The transnational dimensions of identity formation: Adult children of immigrants in Miami

Abstract This article draws on the insights of research on transnational migration to reconsider the process of identity formation among children of immigrants and the patterns of acculturation associated with different trajectories of segmented assimilation. Extending the research on segmented assimilation and identity formation among children of immigrants in early adulthood, the article uses the third wave of the Miami CILS sample to examine the relationship between segmented assimilation outcomes, nationality, and behaviours and attitudes associated with transnationalism. Key findings include evidence associating selective acculturation with greater transnational involvement, but also some limited evidence of downward assimilation associated with higher rates of sending remittances among some nationalities. Rather than contradicting each other, these findings suggest a diversity of transnational patterns that are both stratified by class and contoured by nationality and ethnicity. We conclude by outlining some of the questionnaire items and aspects of research design we would like to see implemented in future studies of transnationalism and its relationship with segmented assimilation.


Introduction
Our globally networked society enables individuals to recombine the points of reference and categories from which they constitute identity. Engagement in transnational, diasporic, and even virtual communities of membership based on shared religious faith, a common sense of national, ethnic or racialized identity, shared global issues such as Ethnic and Racial Studies Vol. 28 No. 6 November 2005pp. 1182Á/1214 human rights or the environment, etc., is now a reasonable possibility for a rapidly growing segment of humanity. Time-space compression, enabled by widening access to transportation and digital communications technologies, has transformed the relationship between space and place so that travel and mobility are no longer prerequisites for engaging with, being influenced by, and affecting the world views and opinions of people in geographically distant locations. Novel possibilities for global, multi-local, and transnational modes of membership arise, even as they disrupt the assumed convergence between geographically bounded locales and parochial constructions of belonging. Thus, globalization and transnationalism make obsolete conventional understandings of identity formation and its processes.
The transnational perspective on migration explores the relationship between this transformation among spaces, places, and identities to reconsider the immigrant experience. For today's migrants, social exclusion and limited economic opportunities in host societies often dovetail with new possibilities for sustaining meaningful relationships with people and institutions in places of origin (Basch et al. 1994;Vertovec 2000). Extensive research shows that migrant families orient significant aspects of their lives around their country of origin. They keep in touch and provide personal support across borders, travel as tourists, send or receive remittances, and discuss politics. Immigrants also engage in transnational collective action, forming and transforming religious, civic, and political institutions to facilitate participation in their homelands' diasporas, and to parley home and host country social issues into transnational platforms of concerns (Rivera-Salgado 2000;Wayland 2001;Guarnizo et al. 2003;Popkin 2003). In addition, transnational institutions may offer avenues for civic and political participation and entrepreneurial pursuits otherwise blocked to today's immigrant newcomers (Goldring 2001;Landolt 2001;Portes et al. 2002).
In a groundbreaking publication, The Changing Face of Home, Levitt and Waters (2002) systematically applied the insights of transnational migration to the study of identity formation and life trajectories of the children of immigrants. Levitt and Waters (2002) invited contributors to examine a range of issues, two of which are of particular relevance here. First, what kinds of attachments do children of immigrants maintain with their parents' homeland? Second, what impact might the transnational attachments of immigrant parents and children have for identity formation and, following segmented assimilation theory, for the resources young adults can rally to confront segmented labour markets, racism, inner-city marginalization, and other obstacles to prosperity and success in the U.S. and other host countries? (Levitt and Waters 2002).
The first question leads to analysis of variables such as parents' socialization of children and children's self-reported ethnic or national identity on transnational outcomes, such as children's symbolic and material attachments to their parents' homeland. The second question posits a transnational family life, reflected in resource circulation and practices of socialization as an intervening variable that may shape cross-cultural notions of belonging among the children of immigrants.
The issues and debates raised in Levitt and Waters (2002) serve as a point of reference for our analysis of the third wave of data on the adult children of immigrants from the Miami sample of the Children of Immigrants Longitudinal Study. The article is organized into four sections. The first sets the context, describing the greater Miami region as a rapidly emerging node in the global system of cities linking the core economy of North America with the primate cities of Central and South America and the Caribbean, replete with its own heightening levels of internal inequality. Second, the literature review connects the scholarship on transnational migration with immigrant identity formation in the second generation as developed in segmented assimilation theory. Given the absence of extensive information in the CILS data on the institutional contexts in which the children of immigrants are embedded, the third section focuses on descriptive findings from the CILS data and analysis of the determinants of transnational practices of adult children of immigrants in Miami. The fourth section discusses the findings and concludes with a few considerations for future research.
