A New Beginning:Early Refugee Integration in the United States
The U.S. refugee population not only has grown dramatically, but the countries from which the refugees are fleeing have also diversified over the last decade. Focusing on five recent refugee groups—Bhutanese, Burmese, Iraqis, Somalis, and Cubans, we examine how premigration characteristics and postmigration integration policies shape early socioeconomic integration in the United States. Our analyses point to three findings. First, early socioeconomic outcomes show only modest differences across refugee groups, despite significant variation in premigration selectivity in human capital. Second, the two possible pathways toward integration are schooling and employment. Third, postmigration integration policies matter. Our findings highlight the role of integration policies, programs, and practices in successful refugee integration, underscoring U.S. refugee policy as a key component of immigration policy.
refugees, immigration policy, premigration characteristics, postmigration integration, socioeconomic indicators
According to the United Nations High Commissioner for Refugees (UNHCR), the world is facing an unprecedented wave of forcibly displaced people, totaling 68.5 million in 2018 (UNHCR 2019b). This number includes forty million internally displaced people within their home countries, 25.4 million refugees, and 3.1 million asylum seekers. This high level of displacement is a significant spike since 2008. Over the last decade, the numbers of forcibly displaced individuals and of refugees have grown globally by 63 and 67 percent (UNHCR 2019b). Although refugees accounted for fewer than 10 percent of the massive flow of international [End Page 117] migrants, which totaled 244 million—3.3 percent of the world's population—in 2015, differences between voluntary migrants and involuntary refugees are numerous (International Organization for Migration 2017). The latter are more likely to have fled their home country in the context of extreme trauma, political persecution, civil wars, armed conflict, or economic deprivation. Refugees therefore constitute a special category of admission in the broader legal landscape of global immigration policies in the twenty-first century.
Although the United States was a leader in the resettlement of refugees for decades, recent trends have been discouraging (Connor 2017). Since 2016, refugee admission has dropped to the lowest level in decades and numbered 22,491 for fiscal year 2018 (U.S. Department of State 2009; Tran 2020). Although the United States alone resettled more refugees than the rest of the world combined for the second half of the twentieth century, the situation has reversed over the last few years. In 2017, countries other than the United States resettled about twice as many refugees as it did (Connor and Krogstad 2018). Although Australia, Canada, and the United States all admitted fewer refugees in 2017 than they had in decades, the United States reported an annual decrease of 70 percent—the steepest drop ever. This shift signals a significant departure in refugee policy under the Donald J. Trump administration (International Rescue Committee 2018).
Despite this decline, U.S. refugee policy remains a key component of immigration policy. By definition, a refugee is a person fleeing their country because of persecution or a well-founded fear of persecution on the basis of race, religion, nationality, membership in a social group, or political opinion (UNGA 1951). Refugees are different from asylees—a related category—because refugees apply for this legal status from a country of first refuge that is outside their country of resettlement (Tran 2020). Unlike economic immigrants, refugees form a distinct group and face a uniquely negative context of exit because many fled wars, violence, political persecution, or religious persecution in their sending countries (Capps et al. 2015). Historically, refugees have been offered more state social assistance on arrival than economic migrants (Bloemraad 2006). This support can include cash assistance, medical care, language and employment trainings (Capps et al. 2015). Overall, integration policies for refugees aim to move them toward self-sufficiency and nonreliance on governmental assistance as soon as possible (Chambers 2017; Portes and Rumbaut 2014). At the same time, the limited assistance often hinders their socioeconomic integration and financial independence.
A robust literature examines the integration of previous waves of refugees, focusing on the experiences of those fleeing communist states such as Cuba, Vietnam, or the former Soviet Union (Gold 1992; Eckstein 2009; Zhou and Bankston 1998; Rumbaut and Ima 1988). Research on immigrant integration has also grown significantly over time (Waters and Pineau 2015). Nonetheless, much less attention, beyond qualitative accounts, has been paid to recently arrived refugee groups such as the Burmese, Bhutanese, and Iraqis. Scholarship on refugees, centering on new groups, has been revived in response to the dramatic rise of refugees and displaced persons worldwide (FitzGerald and Arar 2018; Gowayed 2019; FitzGerald 2019). More generally, these refugee flows form an essential part of a "global system of human mobility" (Aleinikoff 2017).
We have three broad goals for this analysis. First, we ask how socioeconomic integration of refugees into American society varies by national origin in the early years after their arrival. We examine five refugee groups: Bhutanese, Burmese, Iraqis, Somalis, and Cubans. Although Cubans are a more established group, their admission numbers spiked in the period leading to the Obama administration's decision to restore normal relations with Cuba in 2014. Given research on the Cuban experience, we use them to benchmark the experiences of other groups. Moreover, the five groups are racialized as Asian, black, Hispanic, and white in the U.S. context, raising questions on how the racialization process might shape their adaptation. Second, we ask how their premigration characteristics (educational selectivity, occupational attainment, and English proficiency) impact these five refugee groups' integration into American society. Rather than treating refugees as "blank slates" on arrival in the United States [End Page 118] (Gold 1992)—an assumption in much prior work, we ask how premigration skills might promote or hinder early socioeconomic attainment in the host society. Third, we focus on how postmigration integration policies such as language- or job-training programs affect their integration trajectories. Unlike economic migrants, refugees qualify for benefits at the federal, state, and local level that may help rebuild their lives in the United States.
This article analyzes the 2016 Annual Survey of Refugees (ASR)—the first publicly available dataset with a nationally representative sample of 1,500 refugee households admitted to the United States in the previous five years (from 2011 to 2015). Because the survey only captures respondents who had been in the United States for fewer than five years, our focus is to document their early integration. Despite the importance of early policy interventions in supporting refugees, the research on this initial, yet critical, period of adjustment into the receiving society has been lacking (International Rescue Committee 2017).
integration of refugees into american society
We begin by providing an overview of recent trends in refugee admissions into the United States to contextualize our analyses. We then review prior research on how premigration characteristics and postmigration integration policy interact to shape the integration experiences of refugees upon their arrival in the United States.
Recent Trends in Refugee Admissions into the United States
Although the notions of the United States as a land of opportunity and a land of refuge have been central to American national identity, U.S. immigration policy mostly reflects the country's shifting priorities in foreign policies and its engagement with the world (Haines 2010). This is especially the case with the U.S. refugee resettlement program, which began after Congress passed the Displaced Persons Act of 1948. This legislation admitted more than a quarter million displaced Europeans in the aftermath of World War II. As the United States entered the Cold War, the program continued to welcome refugees fleeing communist regimes in the Soviet Union over the decades that followed. The collapse of Saigon in 1975 shifted the composition of refugee flows, which increasingly came from Southeast Asia (Zhou and Bankston 1998). In the aftermath of this wave, the legal basis for the Refugee Admissions and Refugee Resettlement Programs was first established by the U.S. Refugee Act and signed into law by President Jimmy Carter on March 17, 1980.
Under this new program, the United States has admitted more than 3.4 million refugees since 1975. In the 1970s and 1980s, most arrived from Vietnam and Southeast Asia, with annual admissions peaking at 207,116 in 1980. After the fall of the Berlin wall in 1989, a significant share arrived from Europe in the 1990s. The collapse of the Soviet Union in 1992 brought refugees from the Baltic and Slavic states (Lithuania and Ukraine) as well as the Central Asian republics (Kazakhstan and Uzbekistan) to the United States. The early 2000s saw refugees arriving mostly from Africa (Sudan, Somalia and Eritrea), the influx reaching a low of 27,131 in 2002—the lowest in decades. After 2007, most arrived from Asia (Afghanistan and Bhutanese). Annual admissions peaked at 84,994 in 2016, with the majority of refugees arriving from Burma, Bhutan, Iraq, and Somalia that year.
