Abstract
Initial relations between the host society and migrants are likely to influence whether and to what degree migrants receive tangible and intangible settlement support that might affect their children's educational integration. As part of the 1980 Refugee Act, the United States officially began to provide settlement support to one group of migrants - refugees, thus institutionalizing more favorable host-society relations for refugees compared to non-refugee migrants. This article assesses the general idea that post-1980 US refugees will show higher levels of integration than non-refugees by testing the specific hypothesis that refugees’ foreign-born children will attain (by adulthood) higher levels of educational attainment than their non-refugee counterparts. As expected, we find that more schooling is completed among refugees’ children than among non-refugees’ children, all else being equal. We also observe that the level of governmental support at arrival is positively associated with educational attainment among refugees’ children. As expected, schooling differentials also drop in accordance with arrival-period declines in support due to drops in refugee children's schooling. The results highlight the pivotal roles that initial host-society/migrant relations play in fostering refugee integration and underscore the potential societal benefits from adopting and maintaining settlement policies for migrants.
Keywords
Introduction
Initial relations between migrants and host societies seem likely to matter for immigrant integration, with favorable host-society perceptions of migrants and generous support for their settlement assistance facilitating integration processes (Motomura 2021). Thus, countries, such as the United States, that have provided concrete settlement assistance to refugees, but not consistently to other migrants (Haines 2007), would seem more likely to evince greater and faster integration among refugees. However, research on refugees in the United States has yielded inconsistent findings concerning whether they experience integration advantages compared to non-refugee migrants. Some migration scholars view US refugee policy as successfully facilitating economic integration (Jiménez 2011; Waters and Pineau 2015; Hooper et al. 2016; Tran and Lara-García 2020). Others, however, have challenged this conclusion by suggesting that systemic problems in the reception many refugees have received in the United States curtail integration (US Senate Committee on Foreign Relations 2010; Brown and Scribner 2014; Fee 2019; Kyriakides et al. 2019). We suggest here that these seemingly contradictory findings may appreciably owe to refugees having arrived at different time periods and under different auspices that offer varying degrees of settlement support (Zolberg 2006; Motomura 2021). The present article seeks to shed light on this integration anomaly by focusing on refugees coming after passage of the 1980 Refugee Act, which standardized US refugee admission criteria and introduced legal and comprehensive provisions supporting refugee resettlement (Bruno 2016). As we show, however, the degree of this support diminished over time. Thus, we hypothesize that refugees will evince greater integration compared to migrants not receiving such support, all else equal, but that this integration advantage will shrink with reductions in levels of refugee settlement assistance over time.
The United Nations defines refugees as persons who are “outside their own country and who are unable or unwilling to return because of persecution or a well-founded fear of persecution on account of race, religion, nationality, membership in a particular social group, or political opinion” (Tichenor 2002, 147). Since the 1980 Refugee Act, the US government has used this de jure definition to adjudicate applications for refugee status, although it has been critiqued for inconsistencies in practice. 1 In this article, we focus on refugees in the official legal sense of the term (i.e., those who have been granted refugee status by the US government) because since 1980, such persons have, by law, received tangible governmental support of various kinds to facilitate their settlement. 2 Our specific research aims are to gauge empirically (a) whether the sheer availability of such support appears to boost the education of refugees’ children, compared to children of migrants who do not receive such assistance (from this point on, we refer to such migrants as “non-refugees”) and (b) the degree to which variation in the availability of settlement assistance over time affects the educational attainment of refugees’ children. 3
Empirical investigations of such research questions are challenging for several data and measurement reasons. First, comparing the effects of settlement assistance on integration requires examining refugees who arrived in periods with varying levels of assistance and under various contexts of departure (e.g., differing national origins and socioeconomic backgrounds). Second, comparing refugee and non-refugee migrants on important integration measures (like educational attainment) is challenging because the schooling of adult entrants has already occurred in origin countries. We, thus, focus here on migrants’ children's educational attainment. Third, research requires samples large enough to provide adequate bases for detecting differences in children's education resulting from variation in levels of settlement assistance. By using large datasets compiled from the US Census and the American Community Survey across multiple years, we can adequately detect possible differences in educational integration by variation in degree of settlement assistance, thus shedding new light on whether and to what degree the provision of such assistance to refugees boosts their children's educational attainment compared to that of their non-refugee counterparts. If our findings reveal higher educational attainment for refugees’ children, all else being equal, and if the degree of settlement assistance over time is positively associated with the size of such advantage over time, the results would imply that adopting adequate settlement-assistance policies for all migrants, not just refugees, would benefit immigrant integration in general.
The rest of this article consists of six sections. First, we introduce theoretical perspectives on immigrant integration that emphasize how the nature of initial host-society/migrant relationships affects integration. Second, we apply these perspectives to the case of refugees coming to the United States since 1980, including the development of specific research hypotheses about expected relationships between settlement assistance and refugee children's integration. Third, we describe in detail the data sources we use and the measurement of key independent and control variables. Fourth, we present and discuss the findings and then the results of several robustness checks carried out to assess the adequacy of our findings. Finally, we discuss the implications of our analysis for both US refugee and immigration policies.