The setting: Miami as pan-hemispheric keystone Miami provides a unique admixture among nationalities and cultures in a single urban setting that captures the interests of social scientists, politicians, entrepreneurs and major corporations as keenly as it does those of its newest arrivals. Three decades of substantial international migration, coupled with the internal migration of whites from the Northeast and Midwest states and blacks entering the city from other parts of the South, have provided the force behind the city's economic, political and socio-cultural growth and transformation. In the span of one generation, the demographic composition of Miami grew to be intensively multicultural, with an international economy that operates seamlessly between English and Spanish, a vibrant ethnic economy dominated by a powerful Cuban elite, and a highly contested political system.
As indicated by Portes (1992), the Cuban Revolution undoubtedly precipitated this sea change in what had until then been a secondary city in the U.S. urban system. In the years following the Cuban Revolution of 1959, the displaced Cuban bourgeoisie resettled en masse in Miami or eventually found their way to South Florida. Once they had regained their economic footing in Miami, the city became the obvious choice for other Caribbean and South American inflows. As Miami grew in stature and significance, largely through the efforts and investments of the exile Cuban business class joined with the labour of less-privileged immigrant groups, it became a primary destination for Colombians, Nicaraguans, and Haitians seeking to establish themselves in the United States.
As Cuba's entrepreneurial elite settled in Miami, they transformed its economy and polity. The first wave of Cuban exiles brought with them if not the capital, certainly the professional skills and social capital to access investment capital for business ventures. The next major migration from Cuba came with the Mariel boatlift. These immigrants did not have the resources and human capital that were more common among the older arrivals. Nevertheless, with the coethnic community and its economic structure in place, the new arrivals provided their established predecessors with privileged access to a source of low-wage labour and new consumer markets. The Cuban ethnic economy in Miami represents the enclave in its most advanced form. It has emerged in the wake of large waves of migration. Other Caribbean and Latin American immigrants have arrived in Miami in less dramatic fashion, coming in relatively steady flows that have expanded with the maturation of the migration stream in its life cycle. Distinct ethnic economies are emerging from these flows but have not reached the critical mass of the Cuban enclave, with its high levels of capitalization and deep division of labour.
In an era where the internationalization of economic activity thrusts the provenance of global cities to ever-increasing heights, Miami has carved out a specialized niche. Sassen (1994, p. 88) refers to Miami as 'a center for long-distance coordination and management of transactions in Latin America and the Caribbean for firms from any part of the world interested in doing business in these regions'. For Miami's international producer services sector, 'the Cuban enclave represents a significant set of resources, from international servicing know-how to Spanish-speaking personnel. But the particular forms of economic globalization . . . have implanted a growth dynamic in Miami that is distinct from the enclave, although benefiting from it' (Sassen 1994, p. 89). Miami has become an ideal banking centre for Latin America's wealthy elites precisely because the city provides the stability of North American institutions coupled with transactions in their own language (Portes 1992).
Politically, the conservative politics of the Cuban exile community have transformed both local Miami politics and the city's location vis-à-vis Latin American and Caribbean political processes. On the one hand, the city has become a safe haven for Latin America's conservative political elites and a staging ground for their reactionary political projects. Candidates for the Colombian presidency campaign in Miami, and it is said that the Nicaraguan Contra war was often 'planned over tres leches in Miami restaurants' (Grenier and Stepick 1992). On the other hand, Miami Cubans have emerged as an indomitable force in municipal politics and a power in state and some would argue federal elections.
Miami has become an ethnic kaleidoscope where multiple communities vie for voice and interact with each other in complex ways (Portes 1992). Miami is too large not to contain many spatially segregated areas but it is also too small to allow autonomy or deep isolation among its various groups. The resulting situation is less one of interdependence as one of interethnic cohabitation in a confined space. Although racial and ethnic intolerance is overtly condemned by the vast majority of Miamians, discrimination remains systemic, as elsewhere in the U.S. In effect, Miami runs the full gambit of inequality, diversity, residential segregation, high culture, popular youth culture, dilapidated neighbourhoods, corporate dominance, fine estates, wheelers, dealers, hucksters and hustlers as in other major U.S. metropolitan areas. These urban features, however, are all shaped and influenced by its specialized niche as the emerging keystone of the north-south axis of the Americas, producing a local environment with interesting implications for the identity formation of young adults and the possibilities for their transnational behaviours and attitudes.

Bridging scholarship on identity formation: Transnationalism & segmented assimilation
A decade of research on transnational migration has generated rich discussion and debate on migrants' border-crossing relationships . Recent efforts focus on articulating a more coherent set of predictive arguments about the causes and consequences of migration, codification of transnational practices by different types of individual and institutional actors, and consideration of the relationship between transnational practices and immigrant incorporation in the host society (Levitt 2001;Levitt and Glick-Schiller 2003;Portes 2003;Waldinger and Fitzgerald 2004).
