The Departed

New Destinations and Out-Migration by Latino Immigrants

New immigrant destinations have been the subject of much research in recent years. Scores of books and articles have begun to paint a portrait of how these communities formed and key differences between these areas and more traditional immigrant gateways (Flippen and Parrado 2012; Massey 2008). However, the relative newness of the phenomenon has prevented the study of how these communities evolve over time, including how they respond to changing economic and policy conditions. While the forces drawing immigrants to new destinations are as diverse as the destinations themselves (Flippen and Kim 2015; Parrado and Kandel 2008), many immigrants were driven locally by an influx of businesses and native workers and the boom in residential and commercial construction that they engendered. As such, they were hit particularly hard by the 2007 economic recession, which began in the housing sector and resulted in massive losses in construction employment (Hadi 2011).

The implicit assumption in much of the work on new destinations was that these communities would continue to grow and mature into more established immigrant destinations. However, a number of features of the immigrant populations in these areas call this assumption into question. Specifically, compared to those in more traditional gateway areas, immigrants in new destinations are disproportionately more recently arrived to the United States, undocumented, and less established in families (i.e., a higher share of men are either unmarried or unaccompanied by wives, who remain in countries of origin) (Flippen and Parrado 2012). In other words, they are precisely those who are most mobile and most able to relocate in response to changing economic conditions. Whether and how new destinations endure over time thus remains an open question.

Our understanding of international migrants’ mobility patterns after initial entry is limited. While return migration to sending countries has received theoretical and empirical attention, far less is known about relocation within receiving countries. Moreover, most research has focused on the impact of return migration on sending communities, to the relative neglect of how out-migration impacts receiving communities, and the implications of exits for understanding processes of settlement and adaptation.

The departure of international migrants, especially those who return to sending countries, has also been inconsistently incorporated into theories of migration and settlement (Cassarino 2004). The dominant neoclassical economic framework regards migration as a form of investment in human capital. According to this view, individuals migrate when the expected benefits to employment and wages resulting from mobility exceed the expected costs (Sjasstad 1962). As an investment in long-term, lifetime earnings, migration is not generally expected to involve returns. In the absence of substantial reversal in the wage or the opening of employment gaps across locations, return migration from this perspective is in many ways an anomaly and most likely signals return migrants’ failure to achieve their expected labor market gains from relocation.

Alternative perspectives, in contrast, have highlighted that in many situations, circularity and return migration are very much part and parcel of the process of international migration. Piore’s seminal book Birds of Passage (1979), for instance, argued that many international migrants are target earners who engage in circular migration in response to financial needs at home. The circularity of the flow together with capital accumulation abroad were thought to foster the transfer of new skills and entrepreneurial activities to sending communities. Likewise, the new economics of labor migration framed migration as a household risk-diversification strategy to protect against uncertain and unstable economic conditions by providing alternative sources of income (Stark 1982). Under this model, migration was also argued to respond not only to absolute wage differentials but also to relative socioeconomic position and mobility prospects at home (Stark and Taylor 1989). The latter viewpoint directly connects the remittances and savings of migrants, especially return migrants, to business formation in sending areas (Massey and Parrado 1994; Papail 2002; Woodruff and Zenteno 2007).

For both perspectives, return migration is not framed as an anomalous failure to achieve expected lifetime labor market gains. On the contrary, it represents an expression of migrants’ successful capital accumulation at destination that is then translated into absolute or relative social mobility upon return. More recently, scholars have also pointed out that processes of transnationalism lead to sustained connections between migrants and their sending communities, spurring business enterprises that transcend national borders (Portes, Guarnizo, and Haller 2002). Taken together, these perspectives invite a more systematic consideration of the forces and patterns shaping return migration and the selectivity of return migrants and the patterns of settlement that they display.

A more dynamic view of return migration is particularly relevant for the case of Latin American migration to the United States. Massey and colleagues have long described Mexican migration to the United States not as a single event but, rather, as a social process that develops over the life course of individuals (Massey et al. 1987; Massey, Durand, and Malone 2002). The process is described as a sequence of stages that include initial departure, return, recurrent migration, and ultimately settlement or final return. From this view, then, return migration and circularity are defining characteristics of Latin American migration. In addition, a dynamic perspective explicitly recognizes that the fluidity of population movements is significantly affected by macrolevel conditions in the economic and political spheres. It is well documented that immigrant inflows and outflows closely mirror cycles of labor demand in the United States, with immigrants arriving when employment opportunities are plentiful and returning to their home countries during periods of economic slowdown (Angelucci 2012; Mandelman and Zlate 2012). Moreover, immigration policy can also exert a direct effect on the social process of migration, changing not only the legal standing of labor migrants but also the dynamism of the movement (Massey, Durand, and Pren 2015).