Relational Theoretical Perspectives on Immigrant Integration
The ways host societies and migrants define their initial relationships and responsibilities toward one another carry important implications for integration (Motomura 2021). Such dynamics are often suggested in definitions of integration and provide useful bases for understanding possible differences in the educational attainment of refugees’ and non-refugees’ children (Brown and Bean 2006; Waters and Pineau 2015; Jiménez 2017). Emphasis on host-society/migrant relations has emerged from two separate, but complementary, approaches that envision immigrant integration (whether social, political, economic, or spatial) as intertwined with institutional and other aspects of the host society that mirror the society's perceptions, expectations, and receptions of migrants (Reitz 2003). The first approach, most emphasized by Jiménez (2017), highlights the importance of viewing assimilation (or integration) as a “two-way street” that involves migrants coming to resemble the members of host societies and host societies taking on migrants’ features and tendencies over time. A second approach, articulated by Motomura (2006; 2021), emphasizes that host-society initial expectations about eventual immigrant citizenship matter greatly for instigating settlement assistance to facilitate newcomer integration.
Initial governmental expectations that migrants deserve and ultimately will become full-fledged members of society are likely to facilitate the provisions of governmental economic and social support for immigrants (Motomura 2006). Thus, during much of the nineteenth century, migrants in the United States were generally expected, in due course, to qualify for and obtain citizenship, an orientation that undergirded the view that newcomers deserved tangible and intangible settlement support. 4 This view, coupled with the recognition that the nation needed migrants to fuel its growth, led to the enactment of the 1862 Homestead Law, which allowed migrants who filed a declaration of intent (to become a citizen) to take advantage of land grants from the US government (Zolberg 2006). This reception, along with the governmental assistance that characterized US immigration policy until the mid-19th century, stands in stark contrast to more laissez faire (i.e., neutral) views of migrants prevalent in the late-nineteenth and early-twentieth century (US Commission on Immigration Reform 1997). The earlier, more favorable perceptions of newcomers presaged greater acceptance of new entrants and the provision of more positive, inclusive reception contexts that included settlement assistance, which was generally unavailable to early-twentieth-century newcomers, who were often perceived in more hostile, nativist-oriented terms (Higham 1955), thus creating challenges to their eventual integration.
Recent cross-national studies have shown that positive relations between the host society and migrants, at the country level, matter for the adoption of policies to attract and support specific groups of migrants, actions that also induce higher levels of political engagement among newcomers (Bloemraad 2006; Poros 2010). At the individual level, researchers have also found that favorable relational contexts foster more migrant/native friendships, more contact between natives and newcomers, and, subsequently, greater warmth of welcome experienced by migrants (Fussell 2014; Kotzur, Tropp and Wagner 2018; Tropp et al. 2018; Leszczensky and Pink 2019). Regarding US reception of refugees, Portes and Rumbaut (2001, 2014) note that Southeast Asian refugees in the United States experienced more favorable contexts of reception compared to other migrants, resulting in levels of economic integration that cannot be explained by their human capital endowments alone. In an inverse vein, the authors’ segmented assimilation theory suggests that unwelcoming receptions in the form of hostile governmental actions and legal strictures facing migrants, especially unauthorized migrants, severely hamper immigrant integration, with such processes constituting the flip side of positive integration dynamics.
Host-Society/Refugee Relational Dynamics in the United States
This article focuses on comparing legal permanent resident entrants and refugees. The latter sometimes have faced more positive reception factors (i.e., the initial provision of assistance at reception stemming from the host society's willingness to assist newcomers’ resettlement), thus fostering more favorable contexts for integration (Motomura 2021). In 1980, for example, the United States, for the first time, passed permanent refugee legislation (Martin 2011). Called the Refugee Act of 1980, the legislation served two major purposes: 1) to make general and systematic the procedures for admitting refugees to the United States and 2) to provide uniform federal assistance to resettled refugees to facilitate their economic self-sufficiency (GAO et al. 2011; Bruno 2011; 2016). 5 In essence, it is the only US immigration-governing policy that includes specific provisions for immigrant integration.
The 1980 Act established the Office of Refugee Resettlement (ORR) to oversee the distribution of cash, medical, and supplemental income assistance (for refugees with special needs and disabilities), as well funds for a host of support services, including English training, employment services, and citizenship preparation education (Bruno 2016). Beyond such direct benefits and services made available to newly arrived refugees, the US federal government funded voluntary agencies and school districts serving a significant proportion of the refugee population (US ORR 1980). Finally, according to the US Immigration and Naturalization Service (INS), 6 refugees were granted an expedited pathway to citizenship equivalent to legal permanent resident (LPR) status at entry (INS 1988; Martin 2011). Unlike other LPRs, who had to wait at least five years to apply for US citizenship, refugees who adjusted to LPR status (usually one year after arrival) could apply their time as refugees to fulfill the five-year requirement for immigrant naturalization (INS 1980). This benefit is considerable, given that most migrants receiving LPR status (52 percent) do so after having lived in the United States for quite some time under a different legal status (Waters and Pineau 2015). In general, the more positive relational treatments for refugees in the form of greater resource accessibility created an environment in which they and their children were more likely to attain higher levels of integration than non-refugee migrants, all else being equal.