The transnational perspective on migration argues that the proliferation of practices connecting immigrants to distant homelands follows the emergence of social fields */ sets of multiple interlocking networks of social relationships through which ideas, practices, and resources are exchanged, organized, and transformed (Basch et al. 1994;Levitt and Glick-Schiller 2003). Transnational social fields offer possibilities for engaging in ways of being, or actual social relations, and ways of belonging, understood as identification with a particular group or collective in forms suggesting construction of membership extending beyond a single location (Levitt and Glick-Schiller 2003). Transnational practices are also grounded, that is they are bound by the opportunities and constraints found in particular localities (Guarnizo and Smith 1998). In this light, transnational and assimilative trajectories of incorporation are not mutually exclusive and their character and overlap is determined at least partially by the local settings and institutional contexts in which immigrants carry out their lives (Levitt and Glick-Schiller 2003). The urban landscape */ elements such as a city's immigration tradition, its political culture and traditions of civic participation, labour market structures and the ways in which these are written into the built environment */ contour the process of immigrant incorporation, settlement, and identity formation, as well as the character of an immigrant group's transnational practices (Bobo et al. 2000;Halle 2003).
Institutions */ settlement service agencies, community organizations, home country consulates and embassies, faith-based associations and congregations, professional and business associations, etc. */ by virtue of their own material and symbolic arenas and points of reference, may encourage the maintenance or revival of ties to the home country or, conversely, may lead immigrants towards greater involvement in the host society (Menjivar 1999;Itzigsohn 2001;Gold 2002;Landolt 2004).
Cultural identities are constructed through a dialectic of selfattribution (i.e., internal self-categorization) and socially defined or ascriptive conditions specific to the social worlds in which they are embedded (Vertovec 2001). The immigrant second generation may feel their strongest attachments for the ethnic, pan-ethnic, national, racial(ized), or religious dimensions of their social location, for these socially constructed categories are rarely mutually exclusive and, typically, meld into each other. Yet boundary moves across identity categories and the recombination of ways of belonging over the life course are rarely associated with personal choice. In European-derived post-colonial societies, ethnic options and symbolic ethnicity are the privileged domain of white immigrants and their descendants (Gans 1979;Waters 1990) as racialization severely restricts the process of identity formation for people of colour (Kasinitz 1992;Waters 1999).
Rather than focus on individual-level dynamics, the segmented assimilation perspective poses that identity formation for today's children of immigrants results from the convergence of family resources with meso-level contextual variables such as the quality of schooling and youth cultures within the school; the economic, political, and social connectedness or isolation of neighbourhoods; and the social networks and social capital of the immigrant community. It is therefore reasonable to suggest that both, analyses of transnational migration and the theory of segmented assimilation focus on the social contexts and institutional barriers and opportunities faced by individuals, families, and immigrant communities to predict outcomes. This convergence around meso-level factors offers a fruitful point of departure for bridging the two literatures along three lines: the family, race and religion.
Looking first at the family as a social institution, the segmented assimilation perspective emphasizes language retention and parental socialization in the development of ethnic identity among children of immigrants. We find increasing evidence for the formation of transnational families that, via time-space compressing technologies, organize socialization, reproduction, and production around multilocal border-crossing arrangements (Plaza 1999;Salazar Parreñ as 2001;Waters 2002). Researchers have documented the importance of circulating children and the elderly between places of origin and settlement to reduce the costs of social reproduction, promote learning of the mother tongue and, as is often cited by parents, to remove children from what is perceived as a negative and undisciplined social environment in the United States (Orellana et al. 2001;Menjivar 2003).
Second, the segmented assimilation perspective emphasizes the impact of racial discrimination in conferring negative status and leading to daily and lifelong experiences of exclusion. Racism and racial identities thus play a critical role in fragmenting assimilation into divergent trajectories. The literature on racialization and racial formation offers compelling evidence for conceptualizing and operationalizing the construction of racial identities as a global and transnational process (Gilroy 1993;Winant 2001). The racial categories used in the U.S. mix and compete with the racial formations into which immigrants are socialized in their homeland, and into which immigrant parents unwittingly socialize their children at home. In this light, the ability of immigrants of African heritage, for instance, to unsettle U.S. racial formations through diasporic and transnational ways of belonging must be recognized as a powerful resource for children of immigrants as they experiment with identity construction.
Third, just as ties to the parents' country of origin add an alternative reference group to the identities and inner lives of some immigrants so too does religion which, like race and ethnicity, plays a critical role in the construction of identities, perception of meaning, and formations of values . Furthermore, religious practices and notions of belonging are simultaneously vested in both global dynamics and local contexts (van der Veer 2001). As key organizations providing social context, religious institutions may encourage the orientation of religious practice towards transnational concerns or may, alternatively, push believers towards more parochial loyalties and identities (Mullins 1987;Menjivar 1999). The religious practices of children of immigrants may be affected by the influence of religious leaders and their sensitivity (or lack thereof) toward the second generation and their unique situations, secular education, and the growth of 'vernacular' religious traditions (Vertovec 2001).