Figure 1 illustrates these trends by graphing the annual ceiling of U.S. refugee admissions over two decades. Unlike immigration policy, the annual quota for refugees is determined by the president in consultation with Congress, as mandated in the 1980 Refugee Act. This provides both flexibility over and adaptability to the changing socio-political conditions around the world that might necessitate the adjustment of this annual ceiling. For fifteen years, this annual ceiling remained stable between seventy and eighty thousand, and the actual number of admissions fell significantly short of the quota until 2012.
From 2006 to 2016, annual U.S. refugee admissions doubled from 41,223 to 84,994 (Bernstein and DuBois 2018). From 2013 to 2015, admissions actually met the ceiling of seventy thousand. One factor behind this closer alignment between the actual number of refugee admissions and the annual refugee quota was the [End Page 119] higher need for refugee resettlement worldwide. In response, President Obama designated a ceiling of eighty-five thousand in 2016 and of one hundred ten thousand in 2017—a 57 percent increase over 2015—prompted partially by the Syrian refugee crisis, which displaced more than five million people (Gowayed 2019).
Despite decades of generous refugee policy in the United States, the trend has reversed since the 2016 presidential election. In January 2017, the Trump administration introduced a travel ban that barred entry of refugees and immigrants from seven predominantly Muslim countries (for a summary of recent travel bans and ensuing legal decisions, see Bernstein and Dubois 2018, 9). Furthermore, the refugee ceiling was cut more than half to forty-five thousand in 2018 and the proposed number to only thirty thousand in 2019. The 2019 quota is not only just the lowest in decades, but also less than half the number designated by prior administrations in the previous decades (that is, seventy or eighty thousand) under both Democratic and Republican presidents (U.S. Department of State 2009). Moreover, this proposed reduction occurred at a time when the number of refugees worldwide reached 25.4 million in 2018 (Connor 2017; Tran 2020).
How Premigration Characteristics Matter
In 2016, refugees make up about 8 percent of the foreign-born U.S. population (Kallick and Mathema 2016). Although refugees are few in number, they are diverse in national origins, human capital, and class diversity. Moreover, they are among the neediest of migrants because their reintegration into a new society is often fraught with trauma related to exit (UNHCR 2009). Many of them lived in limbo in refugee camps in countries of first asylum for decades (Capps et al. 2015; Dryden-Peterson 2016). Refugees also arrive in their countries of resettlement after extensive prescreening. They often do not choose the country that offers them resettlement, which curbs family reunification—a key channel for immigrant mobility and community creation (Bloemraad 2006; FitzGerald and Arar 2018). A robust literature has focused on the integration of refugees from Cuba (Eckstein 2009), the former Soviet Union (Gold 1992), Vietnam (Zhou and Bankston 1998), Cambodia and Laos (Hein 2006; Tang 2015), Nepal (Gurung 2015), Somalia (Besteman 2016; Chambers 2017; Voyer 2013), Syria (Gowayed 2019), and Liberia (Ludwig 2019).
Refugee integration is often arduous, given the low starting points on arrival (Portes and Rumbaut 2014; Waters and Pineau 2015). Recent refugees, despite varying levels of financial and social support, often take longer to find work, achieve economic self-sufficiency, cultivate communities, and develop a sense of belonging (Capps et al. 2015; Evans and Fitzgerald 2017; Fix, Hooper, and Zong 2017; Kallick and Mathema 2016; New American Economy 2017). They report additional barriers—including language [End Page 120] and culture, lack of social support, and emotional trauma—that often constrain their ability to achieve meaningful employment and mobility in the United States.
Refugee integration research frequently assumes that refugees arrive in the United States as blank slates. The premigration credentials and experiences of those with high human capital are discounted because foreign credentials do not translate well into the U.S. labor market. Among low-skilled refugees, a lack of English proficiency can be a real impediment. Although all refugees face challenges in finding work, barriers vary depending on refugee starting points. In light of this variation, we examine how premigration characteristics shape integration outcomes among refugees. This reveals how differences in the distribution of premigration characteristics inhibits or aids adjustment of refugees to the receiving society. This approach is consistent with research on how selectivity might shape immigrant integration into the American mainstream (Tran, Guo, and Huang 2020; Tran, Lee, and Huang 2019; Tran et al. 2018; Feliciano 2005).
How Postmigration Integration Policy Matters
Although the United States prioritizes immigration policy, it lacks any integration policy (Kasinitz et al. 2008; Portes and Rumbaut 2014; Waters and Pineau 2015). Put differently, debates are contentious about what types of immigrants the United States should admit, how many should be allowed in annually, and how the influx of undocumented migrants from the southern border can be stemmed (Amuedo-Dorantes, Puttitanun, and Martinez-Donate 2013; Donato and Armenta 2011; Waters and Pineau 2015). By contrast, little discussion has been held on what policies can help refugees and immigrants make successful adjustments into American life. Although the federal government is solely responsible for shaping national immigration policy, the task of integration often falls on both local governmental agencies and coethnic communities. This description is accurate with one important exception: integration policies for refugees are formally mandated by the Refugee Act of 1980.
In fact, the most robust integration programs target refugees (Fix, Hooper, and Zong 2017). For fiscal year 2018, the Refugee Resettlement Assistance program—run by the Department of Health and Human Services' Office of Refugee Resettlement (ORR)—was allotted a budget of more than $2 billion (Bruno 2018). These assistance programs include not only cash benefits, but also both medical and social services. Although benefits vary significantly across states (Fix, Hooper, and Zong 2017), most refugees are entitled to four to eight months of federal cash assistance, up to eight months of medical assistance, six months of employment services, five years of citizenship preparation, education and training services, and food stamp assistance through the Supplemental Nutrition Assistance Program, or SNAP (GAO 2011). Refugees are also eligible for Supplemental Security Income, Medicaid, and Temporary Assistance for Needy Families for at least five to seven years after entry (Bruno 2017).
An important goal of U.S. refugee integration policy is spatial dispersion (Bruno 2017). This placement is managed by the U.S. Refugee Admissions Program (USRAP) within the Department of State to ensure the adequate delivery of initial resettlement assistance. As a result, the assignment of refugees is based on whether the type of sponsorship is private (personal) or public (organizational). New refugees with a personal sponsorship (a blood relative) in the United States will be placed within a hundred miles of and within the same state as the sponsor. The geographical proximity helps facilitate the process of integration into local communities. Refugees without any U.S. ties are often dispersed across the country such that no single local community will be burdened by the arrival of refugee groups. Under USRAP's supervision, a national network of public and private nonprofits—often referred to as voluntary agencies—work closely with their local affiliates to assist refugees with basic needs on arrival, including housing, home furnishings, food, school enrollment, training programs, and employment services (Nawyn 2011). In contrast to those with a personal sponsorship, refugees with an organizational sponsorship must be placed within fifty miles of and within the same state as their local affiliate.