Although the 1980 Act was intended to be comprehensive and long lasting, the levels of actual settlement assistance it provided soon started to erode in both the dollar amounts allocated by Congress for the purpose of refugee support and the length of time newly arrived refugees were eligible to receive cash and medical assistance (GAO 1990; Bruno 2011). Based on data from the ORR's annual reports to Congress, Figure 1 shows the declines in funding and length of eligibility available to refugees from 1980 to 1995, the time period of migration examined in this article (ORR 1980–1995). A report by the US General Accountability Office noted that between 1985 and 1989, the average absolute dollar amount that each refugee received declined 48 percent, from $6,921 to $3,600 (GAO 1990). Also, a greater portion of the cost of support was shifted from the federal government to individual states, which, unsurprisingly, resulted in cutbacks in important support services in many states (GAO 1990). Although the federal government justified these changes as a way to incentivize newcomers to more quickly become self-reliant, studies and reports from researchers and governmental committees alike pointed to insufficient funding as a challenge for refugee integration in the United States (US Senate Committee on Foreign Relations 2010; Bruno 2011; Brown and Scribner 2014; Fee 2019).
Level of governmental assistance for refugees from 1980 to 1995. Source: Authors’ calculations from the US Office of Refugee Resettlement's Annual Report to Congress (1980–1995).
Such changes suggest that refugees’ children's educational attainment would reach its peak for those who arrived during the early 1980s, when governmental assistance for refugees was at its highest levels, and that any margin in educational attainment for refugees’ children, compared to non-refugees’ children, would diminish across the late 1980s and into the 1990s. Thus, drops in schooling gaps between refugees’ children and non-refugees’ children would result from changes in the former's education, not from shifts in non-refugees’ children's schooling. Additionally, financial support for both English language education and citizenship training was prominent in the tangible assistance provided to refugees in the 1980 Act. Such tangible support reflected the host society's investment in refugees because becoming naturalized and attaining English proficiency increase structural access to economic opportunities and facilitate social participation (Portes and Rumbaut 2014; Tran and Lara-García 2020). Naturalization and greater English proficiency, thus, would be expected to account for (i.e., mediate) a large portion of refugees’ children's education advantage over non-refugees’ children, who did not receive such benefits.
Research Hypotheses
Based on the above conceptual and theoretical considerations, we formulate the following hypotheses about how educational attainment varies across categories of migration status and by three arrival periods. Hypothesis 1: Because refugees receive more tangible governmental support than non-refugee migrants, we expect refugees’ children to show higher educational attainment than their non-refugee counterparts, all else being equal. Hypothesis 2: Because the levels of support provided refugees in the 1980 Refugee Act began to diminish after its passage, we expect any educational attainment advantage for refugees’ children to diminish across arrival periods (i.e., from 1980–1984 to later periods). Hypothesis 3: We expect any diminishing gap in educational attainment between the two groups to result from declines in the educational attainment of refugees’ children across the arrival periods, not from changes in the education of non-refugees’ children. Hypothesis 4: We expect naturalization and English proficiency to mediate a notable part of any positive effects of refugee status on educational attainment.
Data, Measures, and Methods
Data Sources
The lack of information on either refugee status or parents’ socioeconomic characteristics, or both, in large surveys, along with our need to obtain usable information on the completed education of adult migrants who arrived in the United States as children, requires that we pull together and work with multiple data sources. To obtain all the information and kinds of migrants needed, we employ three sources of official governmental data. The first provides much of the data for our main analytic sample and consists of pooled information on respondents from the American Community Surveys (ACS), 2011–19, conducted by the US Bureau of the Census (Ruggles et al. 2020). Each ACS provides individual-level information on educational attainment and other variables for adults who arrived in the United States between 1980 and 1995, inclusive of or before the age of 18, and who had reached at least age 24 by the time of the particular ACS survey in which they participated.
The second data source involves information from the 1990 and 2000 Integrated Public Use Microdata samples (IPUMs) from the US Bureau of the Census (Ruggles et al. 2020). These data enable us to develop aggregate information on the characteristics of arrival cohorts of migrants coming during the 1980s and 1990s. These cohorts contain the parents of the migrants from the ACS data analytic sample. We link the parental cohorts to the ACS respondents using corresponding arrival years and national origins. The third data source, compiled by the agency formerly called the Immigration and Naturalization Service (INS), contains administrative information on the numbers of refugees and other persons becoming LPRs in the United States from specific countries from 1980 through 1995 (see INS 1980–1995).
Our focus on educational attainment as a key indicator of immigrant integration derives from its vital importance for socioeconomic mobility (Breen and Muller 2020). Here, we examine the education of migrants who arrived in the United States between ages 0 to 18 and who were ages 24 or older by the time period under study. We do not analyze the schooling of migrants who arrived as adults because their education would largely have taken place outside the United States and, thus, would not have been a function of their destination-country integration experiences (Hirschman, Kasinitz and DeWind 1999). We also focus on those who came during the period 1980 to 1995, allowing us to capture as much of the force of official refugee status on migrants’ children's educational attainment as possible. Additionally, many entrants coming to the country after 1995 were still too young to have completed their schooling.
Two pathways exist to apply for refugee status. One involves applying from abroad and the other from within US territories (INS 1980–1995). Successful applicants in the former case are referred to as refugees and, in the latter, as asylees (INS 1980–1995). Both kinds qualify for financial and medical assistance, as well as support services, from the US government and an expedited pathway toward citizenship, although the timeline from when they are eligible for assistance varies due to differences in when and where they applied for status (INS 1980–1995; ORR 1980–1995). In this analysis, we often rely on governmental statistics from INS, which sometimes places refugees and asylees in the same category. During the time period studied, however, asylees comprised only 5 percent of the total applications approved (INS 1995). Given this small number and given that our main research goal is to shed light on how official refugee status affects immigrant integration, we do not distinguish between those gaining refugee status abroad (i.e., refugees) and in the United States (i.e., asylees) and refer to them collectively as refugees.