Our effort to bridge these two research bodies opens up rich possibilities for cross-fertilization and also highlights ambiguities and weaknesses in each approach. Focusing on the transnational migration literature, we find that efforts to capture transnational ways of being can be quite straightforward. It simply involves identifying the material border-crossing practices of the individual and/or capturing the patterns of resource circulation of a set of relevant institutions in which the individual is embedded. This approach was applied by Portes and his collaborators in the Comparative Immigrant Entrepreneurship Project addressing the causes and consequences of transnational migration among Latin American immigrants in the U.S. (Portes, et al. 2002;Guarnizo et al. 2003), and is followed in the CILS.
There is, however, more ambiguity in trying to capture the global and transnational dimensions of ways of belonging. It is, for instance, difficult to discern when ethnic or national identities are also transnational, or when diasporic ties converge with or become panethnic identities. Nonetheless, extending our inquiry to address the border-crossing dimensions of ways of being and belonging may enrich our understanding of identity formation processes among the second generation.
From segmented assimilation theory, the direction of the relationship between the level and character of segmented assimilation outcomes among the second generation and extent of transnational ties or involvements remains unclear. Segmented assimilation among the second generation predicts differential outcomes of selective, consonant, and dissonant acculturation depending largely on family background variables including socio-economic status [SES], strategies of incorporation and adaptation to the host country, and differences in parenting behaviours. Selective acculturation, which is normally associated with higher family SES, also implies greater interest in the sending society and its culture, traditions and heritage. Consonant acculturation is where children of immigrants and their parents simultaneously and actively pursue accelerated assimilation into the mainstream of the host society, while dissonant acculturation represents a generation gap where children of immigrants may disdain their parents' clinging to the ways of the 'old country' while immersing themselves in the youth cultures (and countercultures) of the new.
Consonant and dissonant acculturation do not predict high levels of interest in or involvement with the sending society, whereas selective acculturation does. However, research on transnational networks of youth gangs suggests that even dissonant modes of acculturation can be transnationalized. Efforts by parents to send children 'home' to be disciplined may backfire, as youths act out U.S. urban ghetto scripts in the parents' place of origin (Matthei and Smith 1998;Smith 2002). Furthermore, past ethnographic research has associated transnational practices with the need to find alternative mobility paths for immigrants facing blocked opportunities in the host society (Basch et al. 1994). There are, therefore, different and opposite predictions about which segment of the differentially assimilated second generation should exhibit the highest rates of transnational behaviours and involvements, and about how transnational ways of being and belonging interact with the lives of children of immigrants.
This literature review raises a set of issues extending well beyond the scope of what we can reasonably address using the CILS data. These data offer a limited number of individual-level indicators of transnationalism: counts of trips back to the immigrant parent's country of origin, frequency of sending remittances, and whether or not one feels 'equally at home' in both countries. These are based on the transnationalism index developed by Kasinitz and his collaborators in their study of the second generation in New York (Kasinitz et al. 2002). It does not provide any information on the incorporationist or transnational trajectories of the institutions of civil society in which these young adults participate, parents' transnational practices, and limited information on the transnational dimensions of parental socialization, thus restricting the scope of our analysis. However, the three outcomes associated with transnational behaviours and practices in combination with the battery of other questions addressing segmented assimilation provide sufficient data for exploring, and testing key hypotheses concerning the relationship between segmented assimilation and transnationalism among the second generation entering early adulthood.

a. Children of Immigrants Longitudinal Study */ background and description
The Children of Immigrants Longitudinal Study [CILS] is a comprehensive survey of the second generation that commenced in 1992 in the metropolitan areas of Miami/Fort Lauderdale and San Diego. These sites were targeted because of their relatively long duration as ports of entry for the current wave of U.S. immigration and because of the diversity of national origins among these regions' immigrants, with South Florida mainly attracting migrants from the Caribbean Basin and Southern California attracting migrants from Mexico and the Asian Pacific Rim. The analysis in this work is limited to the Miami sample.
The first round of CILS included 2,842 students in South Florida in the 8 th and 9 th grades who had at least one foreign-born parent and who were either born in the U.S. or had lived in the country for at least five years (Portes and Rumbaut 2001). The first follow-up survey was carried out in 1995 and 1996 when these students were mainly in their last year of high school and produced interviews with 2,225 of the original respondents. The second follow-up was conducted mainly during the years of 2002 and 2003 and produced surveys from 1,841 of the original South Florida respondents. These large-scale survey efforts provided a random sample whose results can be generalized to the wider populations of their respective nationalities and age groups, with the caveat that there was no 'prior knowledge of the second generation's distribution' to serve as a basis for planning the sampling procedures to account for the exact probabilities of inclusion. 1 For a study of this kind, there is risk of attrition bias in the followup surveys resulting from the failure to locate and complete interviews with large numbers of respondents, who might have shared similar characteristics or circumstances that interfered with their inclusion in the follow-ups. To examine the possibility of attrition bias between the initial survey and the first follow-up, an exercise was conducted to compare 'retrieved and lost respondents on their characteristics measured in the first survey' finding that 'the follow-up sample faithfully reproduces the different categories of respondents in the original survey' (Portes and Rumbaut 2001).
