on different data collection points. All three capture the number of moves reported between data collection points, the number of schools attended, residential histories, and addresses. Each covers somewhat different time spans in the lives of the children sampled, yet all collect extensive information about their education, achievement, socio-emotional functioning, behaviors, and health. Information about parents, demographic characteristics, and school environment is also included. The NLSY and the PSID also have data covering those sampled into their 20s and 30s.
Hernandez suggested that this wealth of information would be useful for moving beyond basic questions about the overall impact of mobility, as Raudenbush had suggested, to focus on processes that illuminate how and why mobility matters and for whom. High-priority research issues include
Whether and how a strong preschool foundation can moderate negative consequences of later mobility.
How socioemotional development and physical health are affected by mobility, and how these outcomes relate to cognitive development.
How a full range of ecological conditions, such as family composition, employment, parent-child interactions, school context and processes, and neighborhood context, interact with one another to influence outcomes for children in both positive and negative ways.
How the processes by which mobility influences important outcomes for children differ across population subgroups, including racial/ethnic groups, immigrant children from different countries, and socioeconomic groups.
How sophisticated statistical analysis might improve understanding of the effect of moves at different points in children’s development and multiple moves.
These three data sets, Hernandez explained, measure many of the variables of interest across large national samples, and they provide the basis for a new generation of intergenerational studies, in which state-of-the-art statistical methods can be used to build on the existing knowledge base. They have several limitations, however, Hernandez explained. The time period between data collections is sometimes long—two years or more—significant time gaps in the lives of young children. Even these large national samples are too small to study many racial, ethnic, and immigrant groups. These studies sample households and thus are likely to miss the homeless population. As with all longitudinal studies, these samples are likely to be affected by differential attrition of movers. Hernandez also emphasized that these surveys need to be augmented by qualitative ethnographic studies, comparative case studies, and place-based experiments, which are especially useful for studying the effectiveness of policy and programmatic interventions.
He closed with a reminder that large numbers of children in every social and economic group in the United States experience high mobility rates, and that longitudinal data sets have enormous promise for helping to make life better for them.
Using Qualitative Research
Greta Gibson echoed Hernandez’s final point, suggesting that small-scale qualitative and mixed-method studies can complement large-scale quantitative ones by isolating the risk factors for mobile children and the ways these factors interact in families, schools, and communities. “What, specifically,” she asked, “is bad about mobility for subgroups of students? What exactly goes on in their schools?” Her research, which is primarily on migrant students, has focused on what schools can do to support them. Schools that serve migrant children are often themselves low-performing schools that serve large proportions of low-income, students with limited English proficiency and limited resources. So it is important to understand and describe the differences between settings that promote inclusion and engagement and those that do not. Other important questions include
What characteristics of teachers foster success in meeting the needs of mobile students from linguistically and culturally diverse backgrounds?
Can whole schools or specific programs be identified that are successful in supporting mobile children—what are they doing, and would it be possible to replicate their success?
What is most important to creating a sense of belonging and community for students? Is it caring teachers, positive relationships, access to social capital, adults at school who serve as mentors or who forge links with families?
Gibson described her research with the federal Migrant Education Program. She noted that although the program has existed for 40 years, surprisingly little research has been conducted on its effects. She has followed a cohort of migrant students from ninth grade through high school completion. Questions about the definition of migrant, as well as these students’ mobility, made it difficult to maintain complete data. She found that the migrant students she followed have significantly lower achievement and graduation rates than nonmigrant students, and she used qualitative methods to explore their perceptions of school, sources of support, and other aspects of their lives in an effort to pinpoint the reasons for their academic difficulties.
Through this work, it has been possible to identify schools and programs that have been successful in supporting mobile children and to note that migrant education programs tended to “create spaces of belonging and connection that reinforce both academic success and identity.” Study of successful programs suggests that they had teachers who developed caring relationships with their students, served as role models, helped to bridge gaps between home and school, and acted as liaisons to other resources for migrant students. Gibson pointed out that these findings are consistent with the literature on social capital. However, opportunities for the kind of work she described have been limited, and she closed with a plea for more qualitative study of migrant students’ lives.