Measures of Dependent and Independent Variables
Educational Attainment
We measure educational attainment by the number of years of schooling completed by respondents. The ACS reports respondents’ highest education level completed. There are a total of 24 categories, ranging from “no schooling completed” to “doctoral degree,” which we recode into a numerical continuous variable by assigning each education level the standard number of years of schooling required to complete it. 7
Initial Host-Society/Migrant Relational Factors
We examine how and to what degree two host-society/migrant relational factors influence migrant children's educational attainment. The first is migration status, consisting here of refugee and non-refugee entrants. We identify refugee status, using a proxy measure, as described in detail in the Online Appendix. Essentially, we use individuals’ self-reported national origins and migration years to calculate their likelihood of being admitted to the United States as refugees. Tangible governmental assistance for refugee resettlement and integration has been available for refugees since 1980, but not for non-refugees. To be sure, some members of certain refugee groups (e.g., Hungarians, Cubans, Southeast Asians) were admitted and sponsored by the federal government prior to 1980, but only under temporary refugee programs enacted on an ad-hoc basis (e.g., the Displaced Person Act and the Cuban and Indochinese Refugee Adjustment Acts) and characterized by inconsistency in administration and federal support (INS 1988). Thus, we examine comparisons of assistance for refugees and non-refugees starting in 1980.
The second factor that reflects variation in relational support is arrival period. As detailed above, the amount of federal aid and length of support for refugees were highest during the early 1980s, when the Refugee Act was first passed. Later in the 1980s and into the 1990s, support levels notably declined. When the Cold War came to an end in 1990, the US geopolitical orientation also shifted, which had the effect of further reducing governmental incentives to maintain assistance levels (Martin 2011). We capture these changes in the degree of support and the nature of reception by grouping migrant arrival times into three periods: early (1980–1984), middle (1985–1989), and late (1990–1995) arrival periods.
Control Variables
Sociodemographic Background
Other variables are controls that adjust for sociodemographic differences that may affect educational outcomes, including age at entry, arrival year, and gender. We calculate age at entry by subtracting the respondent's self-reported year of immigration from birth year. This variable captures early developmental exposure to the United States and would be expected to relate to integration negatively, with older arrivals experiencing less integration than younger arrivals (Myers, Gao and Emeka 2009; Lee and Edmonston 2011; Beck, Corak and Tienda 2012). The year a respondent reported arriving in the United States is measured from 1980 (the earliest arrival year included in this analysis) and used to account for a temporal trend of increasing tendencies to pursue higher education in the United States that may affect migrants’ children's educational attainment (Snyder 1993). We also control for gender because it is a known correlate of educational outcomes, especially among the children of migrants (Feliciano and Rumbaut 2005; Bean et al. 2011).
Ethnoracial Status
Because ethnoracial status is related to educational attainment, we also adjust for it in the regression models. We use four ethnoracial groups: Hispanics, Asians, Blacks, and Whites. We sort foreign-born respondents in the ACS into one of these mutually exclusive groups based on their answers to two questions, one asking for their race and the other for whether they were of Hispanic origin. A small number of cases (268) were dropped from the analysis because respondents did not specify a race beyond “other” and reported being non-Hispanic.
Socioeconomic Background
Because parental socioeconomic status (SES) is known to be a strong predictor of educational attainment, it is important to control for it in the analysis (Portes and MacLeod 1999). Due to the lack of individual-level indicators of parental SES in ACS data, we employ a measure of the average SES for the annual cohorts of arrivals with the same national origin and immigration year as the migrants’ children (see Online Appendix for more details). We use three SES indicators: maternal year of schooling, family size, and family income as a percentage of poverty. Together, the country/year cohort averages of these indciators are used as controls in regressions of educational attainment on host/migrant relational factors.
Resources After Migration
Two factors that may potentially shape migrants’ children's educational outcomes and may themselves be influenced by host-country/migrant relational factors are English proficiency and naturalization (Portes and Rumbaut 2014; Tran and Lara-García 2020). We code respondents as English proficient if they reported speaking English as their only language or speaking English well or very well. We also note whether the respondent had obtained US citizenship. Including these variables in the analyses makes it possible to assess one of our hypotheses: whether and to what extent governmental support serving to facilitate refugees’ pathway to English proficiency and citizenship (e.g., English and citizenship trainings) mediates any educational differential observed between the children of refugees and those of non-refugees.
Analytic Model and Plan of Analysis
To model the direction and degree of relationships between host-society/migrant relational factors and migrants’ children's educational attainment, we use ordinary least squares regression, controlling for other factors and including mediating factors. To additionally account for the influences of differences in national origin, we employ origin-country fixed effects models, which adjust for unmeasured differences among respondents associated with their having come from particular origin countries. We anticipate several patterns to emerge from the data. First, we expect refugee status to have a positive effect on educational outcome, even after accounting for sociodemographic and socioeconomic background factors, as well as origin-country differences. Second, we expect educational attainment to decrease across the three arrival periods among refugees, thus diminishing the educational differential between the children of refugees and non-refugees. Finally, we expect naturalization and English proficiency to mediate a notable portion of any positive effect of refugee status on educational attainment.