The possibility of attrition bias was somewhat larger for the second follow-up. In this instance a Heckman correction coefficient (explained in Portes, Fernández-Kelly and Haller, this Issue) was calculated to correct for the primary sources of sample attrition bias. The resulting coefficient, l, was included and found to be statistically insignificant in all of the models presented below. Considering that there was no prior knowledge of the second generation's distribution within these areas, CILS represents the closest attainable approximation to a random probability sample that can be generalized to the populations that it targeted.

b. Univariate results
The Miami CILS sample shows significant disparities in the contexts of exit and reception across immigrant nationalities, variations by the immigration wave within national origin groups, and segmentation of assimilation outcomes. It therefore makes sense to examine the relationship between segmented assimilation and transnationalism by nationality. In the case of the Cubans, since U.S. policy and Miami politics limit contact with family and institutions in Cuba, the Cuban part of the sample is organized by wave of immigration represented by the proxy variable of private versus public school enrolment. The rationale for this partition is explained in Portes, Fernández-Kelly, and Haller (this Issue).
The educational attainments, family, and work situations of the primary immigrant groups in South Florida are discussed at length in the same article. Results show that children of middle-class Cuban exiles who attended private preparatory schools gained the greatest advantages via intergenerational status transmission: zero per cent high school drop outs, higher rates of college completion, fewer instances of early marriage or single parenthood, higher rates of selfemployment, and avoidance of early labour force participation in favour of continued attachment to higher educational attainment and professional goals. At the other end of the continuum, the Haitians and West Indians are disadvantaged, showing the highest rates of early parenthood and lowest rates of full-time employment. Nevertheless, the higher educational attainments among significant proportions of the Haitian and West Indian second generations show an impressive struggle against the uneven playing field of their parents' underprivileged contexts of reception.
The patterns of transnational involvement also vary widely by group, as seen in Table 1. As expected, Cubans (and especially the children of earlier exile waves) are the group least involved with their parents' country of origin. The frigid relationship between Washington/Miami and Havana distorts the relationship that might otherwise depend more naturally on the array of assimilation outcomes among second-generation Cubans. The second-generation immigrant nationalities with the greatest transnational involvement are Colombians and West Indians. This is reflected both in the high values for mean number of visits to parents' home country and in the large standard deviations surrounding these means. The Colombian and West Indian groups contain more respondents reporting greater numbers of trips back than the other major nationalities in the sample.

c. Multivariate models
To relate segmented assimilation outcomes to behaviours and attitudes that reflect transnationalism, we estimated statistical models of the number of trips back to the parents' country of origin and the frequency of sending remittances. Because dependent variables measuring counts generally follow a Poisson distribution, we have selected our estimation methods from the family of models stemming from Poisson regression. But it is, nevertheless, rare for sociological phenomena measured by number of occurrences to follow a Poisson distribution. Thus, we use negative binomial regression [NBR] which is based on a less restrictive set of assumptions and allows for overdispersion in the data. 2 Likewise, robust standard errors are used because they, too, rely on more modest assumptions */ particularly concerning whether the model estimated is 'the true' model (StataCorp 2003). Because there are not yet many examples of such multivariate models in the transnational field, this line of research is, as of yet, too undeveloped to lay claim on one or another model being 'the true' model for all relevant parameters in the population. The percent change column indicates the percent increase (or decrease) in the likelihood of participating in the behaviour represented by the dependent variable associated with each predictor. Frequency of sending remittances is in principle a count variable. However, unlike frequency of trips back to the parents' home country, the respondents in this instance were not asked to state an exact number but instead were asked to select one response from a set of grouped ranges, as shown previously in Table 1. Thus the dependent variable for frequency of sending remittances is ordinal, making ordered logistic regression more appropriate than ordinary least squares or NBR.
The independent variables we included cover basic demographic traits (sex, age, and whether the respondent has any child of his or her own); segmented assimilation outcomes (years of education, family income, unemployment, and having been sentenced for a crime), foreign language use, familiarity with home country, and religiousness. Several of these are drawn from data collected in early and late adolescence (during CILS I and II). Descriptions of all variables used in the following analyses are presented in Appendix A.

Second-generation trips to parents' countries
In keeping with the differences by nationality in the transnational behaviours reported in Table 1, we observe that differences between nationalities are particularly important. Among other predictors for number of trips back to parents' home country, some are significant regardless of whether we control for nationality. These are sex, living in an intact family in 1992, frequency of attending religious services, and whether the respondent was sentenced for a crime. In this population, being a parent is significant and negatively related to number of trips abroad, but only without controlling for nationality. Finally, some predictors are significant only when controlling for nationality (U.S. nativity, preference for a foreign language with friends, and knowledge of parents' home country in 1992).
