6

Issues in Research Design and Analysis of Migrant Integration

ENVIRONMENTAL DISPLACEMENT AND MIGRANT INTEGRATION

Fernando Riosmena (University of Colorado Boulder) focused on methodological lessons learned from the study of immigrant integration in destination societies from his work on health-related integration and

on environmental migration. His presentation also addressed how the assessment of integration can be further complicated when studying a highly heterogeneous population in which some people are displaced by shocks (e.g., environmental shocks) while others are not. Riosmena began by explaining that a relational perspective, or “gap” approach, is needed to understand whether migration is likely to have produced an observed change or outcome in a population. In particular, Riosmena described two major questions that can be answered by this approach. First, what are the differences in the experiences and outcomes of a migrant population from what they would have been if they were perfectly “assimilated” and had the same structure of opportunities as natives (using the outcomes of a comparable receiving population to make this determination)? Second, what might the health of the migrant population had been if they had not migrated (using the outcomes of a comparable sending population)? Selection is an important factor in answering these questions, Riosmena observed, as different health trajectories due to selection could easily be confounded with differential trajectories of the migrating population relative to sending- and destination-area populations. Adjustments to reflect selection are then necessary to understand the differences between groups within and over time.

The relational approach also makes it possible to compare across different waves of immigration, further revealing the effects of selection. For example, the so-called “golden exile” of Cubans to the United States following the Cuban Revolution of 1959 suggests that selectivity and the modes of incorporation might be very different across generations, said Riosmena.

Examining selection requires combining data from origins and destinations. As an example, Riosmena looked at data on height, health, and hypertension drawn from Mexicans living in the United States, Mexicans who returned to Mexico, and nonmigrants in Mexico (Riosmena et al., 2013). These data reveal that migrants from Mexico to the United States had an immigrant advantage relative to non-Hispanic Whites in hypertension and, to a lesser extent, obesity. Riosmena and his colleagues found evidence consistent with emigration selection in height, hypertension, and self-rated health among immigrants with less than 15 years of experience in the United States. They did not find conclusive evidence consistent with sociocultural protection mechanisms.

A more precise comparison of changes in immigrant health with appropriate counterfactuals in both sending areas and destinations further allows for the separation of selection from postmigration changes. Riosmena provided an example from data on smoking among migrants to the United States from Mexico (Riosmena et al., 2017). Comparisons between migrants from five major national-origin groups, a non-Hispanic White counterfactual, and people who stayed in the sending country revealed statistically significant and substantial self-selection and protection among migrants to the United States.

Riosmena then discussed several issues related to the challenges of identifying displacement. It can be difficult to pinpoint or impute displacement for slow-onset events, he observed. In Mexico, for example, displacement related to climate change is embedded in historical flows that have occurred for other reasons; it is difficult to identify them, let alone identify them in a fashion that would allow understanding and distinguishing them and understanding the integration outcomes later on. Events such as hurricanes do displace people rapidly, he observed, but in Mexico, drought is a greater issue, making it difficult to distinguish between who is being displaced and who is migrating for other reasons. Droughts that occur in Mexico seem to be associated with migration to the United States, but it is difficult to identify this factor in a sample of U.S. migrants. A two-step approach can look for the characteristics of people most likely to have been displaced by climate change (Riosmena et al., 2018). But in the large population-based samples used in estimations, that is difficult without data on intentions or motivations to migrate.

A relatively simple solution to this problem would be to ask migrants where they are from. Gathering simple migration histories could make a big difference, Riosmena said. A quasi-probabilistic approach could help identify who is displaced and the role of selection in displacement. For example, Hunter and Simon (2017) used this approach to find that selectivity is stronger in places subject to environmental strains. However, doing this for integration outcomes is more difficult. He noted that understanding sending-area populations is particularly important to assess the impacts of migration on the displaced, even if displacement situations make this difficult.

Riosmena noted that some studies have been able to measure both displacement and integration in ways that have proven fruitful. Using the example of a study by Torres and Wallace (2013), Riosmena noted that the National Survey of Latinos and Hispanics in 2002 and 2003 was an exception to the above observations in that it asked migration circumstance questions, including whether people had to migrate, whether their migration was planned, and how much stress was related to their migration. Riosmena and a group of colleagues also have been working on ways to measure who is immobile, speaking to the idea of “trapped populations” that has been put forth for at least a decade (see, e.g., Foresight, 2011).