Using Administrative Data
Dennis Culhane, who studies homelessness, made the case for the value of administrative data. Primary data are very useful, he observed, but by the time one obtains funding, collects the data, analyzes them, and writes a paper, the policy questions may have changed. Administrative data may have selection issues, but one can study large numbers of observations and subjects longitudinally and can control for many factors. For example, the Kids Integrated Data System in Philadelphia, discussed
in Chapter 2, provides school, birth, and other administrative records for approximately 20,000 children up to age 19. With these data it is possible to follow the subjects through every public system they have touched; in the future, data about their earnings and postsecondary education will be collected as well.
Vast quantities of administrative data are collected every day, as part of the operations of various public services and institutions, and the kinds of data public agencies collect are generally applicable to policy questions. These data are generally more reliable than the self-reports that are used in many studies, and this is particularly important in the study of homelessness, because the kinds of information researchers want—about hospitalizations, days of truancy, and so forth—are particularly difficult for people to recall accurately.
One important disadvantage to using administrative data is access. A variety of federal laws govern the protection of individuals’ privacy in the context of schooling, social services, and health care. However, there are strategies to use the data effectively while observing these rules and protecting privacy, Culhane observed. There are also scientific challenges with these data, which are not collected for scientific purposes. Quality control measures are not the same as they would be in a major research effort, so data are often incomplete and inaccurate. Education data may not include private or parochial schools, for example, and it may be difficult to trace individuals’ trajectories as they move in and out of a jurisdiction.
Nevertheless administrative data provide an excellent way to track residential moves, homelessness and use of residential facilities, attendance patterns, use of special education services, disciplinary actions, achievement data, and school-based health records, for example. Even greater benefits come when these kinds of data can be linked with other social welfare data, such as foster care, juvenile justice, public assistance, mental health, and data on parents’ involvement with these systems. Such data can be used to develop a picture of individual risk factors, as well as to create aggregate measures of exposure to risk for children in a particular area. This is illustrated in Figure 5-1, which shows the results of a factor analysis that included crime, social stress, and structural decline (e.g., housing abandonment) in Philadelphia, providing a visual representation of the concentration of risk factors facing families in particular neighborhoods.
Evaluating Interventions That Aim to Reduce Mobility
Arthur Reynolds proposed a list of important research questions:
What are the basic predictors or determinants of different kinds of moves?
How do the impacts of mobility vary for different types of moves and by population subgroup, family structure, age of the child at the time of the move or moves, poverty status, and so on?
What are the long-term effects of mobility, for example, on dropout and later adult outcomes?
What is the nature of the relationship between mobility and outcomes—linear or nonlinear? Is there a threshold effect?
What is the effectiveness of such interventions as mentoring programs, changes in school district policies, and extra instructional support?
He used an example from the Chicago Longitudinal Study data (described in Chapter 2) to illustrate research that he believes is particularly valuable. He and his colleagues were examining the effect of number of moves on reading achievement, looking at achievement thresholds on the state eighth grade reading assessment. They found that one move cost students about two months’ worth of achievement, and that students who made three or more moves were five to six months behind their peers.
The study controlled for variables that might also affect achievement, and the findings were stable across a variety of model specifications. This threshold effect is also evident in long-term data, Reynolds observed, with students who had moved three or more times ending up by age 25 with a third of a year less education than their peers, even with other factors controlled. He suggested that the threshold effect of multiple moves—which shows up in many studies—is one of the fundamental issues that deserves further exploration.
Another issue suggested by several studies is that of developmental continuity. Preschool arrangements are often fragmented, for example, and the vast majority of children must change schools when they move from preschool to kindergarten. Other structural arrangements exacerbate mobility, rather than working to reduce it. Locating preschools in elementary schools and aligning the curricula, leadership, and supports has shown promise as a way of reducing children’s total moves during the early years of school. He also noted that cost-benefit analyses estimate a 15 percent return on investment (by age 26) for interventions that reduce mobility.
The bottom line for Reynolds was that annual data collection on school mobility at both the national and state levels would be a valuable tool for understanding mobility.