Dummy variables for Nicaraguan, Colombian, Haitian, West Indian/Jamaican, and the residual category of 'other nationalities' are all significantly different from the reference group (public school Cubans), while private school Cubans are not. This finding reflects the United States' travel restrictions regarding Cuba, as it prevents similar rates of visits to the parents' home country among the Cuban second generation as those which occur among all other groups. It also shows that some trips 'home' among a certain proportion of the second generation are the norm under favourable conditions. Differences from the reference category indicate that Colombians, West Indians, and the European and South American respondents grouped under 'Other Nationalities' are the most prone to engage in this form of transnationalism: relative to the Cubans, each of these groups is over 800 percent more likely to have undertaken one or more trips home.
Knowledge of parents' home country in 1992 and frequency of religious service attendance may be viewed as aspects of selective acculturation because they represent an interest in one's own ethnic origins and in the heritage and traditions accompanying it. Religious affiliations and frequency of attending religious services by nationality are shown in Appendix B. Additionally, intact families provide greater opportunity for selective acculturation over the course of childhood since children in this situation are part of solid structures of socialization transmitting the parents' culture. Teenage childbearing is negatively related to trips abroad, but only when not controlling for nationality. The responsibilities of caring for children curtail the freedom to travel, especially for young adults. The nationality differences in fertility among these groups are sufficiently large to account for this factor, so that it becomes insignificant when controlling for national origin.
Being born in the U.S. and preferring to use a foreign language with one's friends are both positively related to trips to parents' countries when controlling for nationality. Concerning U.S. nativity, it is possible that the absence of any early childhood memories of the circumstances of exit and experience of reception means that curiosity about the country of origin will be more intense while any aversion stemming from early childhood trauma associated with migration, particularly for more disadvantaged groups, will have been avoided. It is also possible that those born in the U.S. worry less about challenges to the validity of their passports and citizenship because the supporting documentation of their birth certificates can be produced.
Being sentenced for a crime is negatively associated with trips back, which is consistent with our initial hypothesis that transnational behaviours may be related to the outcomes of segmented assimilation because dissonant acculturation predicts both lack of interest in parents' country of origin and a greater likelihood of downward assimilation. A competing, and equally plausible hypothesis is that incarceration, like premature parenthood, restricts freedom of movement. Those in jail or on parole are simply not at liberty to go wherever they please. In the end, however, ethnicity, sex, and family structure play the key roles in this form of second-generation transnationalism. The existing literature indicates that first-generation immigrant males are significantly more prone to engage in different transnational activities and that the latter vary greatly with the  (Portes et al. 2002;Guarnizo et al. 2003). Present results indicate that these effects extend to the second generation, largely determining willingness to return to the home country or leave it behind. Table 3 presents the results of an ordered logistic regression of frequency of sending remittances to a parents' home country on the same set of predictors. National origin is less salient in this indicator of transnationalism, as illustrated by the greater number of the independent variables that are significant regardless of whether nationalities are entered into the equation. Those that are significant and positively related to sending remittances, whether or not nationalities are included, are sex (male), family cohesion in 1996, language preference with friends in 2002, knowledge of foreign language in 1992, and being unemployed.

Second-generation remittances
As shown in Appendix A, family cohesion is an index consisting of standardized, unit-weighted replies to three items indicating the internal level of family solidarity, as perceived by respondents in early adolescence. The primary difference in these models, when controls for national origins are introduced, is that frequency of religious attendance, having been sentenced for a crime, and age become insignificant. Colombian, Haitian, and West Indian origins absorb the effects of these variables, as these ethnic backgrounds tend to affect both the tendency to send remittances and religiosity and incarceration simultaneously.
Again, most significant predictors are associated with selective acculturation: family cohesion, knowledge of a foreign language in 1992, and foreign language preference when speaking with friends. The fact that these variables were measured ten years earlier than the predicted effect assures a clear causal order. That being unemployed (and not attending school) should be positively related to sending remittances is puzzling. At this point we can only speculate about the possible reasons behind this finding: 1) the experience of unemployment may create more empathy with family members suffering hardship in the country of origin (which would be consistent with the significant relationship of not being born in the U.S. and sending remittances); 2) there is a possibility that unemployment compensation may find its way into remittances more easily than earnings; (3) that many respondents working in the informal economy may have reported themselves as unemployed; or, 4) it is possible that there is more disruption and chaos both in their own lives and in the lives of family members in the country of origin. In other words, disadvantageous circumstances contributing to parents' reasons for emigrating may not have been overcome either by relatives in the home country or by incorporation in the host country. Economic instability may therefore persist within families even as they maintain their ties across borders. Subject to the structural constraints of impoverished neighbourhoods and slack local labour markets in both the sending and receiving countries, resources from the richer economy may be sent back even if the sender experiences poverty and spells of joblessness. To be poor and unemployed in a country with wealth is not the same as being poor and unemployed in an impoverished country. This interpretation supports the argument that emigration and remittances are emerging as the real economic adjustment programme of the poor in many less developed countries, including those that are prime sources of emigration to South Florida (Portes and Hoffman 2003).