The impact of immigration policies on Latin America–U.S. migration patterns can be traced to the unintended consequences of changes to immigration legislation in the mid-1960s. Massey and Pren (2012) argue that the elimination of the Bracero program, which served as a critical mechanism for satisfying the temporary and seasonal labor demands in the United States, and the imposition of numerical caps on immigration from the Western Hemisphere in the Hart-Celler Immigration and Nationality Act of 1965, effectively transformed a long-standing, deeply institutionalized, circular and legal migration flow into an undocumented one. The practical consequences of this transformation became particularly detrimental as restrictive immigration legislation and increased border enforcement disrupted the circularity of the flow. Starting with the passage of the Immigration Reform and Control Act (IRCA) in 1986, the criminalization of unauthorized entry significantly increased the costs of circular migration for the vast majority of low-skill labor migrants. With no legal channels to reenter the United States, voluntary returns became less prominent.

The passage of the 1996 Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA), and the enhanced measures to enforce immigration restrictions it entailed, marked another turning point in Latin American migration. As a result of programs such as Secure Communities, which created partnerships among federal, state, and local law enforcement agencies to promote the interior enforcement of immigration law, the deportation of labor migrants increased considerably post-1996. The number of annual deportations grew steeply over time, from 70,000 in 1996 to more than 420,000 by 2012. More than 70 percent of deportations were Mexican nationals, with the cumulative number of deportations of Mexicans between 2000 and 2012 exceeding 2.7 million (Parrado 2012).

From the perspective of local areas, these changing economic conditions and immigration policies raise many questions related to return migration and out-mobility. There are no estimates of the actual extent of out-migration and whether the departed are returning to their countries of origin or moving to another area within the United States. In addition, the reasons for the move remain unclear, making it difficult to separate responses to changing economic and family conditions from deportations. Finally, we know very little about the socioeconomic background of those departing, limiting our understanding of the implications of out-migration for the adaptation trajectories of those settling in the United States.

Data and Methods

Data for the analysis come from the Gender, Migration, and Health among Hispanics study (for more detail, see Parrado, McQuiston, and Flippen 2005), an original ethnosurvey collected in repeated cross-sections in 2001, 2006, and 2011 in the Durham, Chapel Hill, and Carrboro metropolitan areas of North Carolina (for the sake of expediency simply referred to as “Durham,” where the majority of respondents lived). The current analysis focuses on the 2011 data, which capture conditions during and after the Great Recession, though figures from the 2006 data are also used as points of comparison. Since the relatively recent nature of the Latino community in Durham rendered simple random sampling prohibitively expensive, the project employed targeted random sampling techniques to approximate a representative sample of the Durham Latino immigrant community.

Through Community Based Participatory Research (CBPR) that involves a partnership between academics and community members at all stages of the research process, we identified forty-nine apartment complexes and blocks that housed a large number of immigrant Latinos. We then collected a census of all the apartments in these areas and randomly selected individual units to be visited by interviewers. CBPR members were trained as interviewers and collected all the data. Although our survey may have been less likely to capture established immigrants, the targeted design and community involvement method was more effective than nonrandom methods of recruitment such as snowball or convenience sampling.

In addition to information about the employment and immigration histories of respondents, the 2011 survey included a battery of questions specifically designed to capture out-migration from Durham. Out-migrants are by definition no longer residing locally, so they are impossible to observe at the time of the survey. Building on multiplicity sampling techniques (Sirken 1970), however, we queried respondents about the mobility of prior Durham residents. Multiplicity sampling is a data collection strategy in which a respondent is asked to report on the occurrence of a rare event among a set of individuals known to them (Laumann et al. 1993). A specific multiplicity rule defines the set of individuals respondents are asked to report about. The rule typically follows consanguine or spatial relationships. Respondents are also asked to report on the overall size of the individuals they know within the set defined by the multiplicity rule, which is then used to weigh the number exposed to the event. This strategy has been shown to provide an unbiased estimate of the number of persons experiencing otherwise impossible-to-measure events (Laumann et al. 1993).