The volume of information and ideas that were aired in the course of the two-day workshop was almost overwhelming, one participant observed. Many added their research priorities to the discussion, and one suggested a theme for considering the way forward: capacity building. While each aspect of researching and addressing mobility that was discussed holds promise, all seem to require further development or more resources. Social agencies need to do a better job delivering high-quality integrated services; schools need to provide programs and opportunities for hard-to-serve students. Parents and teachers need support to help stabilize children’s lives and enhance their resilience when they are forced to move. Researchers need not only funding but also, in many cases, increased technical capacity to use sophisticated, perhaps mixed-method designs to tackle the difficult questions that don’t fall neatly into place. This might include designs that consider multiple levels of analysis such as exploring the effects of individual child mobility as well as classroom-level mobility.
Others built on this theme, suggesting, for example, the need for a national mandatory administrative data set. Also cited was the importance of embedding top-quality research in real-world settings: “A disap-
pointing amount of child development research has led us astray because it has taken place in university labs and lab schools that are completely unrepresentative of the general population.”
A related point was the need not only for interdisciplinary research, but also for cross-cutting system solutions. One participant observed that “our education systems are set up in districts, our health systems are set up in counties. We have all these little boundaries and though the mobile families are not even going very far, they are just drifting around across the boundaries.” The methodology for measurement is another research capacity that was identified as needing strengthening. As someone noted, “We are talking as if we could easily, given unlimited funds, go out there and measure kids really effectively. But it is kind of difficult to measure preschool development. There isn’t a strong consensus on how to do that.” At the same time, many of the measures that are typically used with language-minority and immigrant families have never been validated on these samples.
And from a policy perspective, there was a plea for additional data on the magnitude of the problem of mobility. “It would be really interesting to see, for example, what portion of Title 1 schools have 20 percent or more of students who are highly mobile.” Research that more fully documents the impacts of mobility in the older years could encourage policy makers to develop a balanced view of the entire developmental trajectory.
After the more free-ranging discussion of research priorities, Russell Rumberger and Sandra Newman offered their concluding thoughts. Rumberger observed that better methodologies, techniques, and statistical modeling are necessary to identify causal effects. At the same time, it is difficult to isolate mobility as an independent factor. It’s as much a symptom as a cause, he suggested, “and it’s when you study families that you discover the complexity of the phenomenon.” Many of the developments in families that become visible to researchers when some or all of the family relocates cannot be detected using most of the approaches discussed at the workshop. Turning the question around, to consider the many sources of instability in a child’s life and look for ways to deflect them, might be more productive than working endlessly to isolate mobility as an individual factor. For example, if one views mobility as an individual problem for families and children, it is easy to overlook the very important roles that school policies (such as attendance and discipline policies, school closures, and so forth), housing policy, and other institutional decisions play in undermining family stability. Similarly, focusing on broader, more long-term developmental and health outcomes helps to open up understanding of mobility as part of a bigger picture of the factors that affect family stability.
Newman reiterated some of the key findings from the workshop: that moves are not all equal, that the timing of the move seems to matter, and that subgroups experience and react to moves differently. Departing to some degree from Rumberger’s comments, she suggested that the question of whether mobility itself is a unique cause of negative outcomes for some students, net of other factors, is not yet a settled question. For her, the key will be to make further progress in understanding the mechanisms by which mobility causes harm and the conditions in which it is most harmful. It is important to fully examine each aspect of the contexts in which distressed children live. Otherwise, the risk is that “big investments are made in improving the quality of schools in poor communities but the community remains distressed and poor, unsafe, and not a good environment for children.” She ended with a point very similar to Rumberger’s: that to make a difference in children’s lives it will be necessary to address every aspect of their circumstances.
A few final comments brought the discussion to a close. The research on mobility is provocative, rich, and complex. It is emerging, but is still in a fairly immature stage of development in that it lacks rich, robust theories, tailored measurement tools, and sample populations that target the most important questions. Nevertheless, there is a compelling interest in using the research to shape policy and practice. Mobility as a phenomenon has been everywhere and nowhere—there is no single agency that has the lead role in addressing it. Indeed, mobility has in a sense often gone by another name in research and policy discussions: attrition. Children who are missing are difficult to track and to measure.
There is a tension between viewing the glass as half empty or as half full. Policy makers may be poised to address what is clearly a significant problem for large numbers of children and families. For that audience it is important, perhaps, to use the best available knowledge from research. There is indeed a great deal of information about mobility, and policy makers have been slow to recognize the multiple needs of these children and young people. At the same time, however, researchers recognize that many questions still await answers.