The Reception and Placement program, also [End Page 121] run by the State Department, has resulted in significant variations across states. For example, Audrey Singer and Jill Wilson (2006) find that three-quarters of refugees who arrived in the United States between 1983 and 2004 settled in thirty metropolitan areas with significant foreign-born populations, including New York, Los Angeles, San Jose, Chicago, and Minneapolis-St. Paul. Beyond these gateways, new refugee arrivals into smaller communities can have a profound impact, especially when refugees also make up the majority of the foreign-born population in these areas (Besteman 2016; Chambers 2017; Ludwig 2019). In smaller communities such as Utica, New York, or Fargo, North Dakota, the arrival of refugees helps reverse population decline or economic stagnation. In mid-size communities such as Fresno, California, or Springfield, Massachusetts, refugees contribute to ethnic diversity in local communities.
At the same time, the current placement system has several disadvantages. First, the Office of Refugee Resettlement relies on a mix of public agencies and public-private partnerships to administer refugee assistance program (GAO 2011). Despite the billions of dollars spent annually to resettle between seventy and eighty thousand refugees, little research is undertaken on how effective the current programs are in helping refugees find work or achieve financial self-sufficiency (GAO 2011). Second, a personal sponsorship is often more effective than an organizational one (Larsen 2011). Given the lack of familial ties and coethnic community, many refugees with language barriers must navigate the myriad social services available to them at the local level on their own, often with mixed success. Furthermore, formal integration policies, though well intended, often cannot compensate for the lack of community and the feeling of isolation among refugees. Finally, high rates of mental illness and trauma among refugees remain unaddressed, which compromises their emotional well-being and impedes social integration (Ao et al. 2015; Meyerhoff, Rohan, and Fondacaro 2018).
The randomization of refugees in the United States contrasts starkly with practices in other countries, such as the United Kingdom or Canada. In the UK, the recent Syrian Vulnerable Persons Resettlement program matches refugees to a local authority, which is responsible for the refugees' initial placement. Not only is participation of these local authorities entirely voluntary, but the UK system takes great care in achieving a "good" match between the refugees and local areas to prevent secondary migration of refugees within the UK (Jones and Teytelboym 2017). The program is jointly run by the Home Office, the Department for International Development, and the Department for Communities and Local Government. Although refugees do not fully rank the 350 "local authorities" in the UK, the Home Office does ask refugees to express preferences "over the types of local areas that refugees prefer" to resettle (Jones and Teytelboym 2017, 676). Canada adopts a similar matching system under three types of refugee resettlement schemes: government assistance, private sponsorship, and a "blended program" mixing both schemes.
five recent refugee groups at a glance
The five largest refugee groups in our survey—Bhutanese, Burmese, Iraqis, Somalis, and Cubans—merit a statistical and historical overview. Figure 2 presents trends in numbers of refugee admissions from these five sending countries from 2011 to 2018. The graph focuses on the 2011 to 2015 period because it coincides with the years of entry for the sample of refugees in the 2016 ASR. To provide the most recent snapshot, we extend the line to include data from 2015 to 2018. Between 2011 and 2015, approximately fifty thousand refugees from the five groups arrived each year in the United States. Altogether, these five groups make up the majority of all refugees resettled during this period, accounting for a high of 84 percent of total U.S. refugee admissions in 2011 and a low of 67.5 percent in 2015. Since 2016, the number has dropped precipitously. Only 6,180 refugees were admitted in 2018, about 10 percent of the total from the same countries at the peak of the influx in 2014. Figure 2 also shows a shift in the composition of refugees admitted over the same period.
From 2011 to 2015, Burmese and Iraqis were the majority, followed by Bhutanese, Somalis, and Cubans. In 2011 and 2012, Bhutanese and [End Page 122]
Burmese made up the lion's share of new arrivals. From 2013 to 2015, the Burmese refugee flow continued unabated and the Iraqi flow sharply increased. The declines since 2016 are due to changes both in the political priorities of both the Obama and the Trump administrations and in the state of conflict in the sending countries of origin. During this period, the Iraq and Somali conflicts wound down. In 2017, the United States officially ended the "wet foot, dry foot" policy toward Cuba, which led to a decline in Cuban admissions. In 2018, the United States admitted 257 Somalis, 140 Iraqis, and no Cubans, effectively ending the Cuban refugee flow that had begun in 1959 (Portes and Rumbaut 2014). This shifting composition reflects the global dynamics of displacement. Increasing hostility toward asylees, refugees, and immigrants is also a factor behind this drop (Connor and Krogstad 2018; Kerwin 2018). Since 2019, refugee resettlement has dropped to a mere trickle.
On arrival, these refugees are settled in virtually every state in the country. Many also engaged in secondary migration in search of better job opportunities and coethnic communities. Table 1 presents the top three states in which each of the five groups settled from 2011 to 2015. One-third of Bhutanese resettled in Pennsylvania, Ohio, and New York; a similar proportion of Burmese ended up in Texas, New York, and Indiana. Slightly less than half of Iraqis arrived in California, Michigan, and Texas; a quarter of Somalis resettled in Minnesota, New York, and Texas. Cubans are significantly more spatially concentrated—68 percent resettled in Florida, and fewer than a thousand in Texas and Nevada. Despite dispersion policy under USRAP, these numbers show remarkable geographical concentration, Texas and New York being the most common destinations for recent refugees from all five groups.
Bhutanese refugees began arriving in the United States in 2007, following a UNHCR decision to resettle those living in Nepalese refugee camps (Trieu and Vang 2015). Many are descendants of ethnic Lhotshampas—Nepalese migrants settling in southern Bhutan. Lhotshampas differ from the Drukpa ethnic majority in Bhutan in their adherence to Hinduism rather than Buddhism (Rizal 2004). Between 1990 and 1993, about a hundred thousand Nepalese-speaking Bhutanese fled Bhutan for refugee camps in eastern Nepal. They languished in these camps for two decades, unable to repatriate to Bhutan or to receive formal refugee status from the UNHCR (Trieu and Vang 2015). Following a change in policy by UNHCR, a total of 5,244 Bhutanese refugees arrived in the United [End Page 123]
States from these refugee camps in 2008 (Vang et al. 2014; Capps et al. 2015).
The Bhutanese are the second smallest Asian ethnic group in the United States (after Mongolians), an estimated population of only twenty-four thousand in 2015 (López, Ruiz, and Patten 2017). From 2009 to 2011, Bhutanese men worked at higher rates than their U.S.-born counterparts, but the rates are significantly lower among Bhutanese women (Capps et al. 2015). This partially reflects their rather unequal education: a quarter of Bhutanese men have at least a college degree, but about half of Bhutanese refugees, especially women, have yet to complete high school (Capps et al. 2015).
Bhutanese refugees have particularly high levels of depression and suicide. The suicide rate is not only twice that of the U.S. national average (Ao et al. 2015; Meyerhoff, Rohan, and Fondacaro 2018), but also comparable to the rate among Bhutanese still living in refugee camps (Vonnahme et al. 2015; Hagaman et al. 2016). Laura Vonnahme and colleagues (2015) find that men who viewed themselves as family providers tended to demonstrate depressive symptoms at a rate four times higher than men who were not providers. Difficulty in finding work or periods of unemployment can undermine their traditional role as the provider. Depressive symptoms are exacerbated among those reporting to be in poor health, who are also most likely to experience depression (Vonnahme et al. 2015). Scholars still debate the cultural and structural factors for these high rates, but suicide poses a serious health risk to this community.