In our case, respondents were probed in depth about how many brothers, sisters, and other close family members had ever lived in Durham, how many were still residing in the area, and how many had left. For each person identified as ever having lived in Durham, their current location was ascertained. For those who had left the area, the year of departure; city, state, and country of destination; and reason for departure (including deportation) were obtained. We also asked the same set of questions for all blood relatives, other than siblings, who were in Durham at some point and departed. While substantive patterns are similar for siblings and other relatives, we focus our analysis on the former mainly because the socioeconomic correlates in our analysis include household background characteristics that are strongly correlated among siblings of the same household. The correlation across households in background characteristics is likely considerably weaker.

Based on the respondents’ information, we expanded the dataset to construct a siblings file that listed all the siblings who had ever lived in Durham, whether they had left the area, year of departure, reason for the move, and destination. The survey also collected information about the socioeconomic characteristics of the respondents, including country of origin, gender, educational attainment, size of community of origin (rural area, town, or city), how well they speak English, whether they migrated to Durham directly from abroad or via another area within the United States, and current occupation. The multivariate analysis correlates respondents’ socioeconomic characteristics with the likelihood of reporting a sibling leaving the area, reason for the move, and destination. Ideally the survey would have collected detailed information on the human capital, family structure, and migration and employment history of departed siblings. However, that was not feasible within the constraints of the project and also would have been subject to significant recall and other errors.

As such, we use respondent characteristics to model the departure behaviors of their siblings. This serves two purposes. First, because some characteristics were set before migration, they are likely to be strongly correlated among siblings (particularly country of origin, rural origins, and educational attainment). Thus, they provide a proxy for the socioeconomic selectivity of the departed. Second, while we do not directly model individual behaviors, the analyses identify the socioeconomic background of the immigrants exposed to the loss of family members through out-migration, adding to our understanding of the changes in the Durham Latino immigrant community.

We collected a total of 517 surveys in 2011, with 339 and 178 adult male and female respondents, respectively. More than half of the respondents, 60 percent, were of Mexican origin; and 37 percent were from Central America, primarily Honduras (17 percent), with a small portion (2.3 percent) from other Latin American countries. In terms of their educational attainment, the majority of respondents (42 percent) had achieved 6 or fewer years of education, 28 percent reported 7 to 9 years of schooling, 21 percent completed 10 to 12 years of school, and 7 percent reported 13 or more years of education. Respondents were evenly distributed in terms of the size of their community of origin, with 28, 37, and 35 percent reporting growing up in a city, town, and rural area, respectively. Almost 11 percent of the sample reported speaking English well, and 46 percent were internal U.S. migrants prior to arriving in Durham. At the time of the survey, the vast majority of respondents were working in construction (35 percent), landscaping (7 percent), food services (14 percent), and cleaning/babysitting services (14 percent). On average, respondents reported having had 1.33 siblings in the area. Expanding the dataset to create a siblings file produced a total of 1,199 siblings.

We first describe the likelihood of siblings departing Durham over time, the reasons for the move (distinguishing among accidents or illnesses, family considerations, economic concerns, and deportation), and whether it involved return migration to the county of origin or migration within the United States. We then use logistic regression techniques to model the likelihood of a sibling leaving Durham according to respondent characteristics. Finally, we extend the multivariate analysis by modeling, conditional on reporting a sibling departing Durham, whether the move was for family or economic reasons and whether it was a return to their countries of origin or internal U.S. migration.

The Setting

Durham offers a valuable vantage point from which to explore migration dynamics. Growth of the high-tech sector during the 1990s spurred a boom in business and residential construction, heightening demand for construction and other semiskilled laborers, as well as for domestic workers and other service employees for the growing class of professionals in the area. The result of this combination of forces was dramatic. In 1990, fewer than two thousand foreign-born Latin Americans resided in the area; but by 2010, the number had reached close to forty thousand, or 12 percent of the total population.