ESTIMATING THE IMPACT OF NEWCOMER PROGRAMS ON CHILD MIGRANT EDUCATION AND ACCULTURATION

Aimee Chin (University of Houston) noted that 1 in 10 children in U.S. public schools is considered limited English proficient and needs special instructional support to get the same educational opportunity as other

students. English language learners are dispersed across the United States, not only in traditional immigrant receiving hubs. About half are foreign born, and only some of them are forced migrants.

Schools have taken a variety of approaches to help these students learn English, such as bilingual education programs with some native language instruction and English-only approaches with English as a second language instruction combined with structured immersion for content teaching areas. Forced migrants have additional considerations that can require more than help with English acquisition, Chin observed. Many of these students have interrupted formal schooling, such as a teenager who has attended only primary school in a home country, or the quality of their schooling may have been low. These students often arrive with trauma, whether war, violence, or a natural disaster that led them to migrate, or they may have experienced trauma during the migration process. They may be separated from their families, a family member may have suffered violence, or their parents may be in precarious legal situations. Adolescents can be especially at risk. It can be more difficult for them to learn a new language, and they are more likely to have gaps in their education if they are arriving in middle school or high school.

To address these issues, school districts across the United States have introduced newcomer programs to provide extra support (Short and Boyson, 2012). Usually one semester or 1 year long, they typically combine intensive English acquisition with guidance on behavioral norms and interactions at school. Most programs are full day, and some offer counseling and community partnerships to connect students and their families to additional services and resources, including other people in the community from similar cultural backgrounds who can help them acclimate. They may separate students onto a different campus or teach them within a larger school. Such programs seek to provide a bridge to other English language learner programs and to regular schools and programs.

Relatively little work has been done to rigorously estimate the impact of these programs, Chin reported, and such work faces challenges. In particular, comparing the outcomes of students who participated in newcomer programs with the outcomes of students who did not is unlikely to produce causal effects because participants likely differed from nonparticipants in systematic ways. To overcome this problem, she and her colleagues have been developing a plan to study a particular school that has features enabling an effective research design. The school serves grades 6, 7, and 8, and in 2013 it expanded to include grades 4 and 5. Students take entry exams and are admitted only if they score below a certain cutoff—typically 7th- or 8th-graders with English skills at 1st grade or below. Students usually stay in the school for 1 year, though students who are considered to have interrupted formal education may stay an extra year.

Chin’s team plans to use administrative data collected by the school district as part of its implementation strategies. These data, which contain demographic, attendance, coursework, and test score information on all students in the district, can be used to construct longitudinal profiles of migrant students that follow them from entry to later years. The outcomes examined will include attendance, suspensions, test scores, and academic track in the shorter run and high school completion, college attendance and degree completion, and earnings in the longer run. The advantages of using administrative data are that these data are collected for other purposes and would be difficult to get through surveys. A limitation is that, ideally, richer measures of student social and well-being outcomes would be available.

The extension of the program to grades 4 and 5 in 2013 suggests a difference-in-differences strategy to estimate the impact of the policy change, Chin observed. Specifically, the proposed difference-in-differences estimate of the effect of the newcomer program is the after-2013 and before-2013 difference in the outcome among the newcomer students arriving to the United States in grades 4–5 (considered the treatment group as their eligibility for the newcomer program is changed by the expansion) minus the after/before difference among the newcomer students arriving in grades 6–8 (the comparison group; the expansion does not change their eligibility as they were always eligible). This estimate gives the policy impact under the parallel trend assumption—that is, in the absence of the policy change, the arrival cohort’s changes in outcome would have been the same for the grades 4–5 newcomers and for the grades 6–8 newcomers. The data from before 2013 can be used to check the plausibility of this parallel trend assumption, as can data from grade 3 and grade 9 newcomers who did not experience the policy change. In this way, the researchers will be able to assess the causal effect of the program expansion.

A second strategy is to use the eligibility rules that limit the program to students who score below the cutoff on the initial English skills. This suggests a regression-discontinuity-based strategy comparing those just above and just below the cutoff score. This strategy turns out to be difficult to implement because the researchers do not have data on who applied to the school and what their scores were on the admission test. Chin suggested an alternative approach is to use a different test of initial English skills and to compare students with slightly higher and slightly lower test scores. In practice, in part due to sample size and the initial English skill measure that is available, it is not possible to limit analysis to a very narrow neighborhood around the cutoff score. Further, with broader ranges, it is simply not plausible that in the absence of the newcomer program, the outcomes of students below the cutoff would have been the same as the outcomes of students above the cutoff. In recognition of this, Chin noted that the researchers cannot implement a standard regression-discontinuity design in this

context, but they can use this cutoff criterion in combination with another dimension of program take-up to estimate the program impact. Specifically, they take advantage of the fact that students who live closer and farther away from the school have a different likelihood of attending it. The more distant students can be used to “difference out the difference” that would exist anyway between the high-scoring and low-scoring students, thereby separating out the effect that is due to the newcomer program eligibility.