Findings
Descriptive Patterns
According to US government statistics, the highest refugee flow after the 1980 Refugee Act was enacted occurred in that year, when over 200,000 refugees were admitted to the United States (see Figure 2; INS 1980–1995). Subsequently, declines in refugee admission set in until 1983, when the admission rate leveled off at 60,000 to 70,000 per year. From 1989 until 1995, admissions remained at or above 100,000 per year, an increase coinciding with the end of the Cold War and the Soviet Union's collapse. The stacked bar graph in Figure 3 shows the national origins of the top five refugee groups that gained LPR status in the United States for the three arrival periods distinguished here. 8 In 1980 to 1984, the vast majority of refugees who gained LPR status (about 70 percent) arrived from Southeast Asian countries (i.e., Cambodia, Laos, and Vietnam). Across the latter two periods, fewer Southeast Asians arrived, and the numbers of refugees from Iran, Cuba, and the “other” category grew, reflecting increasing diversity in refugees’ national origins. Also, from 1990 to 1995, a sudden influx of refugees from countries that had been a part of the Soviet Union arose, suggesting a ripple effect at the end of the Cold War. Overall, the changing rates of admission and shifting national origins among refugees shown in Figures 2 and 3 indicate the importance of controlling for differences in arrival period and national origin in our statistical analyses.
US Refugee Admissions from 1980 to 1995. Source: Authors’ calculations from US Immigration and Naturalization Service's Yearbooks of Immigration Statistics (1980–1995). Note: These numbers include asylee adjustments.
Refugee adjustment to LPR Status by arrival period and national origin. Source: Authors’ calculations from US Immigration and Naturalization Service's Yearbooks of Immigration Statistics (1980–1995).
We obtain information on refugees’ and non-refugees’ adult children, using pooled data from the 2011 to 2019 ACS. The sample includes a total of 73,642 observations, 34,216 of whom we identify as arriving as part of a majority-refugee migration flow based on arrival-year/origin-country data, as explained above. We classify these individuals as refugees’ adult children. Another 34,631 observations are classified as non-refugees’ children because the migration flows based on their origin country and arrival year contained fewer than 5 percent refugees. 9 Table 1 displays descriptive statistics by migration status (i.e., for refugees and non-refugees) and arrival period. Two patterns stand out. First, refugees’ and non-refugees’ children exhibit different levels of educational attainment, as well as differences in ethnoracial, sociodemographic, and socioeconomic status. Second, while SES composition for non-refugees’ children varies little across the three arrival periods (maternal years of schooling = 10.0–10.1, family size = 4.8–4.9, and family income = 181.8–215.6), the SES composition for refugees’ children varies considerably (maternal years of schooling = 8.8–12.9, family size = 4.2–5.2, and family income = 155.2–284.2), reflecting the shifting national origins of refugees across arrival periods.
Means and Standard Deviations for Years of Schooling and Other Research Variables by Migration Status and Arrival Period, Adult Foreign-Born Children of US Migrants Arriving From 1980–1995.
Source: Authors’ calculations from pooled 2011–2019 ACS data (see Ruggles et al. 2020).
At ages 24 or older, refugees’ children show higher educational attainment than non-refugees’ children (1.5 years, 1.4 years, and 2.4 more years of education during the early, middle, and late arrival periods, respectively). Hispanics constitute an ethnoracial majority in all periods among non-refugees’ children (with their proportions in the samples ranging from 0.63 to 0.67), while among refugees’ children, the large proportion of Asians in the early-arrival period (0.88) decreases as the proportions of Whites and Hispanics rise in the latter two periods. In the late-arrival period, Asians comprise only about a quarter of the refugee group. Corresponding to reported US government statistics on refugees’ national origins, we find that the vast majority of Asian refugees’ children in the ACS sample came from four Southeast Asian countries: Cambodia, Laos, Thailand, and Vietnam. Hispanic refugees’ children are mostly Cuban, although there is a small but notable proportion of Nicaraguans. Most White refugees’ children arrived from countries in Eastern Europe, with the exceptions being Afghanistan and Iraq. The proportion of Blacks is lowest across all arrival periods and migration statuses, and most Black refugees’ children came from Haiti and Ethiopia. Regarding socioeconomic background, very little variation emerges among the non-refugee group across arrival periods, but the refugee group evinces a bifurcated pattern. Refugees’ children who arrived during the later-arrival period exhibit the highest levels of parental human capital (with an average maternal year of schooling of 12.9, a family size of 4.2, and a family income 284.2 percent above the poverty threshold), while refugees’ children who arrived during the early period were generally the most disadvantaged, except on family income (showing a maternal schooling of 8.8 years and a family size of 5.2).
Overall, these results suggest that variation in migration status and arrival period are likely to influence migrants’ integration patterns. The differences across and within migration-status groups in ethnoracial background and parental human capital profiles indicate the need to control for background factors in analyses of differences in children educational attainment. Also, refugees’ children, regardless of arrival period, show decisively higher rates of naturalization and somewhat higher levels of English proficiency than non-refugees’ children. This pattern, too, is consistent with the idea that refugees’ children will show higher educational integration than non-refugees’ children. We later assess the extent to which these factors mediate such education differentials.