The Great Recession, which began in December 2007, significantly altered economic conditions in the area. In North Carolina, payroll employment levels fell in every private industrial sector with the exception of education, health services, and professional and business services; and contraction in the construction and manufacturing sectors was particularly pronounced. Between December 2007 and June 2012, construction and manufacturing employment in the state declined 32.1 and 17.9 percent, respectively, and the unemployment rate increased from 5 to more than 10 percent. The unemployment rate for North Carolina Latinos in 2010 reached 12.1 percent (Gitterman, Coclanis, and Quintero 2012).

One of the most striking features of new Latino destinations such as Durham is the degree to which Latinos were concentrated in precisely those industries that were most strongly affected by the recession. As a result, Latino unemployment in Durham more than doubled, from 4 to 10 percent, between 2006 and 2010. The change in economic fortunes was clearly evident in our data as well, at least among men. In 2006, a mere 4 percent of Latino immigrant men reported being out of work at interview (Flippen 2012), compared to more than 13 percent in 2011. Among women, however, roughly 42 percent were not working at interview in both periods, though the share reporting unemployment (as opposed to more voluntary reasons for nonparticipation) rose over time (Flippen 2013). Like other new destinations, moreover, Durham tends to attract segments of the immigrant population that are the most likely to be mobile geographically: new arrivals to the United States, undocumented migrants, and unaccompanied men.

The sharp deterioration in the city’s economic conditions and the significant changes in its immigration enforcement policies make Durham an interesting vantage point to evaluate the out-migration of Latin American immigrants. In February 2008, the Durham police department initiated the 287(g) program. The 287(g) is an umbrella program in which Immigration and Customs Enforcement (ICE), a federal entity in the Department of Homeland Security (DHS), provides assistance and information to local police officers enforcing immigration laws.

In its Memorandum of Agreement with ICE, the Durham police department followed the more stringent “field model” of enforcement, which included officers who work in the field, rather than the alternative “detention model,” which limited training in the 287(g) program to officers who work in jails. The region’s entry into the Securities Communities Program further enhanced immigration enforcement in the area. Wake County, which includes the capital city of Raleigh, was the first in the state to join Secure Communities, which it did in November 2008. Orange County, which includes Carrboro and Chapel Hill, followed suit in January 2009; and Durham County did so in February 2009.

Secure Communities is a DHS program designed to identify immigrants in U.S. jails who are deportable under immigration law. Under Secure Communities, participating jails submit arrestees’ fingerprints not only to criminal databases but to immigration databases as well, allowing ICE access to this information. By 2015, the Secure Communities program in Durham and Orange counties had produced 701 removals and returns (ICE 2015).

Descriptive Results

Figure 1 reports the share of siblings reported as having departed the Durham area by year of the move. Overall, 14.4 percent of siblings who had ever lived in Durham moved out of the area between 1995 and 2011. However, only 2.8 percent did so prior to the 2007 recession, compared to 11.6 percent afterward. Moreover, the increased prevalence of out-migration was already evident in 2006, suggesting that migrants were responding to deteriorating employment conditions even before the official start of the recession. This pattern also suggests that out-migration was already increasing before the application of the 287(g) and Secure Communities programs in 2008 and 2009, respectively. By 2009, the share of siblings departing Durham had increased to five times the level reported for 2006, reaching 3.3 percent. The prevalence of sibling out-migration remained high in 2010 (2.9 percent) and 2011 (1.7 percent, which, due to the timing of data collection, reflects departures occurring during the first six months of the year).

OPEN IN VIEWER

Figure 1

Percentage of Siblings Reported as Leaving Durham by Year

The coincidence between the timing of the economic recession and the introduction of enhanced immigration enforcement policies makes it difficult to disentangle the forces causing the increase in out-migration. To explore this issue, Figure 2 presents the share of departed siblings who left Durham according to the reported reason for the move, separating departures before and after the recession. Before 2007, out-migration from Durham was dominated by personal considerations, particularly family concerns and injury from accidents, which constituted 55.1 and 12.2 percent of all departures, respectively. At 28.5 percent of all departures, lack of employment was another important impetus behind out-migration from Durham even prior to 2007. However, there were relatively few cases of deportations, which represented only 4 percent of all departures before 2007.