During the general discussion, Ellen Kraly pointed out that people working on the ground focus on their own programs rather than putting them into a broader framework, which produces complications in making a program scalable or relevant to policy. In response, Chin observed that her study is designed to show whether the program works and should be scaled up to other schools. She noted that the school district itself has been debating whether to offer this program in more locations in the district, and that other school districts in the country are adopting or considering adopting this particular model, and the study should have relevance for these decisions. However, she recognized there could be limits to the applicability of the results found for this program to other settings. Chin indicated that the principal at the pilot school is “really special,” which raises issues of whether this model can be replicated elsewhere with more typical leaders. On the other hand, the school has experienced teacher turnover, and the principal has trained new sets of teachers to implement this newcomer program. It does seem possible to replicate this program at other schools, Chin said, although it seems important to have administrators and teachers who are excited to work with this population.

RESILIENCE, RECOVERY, CULTURE, AND ACCULTURATION

Mark VanLandingham (Tulane University) described a project called Katrina@10, which he co-leads with David Abramson at New York University and Mary Waters at Harvard University. A set of five studies is looking at differentials in long-term recovery from Hurricane Katrina, and he and his colleagues at Tulane University are focusing on a Vietnamese community in the eastern part of New Orleans. Many Vietnamese were attracted to New Orleans after the Vietnam War ended in 1975. The first wave were often elites who had worked with the French (and Americans) in Vietnam and were predominantly Catholic, which facilitated their settlement in New Orleans, another major outpost in the French colonial empire. The community became a major enclave in eastern New Orleans, which was heavily flooded during Hurricane Katrina in 2005. Leveraging data that he and his colleagues had collected in that community for a different purpose just prior to Hurricane Katrina, VanLandingham decided

that if they worked quickly they could use these data to build and follow a cohort as a case study of how immigrant communities fare after a major disaster. They used five standard dimensions and measures of recovery: housing stability, physical health, mental health, economic stability, and social role adaptation.

The bottom-line conclusion, VanLandingham said, is that the Vietnamese community recovered much better than other communities on four of five of these measures of postdisaster recovery (VanLandingham, 2017).¹ The Vietnamese returned to New Orleans much more quickly than did Blacks and Whites, even though they experienced similar levels of exposure to flooding and other effects of the disaster. The Vietnamese also experienced much lower levels of posttraumatic stress disorder (PTSD) and had better outcomes on other health measures (Norris et al., 2009; Vu et al., 2009; Vu and VanLandingham, 2012).

Why would that be, he asked. Most disaster researchers who address this question use a concept known as resilience, described by Norris and colleagues (2008) as a process of “linking a set of adaptive capabilities to a positive trajectory of functioning and adaptation after a disturbance.” The dimensions of resilience include economic development, social capital, information and communication, and community competence, which is essentially a measure of the skills present in the community. However, he said, the Vietnamese community did not stand out on any of these measures; the model did not match the data.

VanLandingham suggested that culture was the missing piece. The Vietnamese in New Orleans shared a common history and a common set of events that their non-Vietnamese neighbors did not share, which gave them a view of the world, or a way of interpreting it, that distinguished them from the other groups. He said that recent research on culture has typically employed qualitative approaches, either for conceptualizing or measuring culture, and that work that leverages quantitative methods is much less common. Researchers also steer clear of comparative work because of a reluctance to appear to be disparaging one culture while lionizing another. This reluctance is understandable, given earlier initiatives that used culture to blame people for their own circumstances. One final difficulty in employing culture in an explanatory model is that culture can easily be conflated with selection effects.

VanLandingham said that he and his colleagues relied heavily upon anthropologists’ views of culture as symbolic systems of beliefs, values, and shared understandings that render the world meaningful and intelligible for a particular group of people (Birx, 2010). Recently, he continued, emphasis in the field has moved away from shared beliefs and values and more

___________________

¹ The results for physical health are ambiguous.

toward the “shared understanding” aspect of this definition. These shared understandings can take many forms including:

• Narratives: the telling, re-telling, and refining of key elements of a group’s common history.
• Symbolic boundaries: a way to distinguish members of the group from outsiders.
• Frames: perspectives or orientations, similar to a lens or filter through which people view and interpret their social world.
• Cultural toolkits: repertoires of behavior, such as coping strategies, that vary systematically among groups.