As indicated above, variables that are significant only without controlling for nationalities are age, having been sentenced for a crime, and attendance at religious services. That religious service attendance is positively related to sending remittances and that this relationship should be patterned by differences in nationality points to the involvement of churches as transnational actors. Even after considering the unique situation of Cubans, there is no reason to expect that such efforts are the same across different immigrant populations and their countries of origin.
It is important to note that having been sentenced for a crime is positively related to sending back remittances in contrast to its relationship with trips back home. Of course, being incarcerated or on parole restricts one's freedom to travel. But regarding remittances, we note that the qualitative follow-up to the third CILS survey identified examples of young men from disadvantaged backgrounds resorting to criminal activity to meet what they perceived as the familial obligation of the senior male breadwinner in cases where fathers were absent. 3 Although this might not be the general pattern, it points to the possibility that the importance of sending remittances may in some instances take precedence over the perceived importance of staying within the law.
In this instance, as with unemployment, transnationalism and indicators of downward assimilation are positively correlated. This runs contrary to the earlier conclusion concerning return travel and points to a more complex causal pattern. Common to both cases are the significant and resilient effects of sex, family and national origin. Males continue to be more prone to engage in transnational activities of either kind, regardless of nationality. Solid and solidary families, operationalized by the family cohesion index, are more likely to preserve enduring ties to their kin back home across generations.
Finally, the nationality differences between particular immigrant groups condition their level of religiousness, how prone they are to surveillance and arrest by the police, and their cross-national solidarity. The same nationalities that register higher levels of incarceration in the second generation are those more inclined to continue sending remittances back home, a pattern that reflects an association, among young Miamians, between transnationalism and downward assimilation.

Sense of belonging
Our third indicator of second-generation transnationalism is attitudinal rather than behavioural. Whether or not one feels equally at home in one of two countries is a binary outcome and therefore logistic regression is the appropriate model. The dependent variable taps the extent to which these young adults have abandoned old loyalties, as expected by conventional assimilation theory or still live in two cultural worlds. Among the set of independent variables that we have been examining only two are significant regardless of whether we control for nationality: age and language preference with friends. Table 4 presents these results.
There is a gap of only six years in the age range of the respondents in the Miami CILS second follow-up survey, with 99.2 per cent of the cases within an age range of four years. This is interesting because age was not significant in predicting number of visits back. Whether children are accompanied by a parent on their visits may have something to do with the relationship between age and feeling equally at home in both countries. Young adults travelling independently may acquire a level of confidence or skill dealing with the various problems and contingencies that arise during trips abroad that is not achieved by those who have only been abroad when accompanied by a parent.
A second plausible reason for this age effect is that older CILS respondents were generally born abroad and arrived in the United States at an older age. While all respondents had been in the U.S. at least five years by the time of the first survey, those coming at a later age may have kept more vivid memories of the places they left behind. Supporting this interpretation is the negative (though non-significant) effect of being U.S.-born or feeling equally at home in the parents' country. Family income is significant only when nationalities are not included in the model. That family income is significant and negatively related to feeling equally at home in both countries indicates that relative economic success and well-being in the United States leads to the more rapid extinction of old country loyalties in the second generation. Not surprisingly, this effect disappears when controlling for national origin because immigrant groups vary in economic performance with some doing much better than others. The nationality groups that bear significantly on this dependent variable are West Indian and the residual category of 'other nationalities'. This finding provides some evidence in support of the idea that minority status in the host country combined with lower status and human capital under certain conditions facilitates adaptive outcomes among the second generation associated with transnationalism. In this instance, these conditions include the fact the distances involved with travel between Miami and relevant home countries are modest and the high volume of traffic deflates the per unit costs of travel.
The importance of this finding is that we now see some quantitative evidence based on a large random sample of a situation where transnationalism is associated with lower levels of parental human capital and membership among underprivileged nationalities, who typically face racism in attempting to overcome discrimination in the labour market. This hypothesis, though already ensconced in earlier theoretical work (e.g. Basch et al. 1994;Smith 2002) was not supported by the research on transnational entrepreneurship by Portes and his collaborators (Portes et al. 2002) and, to our knowledge, has been corroborated previously only through case studies or informal observation. As a whole, we see that while intact families and family cohesion, factors generally associated with successful adaptation, promote transnational behaviours, the latter are equally associated with racialized and lower-status groups at risk of downward assimilation. The latter are also more prone to 'feel at home' in the country left behind, arguably because of their difficult experiences in the United States.