OPEN IN VIEWER

Figure 2

Reported Reasons for Migration Pre- and Post-2007

After 2007, however, we see an important shift in the motivation behind departures from Durham. While family concerns remain an important reason for out-migration (26.3 percent), both economic considerations and, to a lesser extent, deportations gain considerably in significance. It is particularly striking that the share of out-migrants perceived as having left due to lack of employment more than doubled after the recession, reaching 63.4 percent of all departing siblings. Likewise, the share leaving Durham due to deportations also increased to 9 percent of departures. Even with this increase, however, the likelihood of out-migration due to deportation remained only 0.15 and 0.37 times as likely as departing due to economic or family reasons, respectively.

Figure 3 presents results from the queries regarding the destination of departing siblings. Results show that respondents overwhelmingly reported that their sibling returned to their countries of origin rather than relocating within the United States. The pattern is very similar before and after the recession with 77.5 and 78.4 percent of siblings voluntarily returning to their home countries before and after 2007, respectively. If anything, the prevalence of internal migration actually declined after the recession from 18.4 to 11.9 percent. The combination of dim economic prospects in other states and rising salience of deportations likely combined to heighten return migration over time.

OPEN IN VIEWER

Figure 3

Destination Pre- and Post-2007

Multivariate Results

While the descriptive statistics document important shifts in the magnitude, motivation, and destination of out-migration over time, it is also important to explore potential selectivity in the flow and its implications for those left behind. Table 1 shows results from logistic regression equations estimated to predict sibling out-migration patterns from respondents’ socioeconomic characteristics. To the extent that the socioeconomic characteristics of siblings are correlated, multivariate results can help us to assess the socioeconomic selectivity of out-migration flows, as many sibling traits are household characteristics acquired before migration.

OPEN IN VIEWER

Table 1

Coefficients from Multivariate Logit Models Predicting the Likelihood of Sibling Out-Migration from Durham and Reason for and Destination of the Move

	Out-Migration		Economic vs. Family Reason		Return vs. Internal U.S. Move
Intercept	−1.586**	(0.269)	−0.207	(0.496)	3.182**	(0.887)
Country of origin (Ref: Mexico)
Central America	−0.271*	(0.186)	0.853**	(0.362)	−0.427	(0.465)
Other nationality	0.339	(0.792)	0.175	(1.745)	−0.416	(1.738)
Survey respondent characteristics
Male	−0.081	(0.258)	−0.830	(0.620)	0.203	(0.793)
Educational attainment (Ref: 0–6 years of education)
7–9 years	−0.028	(0.192)	0.446	(0.377)	−0.201	(0.499)
10–12 years	−0.499*	(0.273)	1.183**	(0.581)	0.654	(0.810)
13+ years	−1.025**	(0.502)			−0.622	(1.130)
Region of origin (Ref: City)
Pueblo	0.006	(0.203)	−0.282	(0.391)	−0.469	(0.562)
Rancho	−0.370**	(0.226)	0.227	(0.446)	−0.772	(0.580)
U.S. characteristics
Speaks good English	0.196	(0.308)	−0.216	(0.636)	−0.107	(0.786)
Internal migrant	−0.047	(0.175)	−0.190	(0.352)	−0.771**	(0.443)
Occupation (Ref: Not working)
Construction	0.535**	(0.279)	0.562	(0.616)	−0.846	(0.698)
Landscaping	0.081	(0.384)	0.778	(0.788)
Restaurant	0.045	(0.313)	−0.039	(0.688)	−1.416	(0.811)
Cleaning/babysitting	−0.095	(0.284)	0.034	(0.600)	−1.233	(0.850)
N	1,199		173		173
Likelihood ratio	24.7**		15.7		13.4

NOTE: Standard errors in parenthesis.

*

p < .10. **p < .05.

The first model reports the coefficients predicting the likelihood of out-migration. Results show that relative to Mexicans, Central American respondents are only 0.762 (exp[−0.271]) as likely to report having had a sibling depart the Durham area. The result is likely a function of both the fact that Central Americans benefited from specific immigration policies, such as Temporary Protected Status (TPS) that afforded them legal status and that the greater distance and difficulty involved in Central American migration discourages exit and circularity relative to Mexicans.

The coefficients for educational attainment show that the likelihood of reporting sibling departure from the Durham area decreases consistently with higher levels of education. Residents with 10 to 12 and those with 13 or more years of education are 0.607 (exp[−0.499]) and 0.359 (exp[−1.025]) as likely, respectively, to report a sibling departing than residents with 6 or fewer years of education. To the extent that siblings’ level of education is correlated, results suggest negative selectivity in the out-migration flow. After controlling for other socioeconomic characteristics, those who grew up in rural areas are also less likely (−0.370) to report a sibling departure than those raised in cities. Again, assuming that sibling places of origin are highly correlated and that return migrants often settle in their communities of origin, the limited economic opportunities in rural areas in Latin America will discourage out-migration from Durham, once other characteristics are taken into account.