Multivariate Analyses of Refugee Status and Control Variables
Table 2 shows results for multivariate analyses from ordinary least squares (OLS) regression. We first examine the zero-order relationship between refugee status and educational attainment shown in the first column (b = 1.85, p < .01). Without controlling for any other factors, refugees’ children attain 1.85 years more schooling than non-refugees’ children. This difference is reduced to 0.62 (p < .01) in the model containing controls (column 2), where we adjust for ethnoracial, sociodemographic, and socioeconomic differences between the two groups. Table 2 also includes results from two fixed effects (FE) models that each control for heterogeneities across arrival year and origin country. While the effect of refugee status remains almost the same in the arrival-year FE model (column 3: b = 0.67, p < .01) compared to the preceding model, it decreases by a fourth in the origin-country FE model (column 4: b = 0.43, p < .05), suggesting that country-level heterogeneity, but not arrival-year heterogeneity, accounts for some of the remaining educational advantage among refugees’ children. Overall, however, results from the origin-country FE model show that even after controlling for measurable background differences across origin countries between refugees’ and non-refugees’ children, a 0.43 year-of-schooling advantage associated with children's refugee status remains, supporting Hypothesis 1 that refugees’ children show a positive net increment in educational attainment.
Regressions of Years of Schooling on Migration Status and Control Variables, Adult Foreign-Born Children of US Migrants, Separately for OLS, Country-of-Origin (CFE) and Arrival-Year Fixed Effects (YFE) Models.
Source: Authors’ calculations from pooled 2011–2019 ACS data (see Ruggles et al. 2020).
*** p < 0.01, ** p < 0.05, * p < 0.1.
The results also reveal the expected relationships between the control variables and educational attainment. The second column in Table 2 shows that ethnoracial status is a particularly strong predictor of education, with Asians (b = 0.95, p < .01) and Blacks (b = 0.53, p < .01) demonstrating significantly more and Hispanics demonstrating significantly fewer (b = −1.23, p <.001) years of schooling than Whites, all else being equal. However, these statistically significant differences in ethnoracial composition on educational attainment are reduced substantially when we control for origin-country differences. The positive effect for Asians shrinks by almost half (b = 0.49, p < .01), whereas it reverses direction (but is not statistically different) for Blacks. The negative effect for Hispanics is also reduced by more than half to −0.47 (p < .01). Controlling for origin-country differences between ethnoracial groups, thus, makes a considerable difference in schooling levels. Stated in substantive terms, about half of Asian migrants’ and all of Black migrants’ schooling advantage over White migrants result from their coming from countries with more highly educated populations (or from countries in which more highly educated individuals were more likely to migrate). For Hispanics, almost three-fifths of their disadvantage is associated with their coming from countries with less-educated populations.
Differences in sociodemographic background also play an important role in educational attainment, but these effects are generally independent of country-level differences (with the exception of arrival year). As expected, age at entry has a negative effect on education (b = −0.15, p < .01). For each year older that migrants (i.e., refugees and non-refugees) arrive in the United States, the education penalty is about one-sixth of a year's schooling. We find a small but positive effect for arrival year (b = 0.02, p < .01) after taking origin-country fixed effects into account, suggesting that, among those who immigrated from the same country, those who arrived later experience slightly higher educational attainment, all else being equal. The effect of arrival year may reflect the overall rise in the education level among the US population over time (Snyder 1993). We also find that male migrants show lower levels of educational attainment than females (b = −0.37, p < .01), which is consistent with the results in the literature on gendered differences in schooling (Feliciano and Rumbaut 2005; Bean et al. 2011).
Finally, we turn to the arguably most important set of control factors: parental SES and family factors, which are here measured at the parental arrival-year/origin-country cohort level because, at the time the ACS surveys data were collected, respondents were adults who generally no longer lived with their parents, making it impossible to obtain individual-level parental SES and family measures. The findings shown in Table 2 (column 2) reveal that, at a cohort level, individuals’ educational attainment is positively related to maternal education (b = 0.32, p < .01) and negatively related to family size (b = −.70, p < .01) and family income (b = −0.34, p < .01). Including origin-country fixed effects (column 4), the effect sizes of the parental SES factors are all reduced (to b = 0.23, p < .01 for maternal education, b = −0.14, p < .1 for family size, and b = −0.21, p < .05 for log family income) but remain significant.
Multivariate Analyses of the Refugee Status Effect Across Arrival Periods
The next analyses rely on origin-country FE models to assess the effect of refugee status on educational attainment across the three arrival periods (Table 3). Model 1 shows that arrival period itself has a positive effect on educational outcome. In general, the children of migrants who arrived during either the early period (b = 0.34, p < .01) or middle period (b = 0.31, p < .01) attained more schooling than those who arrived during the late period. 10 To test the second hypothesis that the education advantage associated with refugee status diminishes across the three arrival periods due to decreasing governmental support for refugees, we include interaction terms for refugee status and arrival period in Model 2. The interaction terms are significant. The educational advantage associated with refugee status decreases from 0.42 (b = −0.05 + 0.47) for the early period to 0.15 (b = −0.05 + 0.20) for the middle period, and by the late period, it has completely disappeared (with the main effect of refugee status not significant). Figure 4a graphs the adjusted educational attainment of refugees’ and non-refugees’ children across the three arrival periods, based on the Model 2 equation. The gap between the adjusted educational attainments of refugees’ and non-refugees’ children diminishes over the three arrival periods, illustrating diminishing refugee educational attainment with declining refugee settlement assistance, thus supporting Hypothesis 2.