As an example of a narrative of survival, VanLandingham observed that the Vietnamese in New Orleans view themselves as triply forced migrants. In 1954, when Vietnam was split into the communist north and the democratic south, many Catholics who lived in the north of the country feared persecution, and about 1 million moved south. Thus, many Vietnamese in New Orleans were displaced after 1954, again after 1975, and again by Hurricane Katrina. Not surprisingly, VanLandingham said, the stories that they tell about themselves emphasize the idea that “we are survivors. There are setbacks. We survive. We lose everything. We rebuild and move on.” He added that this idea of being a member of a group of survivors is in their “cultural DNA.”

Another example of culture provided by VanLandingham involves frames, including a frame of insularity. Many of the Vietnamese who came to the United States had not looked to the South Vietnamese authorities when they had a problem. When public and private institutions were overwhelmed by Hurricane Katrina and unable to meet their basic responsibilities, said VanLandingham, the way the Vietnamese community thought about the world of authority, government, and institutions served the community well: “The cavalry is not coming. It is us or nobody.”

VanLandingham indicated that he is now hoping to generalize these concepts to other populations, which requires that they be operationalized and quantified. Though the Vietnamese community in New Orleans faced a specific set of circumstances, he believes that measurable indices of acculturation can be applied to other groups (see Figure 6-1). He also has been considering the utility of cultural consensus theory for this task, which seeks to identify core beliefs or core perspectives of a group and then determine the extent to which consensus in those beliefs exists in a community.

VanLandingham offered four main conclusions from his presentation. First, culture will influence the recovery trajectory for many forced migrants. Second, culture, which results from history and is dynamic, is often oversimplified. Third, culture is often confused with other elements of social

structure. Fourth, culture should be measurable if it is to be effectively leveraged as a useful construct in disaster research.

During the general discussion, VanLandingham elaborated on the unique resilience of the Vietnamese, noting that their narrative extends much further into the past than the Vietnam War (going back to resistance against the French and, before that, the Chinese) and speaks to their capacity to endure hardships against more powerful countries. However, he worried that such views can contribute to an “essentialism of culture,” both within and outside of the community. Such views could make the Vietnamese themselves less sympathetic toward members of other groups who have histories that are much more difficult than their own. Similarly, such essentialism can offend other groups that see themselves as working just as hard as the Vietnamese but not faring nearly as well. VanLandingham acknowledged that even as positive stereotypes can cause success to beget success (e.g., funders early on wanted to be involved in the Vietnamese community because “they wanted to bet on a winner”), so can difficulty beget difficulty. Jon Pedersen similarly noted that it is relatively easy to have a discussion about culture when culture is seen as something positive that helps resilience. However, such discussions are not as easy when culture has negative effects, as reflected in controversies over the discussion of such factors as essentialism.

Karabi Acharya noted that in the work that the Robert Wood Johnson Foundation has been doing on the issue of well-being (albeit not in the migration context), cultural identity and cultural narrative have come up very strongly in the context of social cohesion and sense of belonging as factors that contribute to well-being. She suggested that the role of cultural identity among Palestinians is very similar to the Vietnamese example providing the narrative of, “This is who we are. This is why we are in this situation.”

Michaela Hynie said policy recommendations that include cultural components are difficult without clear guidance around what it is about a particular set of beliefs or cultural tools that might be beneficial. She also observed that when people talk about integration, exceptionalism could be defined as a marker of poor integration; there is an interesting tension between community self-reliance and the documentation of successful integration.

VanLandingham observed that language may have protected the Vietnamese community in New Orleans from the perils that have affected other immigrant groups; the Vietnamese language is very different from English and is strongly emphasized in the community. He wondered whether such distinctiveness should be encouraged—for example, by supporting Vietnamese-language programs for second- and third-generation children. Katharine Donato cautioned that this might be difficult because culture is dynamic.

Fernando Riosmena, responding to a question about the “Hispanic health paradox,” indicated that it could be explained either by population selection or by ethnic community protective factors (e.g., relating to people being buffered from stress). He pointed to evidence for both selection as well as ethnic community protective factors, with the protection and “negative incorporation” effects stronger for Hispanic women (Riosmena et al., 2017). Riosmena urged more qualitative research to understand the protective factors that arise from such features as a community’s sense of cohesiveness and collective efficacy, and he emphasized that case studies are very important for understanding those factors.

Riosmena also suggested the incorporation of distrust of government into program implementation, which would require framing a program in the right way and taking advantage of community frames. Anna Marandi (National League of Cities) asked what resistance to engaging with government means in practice. For example, it might affect how a community prepares for a disaster or how likely it is to expect and depend on aid. The answer could affect how much help communities receive to prepare for a disaster in advance, she commented.