Interestingly, results show that respondents employed in construction are 1.708 (exp[0.535]) times more likely than nonworking residents to report a sibling departing Durham. Again, the construction industry was not only the main employer of Latino immigrants in the Durham area but also the sector most adversely affected by the recession. Given the importance of social networks to the process of securing a job among immigrants, the likely correlation in sibling industry renders this pattern at least suggestive of greater out-migration among workers in the construction industry. At a minimum, it demonstrates that working in construction is associated with greater exposure to experiencing the out-migration of siblings, a key source of social support in migrant communities.

The next two models report the coefficients from logistic regression models predicting the likelihood of out-migration due to economic circumstances versus family reasons, and whether the move entailed a return migration to the country of origin versus an internal move to another U.S. destination. The models are conditional on a respondent having reported a sibling departure from Durham. The restricted sample size (N = 173) reduces the statistical significance of many of the estimates. However, some patterns still emerge. Even though Central American siblings were less likely to depart the Durham area than Mexicans, they are more likely to do so due to economic rather than family reasons. Again, we attribute this pattern to distance and cost. Returning to home communities to visit or care for family members is less costly for Mexican migrants than it is for Central American migrants, who tend to return only when economic conditions are seriously untenable. We also find that even though higher educational levels are negatively associated with sibling out-migration, conditional on sibling departure, better-educated residents are more likely to report economic than family motivations for exit. Again, we speculate that more educated respondents generally come from families with more advantaged socioeconomic backgrounds, who are likely to be less dependent on others for their support. In these families, it is only when the economic position seriously deteriorates that members return to sending communities.

The final model reports coefficients from logistic regression models predicting the likelihood of returning to sending countries versus moving to another area in the United States. We find that those residents with prior migration experience in the United States are more likely to report a sibling moving internally rather than returning to their home countries. If internal migration reflects family connections in other areas of the United States, it appears that those extended connections reduce return migration to sending communities.

Conclusion

This article investigates the prevalence and motivation of out-migration, and destination of out-migrants among Latin American immigrants in the Durham-Chapel Hill-Carrboro metropolitan area of North Carolina, a new Latino immigrant destination. Most studies of immigrant out-migration tend to focus on return migration to sending countries and are primarily conducted in receiving areas. Relying on an original survey and multiplicity sampling techniques, our study constructed a dataset of respondents’ siblings ever residing in Durham and assessed their out-migration propensities; whether they left the area due to accidents, family considerations, economic conditions, or deportations; and whether they moved internally within the United States or returned to their home countries. We then modeled these outcomes according to respondents’ socioeconomic characteristics to assess the socioeconomic background of those most affected by siblings’ out-migration and to tentatively explore the selectivity of immigrants departing the area.

Durham provides a unique setting to investigate these issues. As an emerging area of destination, it experienced exponential growth in its Latino immigrant population in conjunction with the expansion of the construction and low-skilled service sector industries. Employment in the construction industry was particularly hard hit by the 2007 economic recession. The community, like many others in the nation, also participated in enhanced interior immigration enforcement policies. Specifically, Durham and Orange counties, which coincide with the area under consideration, activated the 287(g) and Secure Communities programs in 2008 and 2009, resulting in more than seven hundred deportations by early 2015 (ICE 2015). Thus, the area provides an ideal setting to investigate the relative role of voluntary versus involuntary departures among Latino immigrant residents.

The analysis documents a dramatic increase in departures from Durham coinciding with the 2007 recession. As many as 14.4 percent of siblings who had ever lived in the Durham area had departed by 2011, and 80.3 percent of those departing did so after 2007. The prevalence of departures was already high in 2007 and 2008, before the activation of programs incorporating local police departments into immigration enforcement policies. We find that before the recession, family considerations and accidents accounted for 77.3 percent of all sibling departures. After 2007, economic considerations increased to represent 63.4 of the reasons given for leaving the area. Deportations rose from 4.1 and 9.7 percent of the reasons for departures before and after 2007, respectively. Overwhelmingly, those departing returned to their home countries (close to 80 percent), with little change before and after the recession.