Regressions of Years of Schooling on Migration Status (MS), Arrival Period (AP), and the Interaction of MS and AP, Adult Foreign-Born Children of US Migrants, Country-of-Origin Fixed Effects Models.
Source: Authors’ calculations from pooled 2011–2019 ACS data (see Ruggles et al. 2020).
*** p < 0.01, ** p < 0.05, * p < 0.1.
We base Hypothesis 3 on the same reasoning. Due to decreasing governmental support for refugees across arrival periods, we expect a concomitant decline in educational attainment for refugees’ children. Focusing on arrival periods, we find evidence of this decline in Model 2. Refugees’ children who arrived during the early period, when the support level was high, do significantly better than those who arrived during the late period, when the support level was lower (diff. = 0.70 [0.23 + 0.47]), and refugees’ children who came during the middle period also do significantly better than those from the late period, but to a lesser extent (diff. = 0.49 [0.29 + 0.20]). Figure 4a illustrates this pattern clearly, showing steady declines in adjusted educational attainment among refugees’ children across the three periods. Arrival period also has an effect on the educational attainment of non-refugees’ children. However, rather than a steady decline across the three periods, the drop-off in education occurs only during the late period.
Finally, to assess Hypothesis 4, we estimate a model that introduces binary variables for naturalization and English proficiency to test whether and to what extent acquiring these resources mediates the effects of refugee status for those who arrived during the late and middle period. Model 3 shows that the educational differentials between refugees’ and non-refugees’ children are 0.47 for the early period (diff. = −0.18 + 0.65) and 0.18 for the middle period (diff. = −0.18 + 0.36). The results are nearly identical to the educational differentials calculated in Model 2 of Table 3. Overall, results from Model 3 suggest that although naturalization (b = 1.45, p < .01) and English proficiency (b = 2.95, p < .01) each exert strong and positive effects on education for migrants’ children, they do not appear to mediate the interaction effects between refugee status and arrival period.
Interestingly, however, including naturalization and English proficiency serves to mediate some of the arrival-period effects for non-refugees. Comparing the non-refugee changes in educational attainment owing to arrival period in Models 2 and 3, we note that the education gain for non-refugees’ children who arrived during the early period, compared to the late period, becomes non-significant in Model 3. The gains for those arriving in the middle period compared to the late period also drop appreciably (b = 0.15, p < .01 in Model 3 vs. b = 0.29, p < .01 in Model 2), suggesting that for non-refugees, different rates of naturalization and English proficiency notably explain some of the educational differentials by arrival period. Late-arrival non-refugees’ children experience an educational disadvantage compared to their early- and middle-arrival non-refugee peers, due in part to lower English proficiency and naturalization rates. If we examine Figure 4b, which graphs adjusted educational attainment by migration status and arrival period, based on the equation of Model 3, the mediation effect on arrival year among non-refugees’ children is more apparent. We no longer see as much of a difference in educational attainment among non-refugees’ children who arrived during the late period compared to those who arrived during the two earlier periods.
Robustness Assessments
The ACS and IPUMS datasets used in this analysis are not without limitations. For example, they are cross-sectional, not longitudinal, thus necessitating that we construct later panels of children who could be linked to parents only by cohort. Also, the datasets, while indicating whether respondents were foreign born, do not provide information about initial migration status, hence the need to use a third data source from INS to impute refugee status. As large-scale national surveys, the ACS and US census offer major advantages, however. Chief among them is that they provide sufficiently large, randomly selected samples of migrants that 1) enable our utilization of a comparative research design, 2) estimate relationships among variables that achieve considerable stability, and 3) support generalizability of findings. Moreover, to our knowledge, no other data alternatives contain enough cases to assess the hypotheses set forth here. But given the data limitation, we undertook a number of robustness checks. First, we checked the adequacy of our method for refugee identification by comparing the ACS population estimates to governmental statistics on the foreign-born population arriving from countries with high refugee concentrations (see below for more details). We also used origin-country fixed effects models to account for educational differences due to national origins. Additionally, we replicated our analyses, using a smaller, more geographically localized dataset on migrants’ children. Lastly, we assessed sampling errors. None of these assessments affected the results. The procedures used in the assessments can be found below and/or are elaborated in the Online Appendix.
Origin-Country Fixed Effects
The results of the origin-country FE analyses show that, even after controlling for country-level differences, refugee status is associated with higher educational attainment for the foreign-born children of migrants. Some recent analyses note that controlling for origin country eliminates any positive effects of refugee status on children's educational attainment (Fitzgerald and Arar 2018). We find that such is not the case, although our target population here is more limited (i.e., those coming under the auspices of the 1980 Refugee Act and during the arrival period that extends from 1980 to 1995). As expected, refugee status differences do not disappear in our origin-country FE models, suggesting that a part of the refugee status effect cannot be accounted for by differences in background characteristics between refugees and non-refugees alone.
The Adequacy of Refugee Identification
To gauge the extent to which the 2011–2019 ACS samples used to assess educational attainment are satisfactorily representative of the numbers of migrants arriving from 1980 to 1995, especially those coming from refugee-sending countries, we compared the country-specific, year-specific numbers of foreign-born persons in the ACS samples (adjusted to population totals, using person weights) to the numbers in the admission statistics published by the INS (see Online Appendix Table A for an extensive list of admission statistics). We would expect some attrition in the numbers of arrivals in the ACS, due to mortality, return migration, and errors in reporting, but overall, the ACS numbers should show only slight deviations from INS statistics if attrition is not large, which is indeed the case (see Online Appendix Table B). 11 The ACS numbers are, on average, 10 percentage points below those indicated by INS statistics, as one would expect, given a not-unreasonable rate of attrition. The ACS country-specific entry numbers are generally consistent in showing reasonably lower figures compared to the INS county-specific admission statistics.