Burma (or Myanmar) is a diverse country with regards to ethnic and religious minorities.1 The people of Burma are one of eight ethnic groups. Burman is the largest, followed by Chin, Kachin, Karen (Kayin), Mon, Arakhan (Rakhine), Shan, and Karenni (Kayah). Burman, Mon, Arakhan, and Shan are primarily Buddhist, whereas Chin, Kachin, Karen, and Karenni are primarily Christian. Following the uprising on August 8, 1988, the oppression of ethnic minorities and political dissidents by the ruling military regime led to the displacement of more than one million people who settled in massive refugee camps in neighboring countries, including Bangladesh, India, Malaysia, and Thailand (Vang et al. 2014). There, many lived in limbo for decades; only 3,528 were resettled in the United States between 1984 and 2004 (Furukawa and McKinsey 2009). The dramatic rise of Burmese refugees in the United States is a fairly recent phenomenon: 77,265 were admitted from 2005 to 2011 (Vang et al. 2014).
To date, the Myanmar conflict remains one of the world's largest humanitarian crises, the Rohingya conflict in particular. The Rohingya people are a predominantly Muslim ethnic minority in Rakhine State at the northern edge of Myanmar. In 2017, seven hundred thousand Rohingya [End Page 124] fled Myanmar to Bangladesh to avoid ethnic cleansing, mass killings, sexual violence and widespread arson from the Burmese security forces. The number of Rohingya refugees of Muslim background was estimated at more than one million by the end of 2018 (UNHCR 2019a). Despite this group's significance, they are not among the Burmese refugees we examine here, given the timing of migration.
The Burmese—estimated population of 168,000 in 2015—are one of the smallest Asian ethnic groups in the United States (López, Ruiz, and Patten 2017). According to an analysis of Current Population Survey (CPS) data (Trieu and Vang 2015), the population is a fairly young cohort, 67 percent under the age of forty. The majority are first generation (58 percent), though the 1.5 and second generation is growing (38 percent). Language is a major barrier for Burmese, a majority speaking either no or very poor English. About a third of those in the United States live below the poverty line with an average income of $25,901 (Trieu and Vang 2015). Both men and women report working at a lower rate than U.S. natives (Capps et al. 2015), despite their strong desire for work (Trieu and Vang 2015).
Burmese refugees struggle to achieve economic self-sufficiency (Trieu and Vang 2015). As with other refugees, resettlement agencies provide crucial support in their initial resettlement. When state support ends, Burmese turn to voluntary agencies and self-help organizations, which emerged as a way to organize and leverage community resources to aid their integration (Trieu and Vang 2015). Yet these agencies are often inadequate to serve diverse cultural and linguistic needs of new Burmese arrivals. This internal diversity poses an additional challenge for the Burmese because social service providers in the United States often adopt a one-size-fits-all policy that fails to account for the internal cultural and social differences (Brown and Scribner 2014; Kerwin 2018).
The fall of the Saddam Hussein regime and the ensuing instability of the Iraq war brought a steady flow of Iraqis, mostly from Baghdad, to the United States (Capps et al. 2015). Many settled in California, Michigan, and Texas. At the city level, San Diego, Dearborn, and Detroit have the largest populations (Shoeb, Weinstein, and Halpern 2007). Iraqis, however, do not form a cohesive community in the places they settle. Internal divisions along socioeconomic lines and along ethnicity—Arab, Kurdish, and Chaldean—are important markers of difference (Shoeb, Weinstein, and Halpern 2007). In many of these places, Iraqis also identify discrimination by U.S. natives as a potential barrier to integration (Jamil et al. 2012; Shoeb, Weinstein, and Halpern 2007)
Iraqis are a relatively educated refugee group. Relative to the overall U.S. population, Iraqis report similar college completion rates but higher levels of high school drop-out rates (Capps et al. 2015). At the same time, they report the lowest employment rates of any refugee group in the country, men and women each working at lower rates than the U.S. natives. On average, Iraqis report working low-wage jobs, often earning annual incomes of $20,000 or less (Capps et al. 2015; Shoeb, Weinstein, and Halpern 2007). Additionally, Iraqi refugees report unemployment rates double that of Iraqi immigrants with comparable educations and credentials (Jamil et al. 2012). Potential hindrances to employment for Iraqis include language skills, knowledge of the U.S. professional world, and health (Jamil et al. 2012). Given the limited assistance from the state, Iraqi refugees are likely drawing on their savings or financial resources from their home country to make ends meet. They might still be engaged in businesses in their sending country or work in the informal economy within their ethnic enclave, resulting in an underreporting of their employment. Iraqi refugee children also struggle because of the considerable gaps in their educational history (Bang and Collet 2018). These gaps result from transitionary periods characterized by significant instability in Iraq—where an active armed conflict interrupted school attandance—and by educational restrictions imposed in transitional countries. Integration into local U.S. schools is further complicated by the emotional trauma experienced during displacement (Bang and Collet 2018).
Somali refugees began resettling in the United States following the ouster of Mohamed Siad [End Page 125] Barre in the 1990s (Chambers 2017; Golden, Garad, and Boyle 2011). From 2010 to 2016, about nine thousand Somali refugees arrived each year. They are most likely to first settle in Minnesota, New York, and Texas. Other sizable Somali communities have formed in Lewiston, Maine, and Columbus, Ohio (Voyer 2013; Waters 2013; Chambers 2017). Minnesota is the destination for a considerable secondary Somali migration flow from other states and from Toronto, Canada (Horst 2006). The Twin Cities boast the largest Somali populations, especially within St. Paul and Minneapolis (Chambers 2017; Waters 2013). Detailed data on Somalis are not available at the city level. However, East Africans2—a census category that includes Somalis—accounted for 23.5 percent of the immigrant population in Minneapolis and 15.9 percent in St. Paul from 2009 to 2013 (Chambers 2017). They are also the largest and second largest ethnic group in these cities, respectively. Lewiston, Maine, where many arrived after initially settling in Georgia in the 2000s, is now home to about five thousand Somalis.
Once settled, Somalis often find low-skilled work, supplementing their income with informal entrepreneurial work (Chambers 2017). Given their concentration in blue-collar work, Somali refugees report extremely low wages, with average annual earnings ranging from $13,370 in Saint Paul to $11,414 in Minneapolis, Minnesota. These numbers also partially reflect the relatively high rates of informal work (Chambers 2017). In the Twin Cities, many Somalis are employed in meat-packing operations, whereas factory and warehouse work are the most common sources of employment in Columbus (Chambers 2017). More entrepreneurial Somalis also start their own small businesses, Somali businesses appearing in large malls (suuqs) and storefronts (Chambers 2017). Despite their economic hardships, financial remittances—xawilaad—back to Somalia or related refugee camps are rather common (Horst 2006; Lindley 2009). Over the last decade, this inflow of financial remittances has totaled $1.3 billion annually, 16 percent coming from the United States (Chambers 2017; Waters 2013).
Somalis encounter significant variation in their reception. In the Twin Cities, they have access to a robust system of nongovernmental organizations and state assistance such as mental health services and business development groups (Chambers 2017). This positive reception reflects the city's long history in welcoming refugees, including Jewish refugees after World War II and Hmong refugees in recent decades (Chambers 2017). Columbus, by contrast, offers a more negative reception, including stronger anti-immigrant rhetoric and more limited service provision (Waters 2013). In Lewiston, despite welcoming official rhetoric, concerns are circulated about their fiscal burden on the city (Voyer 2013). Across cities, Somalis also report discrimination experiences in most settings, usually on the basis of their racial or religious background (Voyer 2013; Waters 2013; Chambers 2017).