Central Americans were less exposed to sibling departures than Mexicans, though out-migration was more often attributed to economic rather than family considerations for the former. The greater difficultly involved in reaching the United States from Central America likely contributes to both of these patterns. An examination of the human capital and migration histories of respondents with out-migrating siblings also hints at the selectivity of the departing flow. Respondents who herald from rural areas are less likely to report out-migrant siblings, as are those with higher levels of education. Those employed in construction, in contrast, are more likely to report out-migrant siblings. There is some evidence that prior experience with internal U.S. migration is associated with movements within the United States rather than the return to countries of origin.

Taken together, these patterns are supportive of the argument that immigrants’ characteristics interact with conditions in both sending and receiving areas to shape patterns of return migration. Those whose easiest option is to return to underperforming rural areas are less likely to leave Durham than those who grew up in cities. Those with higher levels of education are better able to weather economic downturns in the United States and are particularly unlikely to leave for family-related reasons. Finally, leaving Durham for another U.S. location is more likely among those whose family networks include internal migrants, as opposed to those more likely to have migrated directly to the Durham area from abroad.

The analysis has important implications for the literature on immigrant adaptation and incorporation. Return migration and other forms of out-migration are often neglected in accounts of the socioeconomic mobility of immigrant origin populations, including both contemporary flows from Latin America and Asia and also historical flows from Europe. The issue has received some attention in the health literature where the Latino advantage in some health and mortality indicators has been attributed to the negative selectivity of those departing the United States (Palloni and Arias 2004). Our analysis highlights that return migration and other forms of out-migration are an important factor affecting the settlement and development of Latino immigrant communities in the United States. The concentration of the most mobile immigrants, particularly the newly arrived, undocumented, and unaccompanied men, in new destinations could make their immigrant populations especially vulnerable to sizable shifts in composition in response to changing conditions.

As such, research that seeks to examine the process of immigrant incorporation over time or across generations must take into consideration the complex dynamics of in- and out-migration. Not only do processes of selection shape both types of movements, but these processes also likely differ for in- and out-migration, as well as for different national origin groups. Our results show, for example, that the considerable out-migration in response to the recession is an indicator that efforts to assess the impact of macroeconomic change in the United States that rely on unemployment and nonparticipation figures understate the true magnitude of job loss for Latino immigrants.

Our analysis also adds to our understanding of contemporary migration from Latin America. While the militarization of the border and enhanced enforcement of immigration restrictions throughout the United States have discouraged circular migration and tipped the scales in favor of settlement, the flow is far from unidirectional, particularly for Mexicans. Even though departures increased dramatically after 2007, they were also an important feature of the Durham Latino immigrant community even during a period of booming employment in immigrant-intensive industries. The large share of migrants who return to their home countries for family and other noneconomic reasons, in good times and bad, strongly suggests that were immigration policies to shift away from blocking entry toward the approach taken in earlier generations, a sizable segment of the Latino migrant population would return to the more fluid and circular pattern that predominated in the pre-IRCA era. The fact that considerable return migration is evident even from a metropolitan area were immigrants are concentrated in relatively well-paid fields outside of agriculture indicates that temporary legal worker programs could be efficacious in the contemporary service-driven economy.

Another important implication of the analysis is that despite the massive investment in financial and human resources to enforce immigration control in local areas, and the tremendous hardship that these policies impose on immigrant communities and families (Brabeck, Lykes, and Hunter 2014), deportations had a surprisingly small impact on return migration. The vast majority of Latino immigrants leaving Durham returned to their home countries on their own. While the majority of returns were in response to deteriorating economic conditions in Durham, it is noteworthy that more migrants voluntarily returned to attend to family matters than were deported. To the extent that the objective of immigration policy is to regulate the labor market so that it is in line with economic opportunities, relying on the natural forces of circularity and adaptation to changing economic conditions appears as a much more effective mechanism to maintain well-functioning local labor markets than forced removal. For deportations to reach the same level as voluntary returns, they would have to increase more than tenfold. Our findings indicate that given the tremendous costs associated with interior immigration enforcement, relying on deportations to regulate the size of immigrant communities is an untenable, ineffective, and unnecessary strategy.