Discussion and Conclusion
The view that host-society expectations and treatment affect how well migrants thrive in their new societies is a common perspective that emerges from relational understandings of immigrant integration (Portes and Rumbaut 2001; Motomura 2006; 2021). Migration scholars have generally characterized US host-society/migrant relations in the modern era as “laissez faire” or even hostile in nature (Higham 1955; Zolberg 1988; U.S. Commission on Immigration Reform 1997). One notable exception, however, is the US governmental provisions for refugee resettlement, established by the 1980 Refugee Act, which some scholars have pointed to as a rare policy that facilitates integration (Bloemraad 2006; Jiménez 2011; Waters and Pineau 2015; Hooper et al. 2016; Tran and Lara-García 2020). Nonetheless, the extent to which integration has occurred for new refugee arrivals has been questioned (Brown and Scribner 2014; Fee 2019). Results from our analysis suggest that many of the problems for refugee integration are rooted in the governmental disinvestment that began a few years after the inception of the 1980 Refugee Act itself. Although intended to be comprehensive and enduring (Bruno 2016), support for refugee resettlement (in dollars, as well as length of eligibility) has repeatedly declined since the early 1980s, creating increasing challenges for refugee integration.
This article has pulled together data from multiple sources and time periods to develop findings that are suggestive for future immigrant integration policies. Focusing on the children of refugees and other migrants who arrived to the United Stated some 25–40 years ago enables us to not only begin our analyses at the time when the historic 1980 Refugee Act was passed but also capture the integration experiences of now-adult children of migrants who accompanied them in their migration several decades ago. This long-term view is critical for being able to compare refugees’ children's educational attainment to that of their counterparts coming under other migration regimes.
The existing literature on US refugee admissions demonstrates that national origins and race, along with other geopolitical considerations, have affected whose applications for refugee status are approved by the US government, especially during the Cold-War era, and how newcomers are treated upon arrival (Zolberg 1988; Lennox 1993; Menjívar 2006; Esbenshade 2011). Although this article's goal is to assess how refugee status and the degree of governmental assistance affect integration, rather than to investigate ethnoracial and other selection biases in refugee admissions, we do control for ethnoracial and other background factors in our analyses. Results show that controlling for race and national origins, along with other factors, somewhat, but not totally, accounts for the educational advantage attained by refugees’ children over that of non-refugees’ children. Overall, we find that refugees’ children fare better educationally than non-refugees’ children, particularly if they arrived shortly after the 1980 Act was passed and implemented and when its benefits were at the highest level. Our findings, thus, suggest that the integration of refugees’ children can be enhanced by governmental provision of settlement assistance, especially when such assistance reaches the levels provided in the early 1980s.
In summary, the results obtained in this article suggest not only that tangible governmental financial and social support generate economic stability for migrants in the short run, as previous research on refugees has shown (see Capps et al. 2015; Evans and Fitzgerald 2017; Fix, Hooper and Zong 2017), but also that better host-society/migrant relations make for longer-term positive effects on the integration of the next generation. The decreasing assistance for refugees during the past 25 years helps explain why studies examining more recent refugee groups find more integration challenges facing refugees (US Senate Committee on Foreign Relations 2010; Fee 2019). The Welfare Reform Act of 1996, for example, further squeezed refugees’ sources of support by cutting funding for public welfare programs, for which refugees were eligible (on top of refugee assistance), and by restricting length of eligibility for noncitizens, including refugees (Bruno 2011). Capturing how the dynamics of general social welfare policy affects the integration of newcomers of different migration statuses may be a worthwhile endeavor for future research.
More broadly, this article's findings suggest that the United States and possibly other nations could benefit from modifying their longstanding laissez-faire approaches toward newcomers. Although the reason often given for the US government's disinvestment in refugee assistance is to encourage refugees to become economically self-reliant (Martin 2011; Brown and Scribner 2014; Bruno 2016), the evidence presented here for refugees arriving in the United States in the 1980s and early 1990s suggests that immigrant socioeconomic integration is enhanced in the long run through settlement assistance and social support. This takeaway is particularly relevant today, when many nations grapple with finding and implementing policies that might effectively deal with global refugee crises (Weiner 1997; Haas, Castles and Miller 2020). Research on public opinion about refugees has found that, in the United States, the public has become more receptive to admitting and providing more support to refugees and some other migrants overall (Sana 2020). While such positive changes in intangible host-society/migrant relations are encouraging, immigrant integration policies that provide tangible settlement assistance can substantially raise migrant integration, thus benefitting the overall society.
Footnotes
Acknowledgments
We thank Susan K. Brown, Kate Hee Choi, Suzanne Model, Jennifer Van Hook, and Veena Kulkarni for helpful comments on earlier drafts.
Authors’ Note
This is a revision of papers presented at the Annual Meetings of the Population Association of America, May 5-8, 2021 and the American Sociological Association, New York, NY, August 10–13, 2019.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Center for Research on International Migration at UC Irvine and the National Science Foundation (grant number DGE-1839285).
Supplemental Material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