Cuban refugees have been in the United States for more than half a century. Strictly speaking, they do not qualify as a new refugee group. For three decades following the rise of Fidel Castro, the United States admitted Cuban refugees on the basis of the Cuban Admissions Act of 1966, which granted any Cuban the expedited and virtually automatic right to legal permanent residency, and thus an eventual pathway to citizenship once they were in a position to apply for naturalization (Barrios 2011). The fall of the Soviet Union in the 1990s led to a significant increase in Cubans fleeing the island by boat, many of whom were apprehended by the Coast Guard in territorial waters (Barrios 2011). This surge in balseros, or rafters, gave rise to the wet-foot, dry-foot policy in 1994. Under this agreement, Cubans who reached U.S. soil would be given asylum; those caught at sea were sent back to Cuba (Henken 2005).
Cubans are among the most studied refugee groups (Portes and Bach 1985; Portes and Rumbaut 2001; Portes and Stepick 1993). Cubans, [End Page 126] whose U.S. population numbers some two million people, are the fourth largest Hispanic group behind Mexicans, Puerto Ricans, and Salvadorans (Krogstad 2017). The United States has resettled more Cuban refugees than any other group except the Vietnamese (Hooper et al. 2016). Cuban refugees in the United States benefit from a support network of established coethnics. New arrivals are able to access strong networks of economic and cultural support established by previous generations of Cubans. In Miami, the most populous area of Cuban settlement, several vibrant ethnic enclaves include Cuban grocers, barbershops, professional organizations, and other small businesses (Eckstein 2009; Henken 2005).
In 2014, the Obama administration announced plans to normalize relations with Cuba. This led to a final surge in Cuban refugees seeking to take advantage of the wet-foot, dry foot policy in the period leading to 2017. In 2016, 56,406 Cubans entered the United States through ports of entry—31 percent more than the previous year (Krogstad 2017). In January 2017, the Department of Homeland Security officially ended the wet foot, dry foot policy as a part of this process. This change made Cuban nationals entering the United States illegally subject to removal for the first time since the Cuban revolution in 1959 (White House 2017). In 2018, the number of refugees from Cuba was zero, effectively ending this continuous flow of refugees.
data and methods
To examine patterns of labor-market integration among recently arrived refugees, we use data from the 2016 Annual Survey of Refugees.
The 2016 Annual Survey of Refugees
Since the 1980s, the Office of Refugee Resettlement has administered the ASR annually. However, only the 2016 version has been made available to the research community. Its public availability offers a unique opportunity to understand the early integration of refugees. For our purpose, the dataset provides a good sample size for five recent refugee groups, along with detailed survey questions about both the premigration characteristics and postmigration integration policies, which are the core of this article's empirical contributions.
Prior work on refugee integration in the United States has relied on a patchwork of datasets. These include national surveys, administrative data sources, and some primary sources (Bernstein and DuBois 2018, table 3). National samples include the American Community Survey (ASC) from the U.S. Census Bureau, the New Immigrant Survey, and the ASR. Administrative sources include the Worldwide Refugee Admissions Processing System, which provides selected characteristics on all refugees, including their initial state of resettlement in the United States. Other administrative sources are refugee resettlement program data from the Bureau of Population, Refugees and Migration in the Department of State and ORR. Last, some researchers and refugee-assisting agencies collected surveys and interview-based data on specific refugee groups, but these data sources are either smaller in scope or not nationally representative.
The 2016 ASR is the only national survey that specifically tracks the progress of refugee integration, the results of which are used to fulfill congressionally mandated reporting. The 2016 ASR contains a representative sample of 1,500 refugee households that entered the United States between fiscal years 2011 and 2015. This sample was drawn using a stratified probability sample from 141,000 principal applicants (PAs) contained within the ORR Refugee Arrivals Data System. For refugee families, the PA is the family member whose refugee case is used as the basis of the admission application. The surveys were conducted by telephone interview after an introduction letter and $2 cash incentive were sent by postal mail. These interviews were offered in the seventeen most common languages spoken by refugees, covering 77 percent of the linguistic variation of the refugee population that has arrived in the last decade. All respondents received a $25 gift card via first-class mail for their participation in the survey.
The analysis is limited to 1,500 respondents (eighteen years or older) who have PA status. Although the survey sampled 4,037 refugees, including household members within the same refugee family, we focus only on the 1,500 PAs because they are most likely to be the head of household within their family. By definition, these respondents are first-generation immigrants. [End Page 127] The majority of household heads are men (1,273); one in six are women (227). By contrast, most second household members enumerated are women (957), typically the spouse of the household head. These numbers reflect traditional gender norms within these refugee communities, in which men are more likely to be the household head.
The 2016 ASR has a number of strengths. It contains the most recent data on a number of refugee groups. It also includes key variables on both pre- and postmigration characteristics. The premigration variables—English proficiency, education, and occupation—help us understand issues of immigrant selectivity and how it affects their integration. The postmigration variables—training programs and cash assistance—provide a unique glimpse into the role of integration policy in shaping refugee experiences. At the same time, the 2016 ASR has a few limitations. First, it has an adequate sample size for only five groups and does not fully capture the heterogeneity in refugee experiences. Second, it focuses only on the early period of adjustment for refugees in the United States. Thus it provides no information on the long-term outcomes of these groups. Third, it includes many variables on socioeconomic integration, but no information on other outcomes, such as political participation or residential patterns. Fourth, the sample size is too small for reliable state- or county-level analyses in light of the important role of local policies in incorporating refugees. Although the ASR collected respondents' state of residence, this variable was recoded to regions to protect the anonymity of respondents. Finally, it provides no information on where and for how long refugees lived after leaving their country of birth and before they were resettled in the United States.
We examine eight outcome variables on the early socioeconomic integration of refugees. Because education and work are the two pathways to integration into American society, the first four dependent variables are attending school within the past twelve months, working at a job anytime last week, number of hours worked at all jobs last week, and hourly wages at primary job last week. Because hourly wages are not normally distributed, we use the log version of hourly wages. The next four outcomes capture postmigration occupational attainment by focusing on four occupational categories: professional, sales, service, and blue-collar work. The first variables on attending school and working, along with the last four variables on occupations, are dichotomous; hours worked and hourly wages are continuous.
Our first set of independent variables are the national origin of refugees. This variable is categorical with six categories: Bhutanese, Burmese, Iraqi, Somali, Cuban, and Other. Other is a residual category that combines refugees from the other sending countries with fewer than one hundred respondents in the sample. It is inherently not meaningful as a category, but its inclusion allows us to retain the largest sample size possible for our analyses. One primary goal of this article is to compare the integration experiences among these recent refugee groups, so we are interested in how each refugee group fares relative to the others. In a standard analysis, we would like to compare the socioeconomic outcomes of these five refugee groups with the native-born majority group (non-Hispanic whites) to benchmark their progress. One limitation of the ASR is its exclusive focus on refugees: it does not include any nonrefugee groups. As a result, we use Cubans—the most established among the five refugee groups—as our reference category in our analyses because we would like to know how the other groups fare relative to Cubans.
National origin is important because we expect refugee groups that arrived with more resources and positive context of reception to become better integrated into American society than groups with fewer resources and a negative context of reception (Portes and Rumbaut 2001). We use Cubans as the reference group to probe the role of ethnic enclaves in shaping the early integration of refugees (Martén, Hainmueller, and Hangartner 2019). Some of these refugee groups arrive in localities with few coethnics and this lack of coethnic community would hinder their initial integration. Furthermore, despite ethnic heterogeneity within each of these national groups, individuals with different ethnicities [End Page 128] from the same national origin can face systematically similar experiences, such as contexts of exit in the sending country and contexts of reception in the United States.
We include three additional sets of independent variables in our analyses. The second set focuses on demographic characteristics: age, the quadratic term of age, gender, age at arrival, marital status, parental status, legal permanent residency, region and length of residency. Age of arrival is categorical with four categories. Marital, parental, and legal statuses are all dichotomous. Region is categorical with four categories. Length of residency is continuous and measured by the number of months at the current address.
The third set of variables controls for premigration characteristics prior to arrival in the United States. These include years of schooling, English proficiency, and prior occupations in the home country to show how selectivity affects refugee integration experiences. Years of schooling is continuous whereas English proficiency and occupation are both categorical. Beyond four major occupation categories is another for students and a residual one for Other.
The fourth set of variables controls for postmigration integration policies. These include enrollment in English-language, job-training, refugee assistance, and nonprofit cash assistance programs over the last year. The variables are dichotomous. These integration policies provide refugees with initial assistance on resettlement and other training to facilitate their reentry into the labor market and economic self-sufficiency. We are interested in English-language and job-training programs as potential paths to integration. These two variables are based on two survey questions that asked whether the respondent has attended an English-language training program or any job-training program within the past twelve months. To account for sequencing of job training, we analyze job training in the last year in relation to employment in the last week. As a result of this temporal order, we are confident that job training precedes employment, as opposed to the other way around. On school attendance, an English-language program should precede school or university attendance because English proficiency is often a prerequisite for university enrollment. Additionally, the possibility that the respondents might conflate attending school or university with attending an English-language training program is unlikely, given the survey's careful wording: "Within the past twelve months, has [the respondent] attended school or university (other than to take English-language training or the job-training class indicated in the previous question)?" Finally, although we are interested in the impact of refugee sponsorship (private versus public sponsorship) on integration outcomes, the ASR has no such variables. Table A1 provides all descriptive statistics for the independent variables.
The analyses proceed in two stages. First, bivariate analyses provide a snapshot of early socioeconomic integration for the five refugee groups. Second, multivariate analyses focus on how the three sets of independent variables shape the integration process. For dichotomous outcome variables, we use logistic regressions with simulated standard errors and report the odds ratios. For continuous outcome variables, we use standard ordinary least squares regressions. For dichotomous outcome variables, the logistic regression models for measuring group differences are as follows:
where denotes the log odds of the probability of a particular socioeconomic outcome (Y) for respondent i. Ni is the national origin for respondent i—the main variable of interest. Di is a vector of demographic control variables for respondent i. Pre_Mi denotes a set of variables on respondents' premigration characteristics. Post_Mi denotes a set of postmigration integration policies. Because Cubans are the most established in the United States, they serve as the reference group and the benchmark for the integration of other groups in multivariate analyses. These analyses also adjust for the stratified survey design by using the appropriate final weights provided by the 2016 ASR. [End Page 129]
All missing data were imputed using a multiple imputation procedure. Multiple imputation is a flexible, simulation-based statistical technique for handling missing data (StataCorp 2015; Rubin 1987). This procedure rests on the assumption that data are missing at random, conditional on the observable individual-level covariates (StataCorp 2015). We adopt the MICE (Multivariate Imputation via Chained Equations) package in R. The percentage of missing values across the variables we used ranged from 0 to 22 percent, though only 3 percent of total records were missing. Following Stef van Buuren (2018), we use the default techniques for the four types of variables. The imputation process follows predictive mean matching for continuous variables, logistic regression for dichotomous variables, proportional odds model for ordinal variables, and multinomial logistic regression for categorical variables. The predictors used in the imputation equation include the individual-level covariates in the final models. For wages, we first imputed the missing values before transforming the variables into log wages. To lessen the Monte Carlo error from the simulation, we generated two separate datasets based on five and twenty imputations, but the results are substantively similar, suggesting that our findings are robust in regard to the missing data that were imputed. In addition, the distribution of imputations and the convergence of the algorithm that produced them were inspected visually to ensure that these were reasonable. This article reports results from our imputation with m = 20. We estimate a full set of coefficients for each of the twenty imputed datasets before pooling using Rubin's rules across the twenty estimates using the pool() function in MICE. In R, one standard way to report standard errors when converting logits to odds ratios is simulation, as we have done. The standard errors and the confidence intervals are uneven in our odds ratio plots because these represent the simulated distribution (n = 1,000) of the odds ratios using the covariance matrix at the 95% level, so these are actually more accurate to the distribution of our data. Our final sample is 1,496 instead of 1,500 respondents because four of them reported being younger than eighteen.
patterns of early socioeconomic attainment among refugees
In this section, we begin with our descriptive findings by refugee group. We then present our regression models, which use premigration characteristics and postmigration integration as key predictors of schooling and employment.
Table 2 presents descriptive statistics for the eight dependent variables by refugee group. Somalis (19 percent) and Iraqis (14 percent) are
[End Page 130]
the most likely to attend school. Cubans are the least likely to do so. By contrast, Cubans (77 percent) and Burmese (75 percent) are most likely to work. Iraqis and Somalis are least likely. Among those who reported being employed during the week prior to the survey, Cubans report working the highest number of weekly hours and Somalis the fewest. Despite variation in educational attainment across groups (see appendix table A1), the average hourly wage is universally low for all groups (approximately $12 per hour), indicating that refugees are mostly concentrated in low-wage work.
On occupation, refugees from every group disproportionately report being in service and blue-collar work. Many are in sales, but few report working in a professional occupation. First, on professional occupation, Burmese and Somalis report the lowest rate (3 percent) and Bhutanese the highest (8 percent). Second, Bhutanese, Iraqis, and Somalis are about twice as likely as Burmese and Cubans to be in the sales industry. Third, at least a quarter of respondents from every group report working in the service sector. Finally, Burmese are the most concentrated in blue-collar work (42 percent) and a third from the other refugee groups are also in blue-collar work.
Figure 3 presents educational achievement for both refugees and nonmigrants in sending countries by national origin. Refugees from Burma and Somalia are negatively selected. 9.1 percent of nonmigrants in Burma have a bachelor's degree or more, relative to only 3.9 percent among Burmese refugees. Similarly, the proportion with a bachelor's degree or more is 5.8 percent in Somalia, but only 4.1 percent among Somali refugees in the sample. By contrast, refugees from Bhutan and Iraq are positively selected. Bhutanese refugees in the ASR are twice as likely as Bhutanese nonmigrants to have a college education, whereas Iraqi refugees are three times more likely than their counterpart nonmigrants to have one. Among Cubans, the slight edge in education is among the refugees in the sample over nonmigrants. These patterns of selectivity, in turn, should [End Page 131] shape refugees' early integration. In particular, we expect Burmese and Somalis to report the worst outcomes and Iraqis and Bhutanese to report the best outcomes in their early integration, given human capital on arrival in the United States. For Cubans, we hypothesize that the established coethnic community will provide them with an advantage, given the availability of employment opportunities for new refugees. Finally, variation is substantial in the rate of legal permanent residency by national origin, from 49 percent among those from Bhutan and to 98 percent among Iraqis. This variation is consistent with findings in other work, which attributes these differences to sociodemographic characteristics and initial resettlement location, among other factors (Mossaad et al. 2018).
Early Indicators of School and Work Among Refugees
Table 3 presents multivariate results from logistic regressions predicting schooling and employment for the five refugee groups while accounting for the three sets of demographic, premigration and postmigration covariates. We estimate two models for each dependent variable. The first examines the role of premigration characteristics in shaping integration, controlling for national origin and demographic characteristics. The second introduces postmigration integration policies to examine their impact on pathways to integration.
On school attendance, model 1 shows that Somalis are 4.2 times more likely than Cubans to attend school, controlling for the full set of demographic variables. Being married or being a legal permanent residency reduces the likelihood of school attendance by about half. Among the premigration characteristics, prior education positively predicts school attendance. This suggests that those who are more educated on arrival are more likely to pursue school to regain their credentials. Finally, those in blue-collar work in the home country are half as likely to report being in school as those working in service occupations.
Model 2 introduces postmigration integration policy. First, Somalis are 3.2 times more likely than Cubans to attend school. The coefficients for marital and legal status change slightly, but remain significant. Among the postmigration variables, the strongest predictor of being in school is enrollment in an English-language training program. Those who indicate attendance in such a program are five times more likely to be enrolled in school.
Models 3 and 4 show that Cubans are significantly more likely than the other five groups to report working, controlling for other observable covariates. Specifically, Cubans are three times more likely than Bhutanese and Burmese, six times more likely than Iraqis, and four times more likely than Somalis to be employed. Men are about four times as likely as women to work and those with lawful permanent residence (LPR) status are 1.4 times more likely to work. Most surprisingly, none of the premigration variables reach statistical significance in model 5, suggesting that refugees are treated as "blank slates" on arrival in the United States. Their prior credentials and work experiences do not predict their short-term socioeconomic outcomes. Among postmigration integration policies, refugees participating in an English-language training program are one-third less likely to report working whereas those who had received job training in the last twelve months are 2.8 times more likely to be employed in the week before the survey.
Overall, we observe an association between integration policy and economic integration. Moreover, these findings suggest that participation in different training programs puts refugees on different integration paths. Language-training enrollment positively predicts school attendance whereas job-training participation predicts work. In fact, the coefficient for English-language training program is positive for schooling (in model 2) and negative for employment (in model 4), pointing to two possible integration paths training programs can provide. At the same time, our findings cannot address the effectiveness of these programs. We do not know whether the respondents attended job-training programs for the jobs they eventually receive, although job-training programs likely target specific industries or sectors (health care or technology, for example). [End Page 132]
[End Page 133]
Table 4 shifts the focus to hours worked and hourly wage among those who are working. The most consistent and significant effect across all four models is gender: women not only work eight fewer weekly hours but also report earning hourly wages3 that are less than one third those of men. On hours worked last week, model 1 shows that, on average, Cubans work 6.2 more hours than Iraqis and 4.5 more than Somalis. In model 2, Cubans still report working 5.6 more weekly hours than Iraqis. Among the premigration variables, every additional year of schooling in the home country is associated with a quarter fewer work hours in model 2, but has no impact on log hourly wages.
On log hourly wage, no significant differences by national origin are observed. This flattening of hourly wages, despite educational selectivity across groups, is likely for one of two reasons. First, refugees are disproportionately concentrated in low-wage, entry-level jobs, at least in the short term, where wages vary little. Second, significant barriers in the conversion of foreign credentials means that refugees with [End Page 134]
[End Page 135] more education in the home country cannot translate their training to U.S. employers. That none of the other premigration variables is significant is further evidence of the devaluation of foreign credentials in the United States. On postintegration policies, those receiving nonprofit cash assistance report 8.9 fewer weekly work hours but those who participated in job-training programs report earning 21 percent ([exp (0.190)–1]*100 = 20.9) more.
Figure 4 presents odds ratio plots and predicted value plots for socioeconomic attainment based on models 2 and 4 from tables 3 and 4, controlling for other covariates. This figure shows a clear pattern by refugee group. Somalis are most likely to be in school and Cubans are most likely to work. Iraqis are not only the least likely to work relative to Cubans, but also report the lowest work hours when they do. Given their relatively high level of education, Iraqi's lower propensity to work reflects the group's higher reservation wages—the lowest [End Page 136]
wage rate at which an Iraqi refugee is willing to accept a particular job, especially in light of the low-wage work that is more readily available to them.
Occupational Distribution Among Refugees
Table 5 presents a mobility matrix of pre-and postmigration occupational patterns. Specifically, respondents report one of these four occupational categories: professional, sales, service, and blue-collar work.4 Among those who worked as a professional in their country of origin, only 22 percent report working in a professional occupation in the United States. Put differently, more than three-quarters of professionals experience downward occupational mobility on arrival in the United States. In addition, 32 percent of those in the sales sector, 43 percent of those in the service sector, and 46 percent of those in blue-collar work in the home country report working in the same occupational category after migration. These findings point to stability in low-status jobs and major shifts in high-status jobs between pre-and postmigration reported occupations.
Table 6 presents the multivariate regressions on postmigration occupational attainment, controlling for the full set of covariates. Cubans are five times more likely than Iraqis and twelve times more likely than Somalis to work in a professional occupation. By contrast, Somalis are three times more likely than Cubans to be in sales. No significant differences by national origin are apparent in other occupations. Among other covariates, females are 2.5 times more likely to report being in a professional [End Page 137] occupation and males three times more likely to be in blue-collar work. This female professional advantage among recent refugees is due to women's higher concentrations in teaching and healthcare professions.5 Age at arrival also matters. Relative to those who arrived after the age of fifty-five, refugees who arrived before the age of twenty-five are twenty times more likely to work as a professional or in the blue-collar sector. By contrast, refugees in the former age group are 8.3 times more likely to do service work. Finally, those with LPR status are 2.5 times more likely to be a professional and 1.4 times more likely to be in service.
The impact of premigration characteristics varies by occupation. First, those with higher English proficiency are ten times more likely to be in a professional occupation. Second, those with more education report higher likelihood of working in sales. Third, those who worked in the blue-collar sector or who reported being a student in the home country are less likely to be in service work. Among postmigration integration policies, none has an impact on occupation in the United States. The only exception is blue-collar work: refugees who participated in a job-training program are 1.5 times more likely to work in this sector. Active integration policies like job training seem to channel refugees into blue-collar work as opposed to other professions.
Figure 5 presents odds ratio plots for occupational attainment based on results in table 6. It reveals a clear pattern of occupation distribution by refugee groups. Cubans are more likely to be in a professional position than Iraqis and Somalis whereas Somalis are most concentrated in sales. Finally, all five groups are roughly equally concentrated in sales and blue-collar work.
How Selection Shapes Enrollment in Training Programs
Although we identify English-language training and job-training programs as two possible pathways for refugee integration, it is possible that some selection process might shape enrollment in each program. We address this issue by examining the characteristics that might predict enrollment in one program versus the other. Because participation in the two training programs predicts school and employment in the United States, we also focus on the role of premigration school and work patterns in shaping this postmigration decision. For example, refugees with low English proficiency and those who were in school before migration might be more likely to enroll in an English-language training program in the United States. By the same logic, those who were employed before migrating will be more likely to enter a job-training program in the United States.
Empirically, we fitted two sets of logistic regressions with the enrollment in an English-language [End Page 138]
[End Page 139] training or a job-training program as the dependent variables. We control for national origin, the full set of demographic variables, and premigration characteristics. We summarize these findings (full results available on request).
In regard to English-language training programs, Somalis are more likely than Cubans to enroll. Women are 1.75 times more likely than men to be enrolled in English classes. Moreover, those with greater English proficiency are less likely to enroll (self-select out of classes). Besides these, no other coefficients are significant, including the premigration school and premigration employment variables. This is evidence that those enrolling in English classes were not simply students in the country of origin and are therefore continuing a path into school by self-selecting into English classes.
The job-training program also has few significant predictors in the model, other than premigration employment and LPR status. Those who report working in the sending countries are twice as likely to enroll in a job-training [End Page 140] program in the United States. Moreover, those with LPR status are 1.5 times more likely to enroll in job training. Iraqis are significantly less likely to be enrolled than Cubans—evidence that some selection is driving a part of the effect that job training has on the likelihood of employment. Those who worked in their home country are more likely to be in job-training programs. These participants, in turn, are also more likely to pursue work as an integration pathway in the United States.
discussion and conclusion
This is a crucial moment to study refugee integration in the United States given that refugees across the world face an uncertain future. Not only has the refugee population grown dramatically, but the countries they flee have also diversified over time. As the need for refuge has peaked globally, the annual quota for refugee admissions in the United States has plummeted to the lowest level in decades. The United States is not alone in this retreat; this trend is universal across many affluent democracies in the Global North (FitzGerald 2019; Tran 2020).
Our analyses point to three findings. First, despite significant variation in selectivity and premigration educational profile, English proficiency, and occupation, we find only modest differences across the five refugee groups. The absence of any strong associations between premigration characteristics and postmigration socioeconomic outcomes in the United States indicates an active process of human capital discounting (the nontransferability of human capital) for recently arrived refugees, at least in the short term.
Second, the two possible pathways to integration are education and work. The majority of every refugee group reported working in the week before the survey, whereas only a fraction pursued schooling. When they work, they are mostly concentrated in low-wage sectors such [End Page 141] as service and blue-collar work. Iraqis and Somalis are more likely to be in school and Cubans to work. Bhutanese and Burmese take both paths.
Third, postmigration integration policies matter. Both English-language and job-training programs are positively associated with the likelihood of attending school and of working, respectively. At the same time, English-language training program is negatively associated with the likelihood of working. Taken together, these patterns point to two distinct paths of incorporation. Refugees enrolled in English courses are more likely to end up in school; those in job-training programs are channeled into employment, given their desire to retrain themselves for the U.S. labor market.
Although federal refugee cash assistance has no impact on integration pathways, nonprofit cash assistance is negatively associated with weekly work hours. We interpret nonprofit cash assistance as an indicator of both prolonged need for support and delayed self-sufficiency. Put differently, negative selectivity is likely among refugees who have to rely on such support given that they might have exhausted the federal refugee cash assistance, while not having been able to find work. In fact, this pattern fits the qualitative descriptions of Burmese refugees in selected case studies. Overall, our findings show how selected federal, state, and local policies can shape the opportunities available for refugees on arrival and eventual incorporation.
We also find clear evidence of downward occupational mobility, especially among those with higher education and occupational status in their country of origin. The selective recognition of foreign credentials has long been identified as a crucial mechanism for variation in the labor pathways of immigrants (Bratsberg and Ragan 2002; Friedberg 2000; Sumption 2013). This process is most apparent among Iraqi refugees. Although almost 30 percent had a university degree on arrival, they were far less likely to be employed than Cubans. The lower rates of labor-market participation were not offset by schooling, because Iraqis were not more likely to be attending school than other groups, net of controls. This general pattern holds across the other groups in the sample, even among the group with lowest human capital on arrival—the Burmese. Refugees may not be premigration blank slates, but their context of reception on arrival in the United States effectively renders them so, at least in the short term. We do not know how educational selectivity might affect the long-term prospects of refugees, but our findings show that high levels of human capital on arrival do not translate into work or school success.
National origin also matters, but less consistently. Relative to Cubans, all other groups are less likely to work and, among those employed, some report working fewer hours. We think of these findings in two ways. For some groups, they could indicate discrimination in the labor market, which would align with qualitative accounts of discrimination in the literature, especially among Iraqis and Somalis (Chambers 2017; Jamil et al. 2012; Voyer 2013). Second, fewer weekly work hours can indicate the difficulty some groups have in finding full-time work without the support of an ethnic community due to linguistic barriers to mainstream employment. By contrast, Cubans have an advantage because their established coethnic communities not only offer employment opportunities, but also can insulate refugees from labor-market discrimination (Eckstein 2009; Portes and Stepick 1993). Because the 2016 ASR does not include questions on coethnic communities or discrimination, we cannot empirically evaluate these competing explanations.
Postmigration integration policy significantly helps refugees settle into American life. Those participating in some sort of training programs were far more likely to be in school or working, English-language training offering a path into school and job training a path into the labor market. Job training was also associated with higher wages for refugees. We could speculate on three possible mechanisms behind this. First, training programs might have a network effect, connecting prospective students and workers with prospective schools and employers. Second, training can either make refugees' prior credentials more visible to U.S. employers or offer refugees skillsets and credentials for the U.S. labor market. Third, refugees who want jobs seek out training programs, [End Page 142] suggesting that they are motivated to work and selected in other "unmeasurable" characteristics.
The contrast between the differential impacts that premigration human capital and postmigration training programs have on refugee integration highlights the crucial role of integration policy in shaping refugee outcomes in the short term. That we welcome refugees at all is so important, but the resources and training programs available to them on arrival matter just as much for their success. Further, we may also be undervaluing skills that refugees—especially the most educated—bring with them to the United States. This could be remedied by the creation of a mechanism for credentials from other nations to be evaluated and validated on the refugees' resettlement in the United States (Sumption 2013).
Our analysis is limited to short-term integration outcomes, but highlights the crucial role of integration policies, programs, and practices in setting refugees on a path to success in this country. Although the inclination has been to do less and less for those who have been fortunate to escape violence and persecution, the evidence indicates that we should be doing just the opposite if we want refugees to lead productive lives. New beginnings, after all, are about second chances. A helping hand can go a long way for the poor, the tired, and those yearning to breathe free. [End Page 143]
Van C. Tran is associate professor of sociology and deputy director of the Center for Urban Research at the Graduate Center, City University of New York, United States.
Francisco Lara-García is a PhD candidate in sociology at Columbia University, United States.
We are grateful to Catalina Amuedo-Dorantes, Katharine Donato, Sheldon Danziger, Heba Gowayed, Philip Kasinitz, Jennifer Lee, Suzanne Nichols, Andreas Wimmer, and participants at the Legal Landscape of U.S. Immigration in the Twenty-First Century conference at the Russell Sage Foundation for helpful feedback on an early draft. Maria Abascal, Flavien Ganter, Tiffany Huang, and Greer Mellon provided critical advice on technical aspects of the analyses. We also thank Moti Benti for his research assistance. Three anonymous reviewers' feedback strengthened our manuscript. All remaining errors are our own.
1. In this article, following current media convention, we use Burma and Myanmar interchangeably.
2. The category of East Africans comprises individuals from Burundi, Comoros, Djibouti, Madagascar, Malawi, Mauritius, Mayotte, Mozambique, Reunion, Rwanda, Seychelles, Somalia, Tanzania, Uganda, Zambia, and Zimbabwe.
3. Because the dependent variable is log hourly wage, the percent decrease in wages for the female coefficient in table 4 is equivalent to [exp (0.258) – 1]*100 = 29.4 in model 4.
4. Two additional categories, as mentioned, are students and Others. We exclude these cases from this table.
5. Results are not shown but are available on request.