Discussion and conclusions
In the process of maturation and identity formation that ratchets the lives of second-generation youths into differential pathways of success and failure, the establishment and pursuit of ties to their parents' homeland provides yet another venue of opportunities and risks that are not typically open to their native counterparts. Theory and research have produced contradictory predictions and findings concerning the likelihood of immigrants pursuing transnational ways of being and ways of belonging. In the process of testing relationships between variables associated with segmented assimilation outcomes in early adulthood, we have found some evidence to support several alternative, though not necessarily contradictory, arguments. Specifically, we find evidence pointing to different transnational patterns that are stratified by class and contoured by nationality and (by extension) ethnicity and gender. Hence, there are aspects of, and empirical contingencies within, the second-generation experience that call for additional work taking segmented assimilation as a broad frame of reference for the advancement of meso-level theory on the transnational aspects of identity formation. Our analysis confirms that transnational ways of being and belonging intersect with different assimilation trajectories depending on individual resources, social networks, and the institutional obstacles and opportunities garnered by families and immigrant communities.
The findings that emerged are not only interesting and provocative, but also particularly significant because they are based on a large random sample of second-generation young adults. The generalizability of our findings is noteworthy. In particular, seeing what appears to be evidence of sustained transnational connections via remittances between the poor and marginalized shows that just because there is an international division of labour does not mean that labour and its reproduction is divided internationally when it comes to families and households diversifying the geographical scope of their income-earning activities. In the same vein, the possibility that the informal and illicit economies in the developed world fuel the growth of remittances points to the development of transnational communities as highly stratified social formations */ interstices where the class structures of host and sending societies converge.
The institutional barriers of host country labour markets, the redlining of impoverished neighbourhoods and the maintenance of highly unequal educational opportunities derail many youths' chances for upward mobility through conventional mainstream channels. Parts of the class structures of Haiti, the West Indies, Nicaragua, Cuba, and Colombia persist in South Florida because there are two sides to the institutionalized channels of opportunity and upward mobility. The other side, in effect, is an organized system of social and economic exclusion. These are the outcomes referred to in the literature as 'downward assimilation' and we see these converging, in several forms, with the persistence of transnationalism in the second generation.
Religion is another social arena of identity formation among the second generation (as it is with many young adults). The CILS-III questionnaire items on religion enabled us to test the hypothesis that religiousness, as measured by frequency of religious service attendance, is related to our indicators of transnational involvement even after taking into account segmented assimilation outcomes. That religiousness is positively related to travel back to sending countries, positively related to sending remittances, and not related to 'feeling equally at home' in both countries perhaps suggests a different type of stance towards one's parents' country of origin depending on whether members of the second generation are more or less committed to a transcendental ideology. 4 Transnational behaviours associated with greater religiousness may reflect an altruistic dimension in these young persons' lives which is a causal force different from that associated with negative adaptation outcomes. Such interpretation is congruent with the parallel finding that family structure and family solidarity also lead to transnational activism. Drawing on the literature on social capital (Portes 1998;, we may say that, aside from the reactive identities promoted by downward assimilation, 'bounded solidarity' promoted by family ties and 'value introjection' founded on religious belief are key factors leading to the continuation of ties with the home countries across generations.
Finally, demographic characteristics Á/ age and sex Á/ also play a role with all other factors controlled, males are significantly more prone to engage in transnational behaviours, while older respondents, who generally arrived in the U.S. at a later age, are more likely to express greater attachment to their countries of origin. Both results are reasonable and conform to our understanding of the dynamics of transnational ties for children of immigrants.
Prediction in the social sciences is risky. Thus, we feel that any attempts to specify the conditions under which transnational practices or attitudes will be embraced by the second generation should be offered only in a highly circumspect manner or confined to the obvious. First, states and the policies they pursue or enforce matter (Smith 2003). Partition, or its functional equivalents, block transnational flows just as the U.S. trade and travel embargo on Cuba stifles the transnational ties that would otherwise traverse the Florida Strait. Second, the factors that facilitate international migration, travel, and trade generally make transnational practices easier, particularly the transportation and telecommunications infrastructure, producer services, and proximity as has been discussed extensively in the past literature on transnationalism. Comparing different migration flows during different historical periods underlines the variations produced by the relative presence or lack of such limiting factors (e.g., Smith 2000).
Particularly important in the case of adult children of immigrants in Miami is that its role as a centre of finance and trade between Latin American and the Caribbean and the broader world economy (particularly the United States and Canada) and the fact that it is the main point of departure for Caribbean tourism has given it a highly developed infrastructure for travel, trade and communications throughout the Americas with particularly low per-unit costs. Thus, while the findings are generalizable to the young adult second generation in contemporary Miami they reflect, in part, the unique conditions of that metropolis.
CILS was designed to serve several vital research purposes simultaneously, rather than being aimed exclusively at the issue of segmented assimilation and transnationalism. We thus conclude with a 'wish list' for future research. It is important to point out first that the focus on nationalities, language, religion, family background, and segmented assimilation outcomes granted us considerable leverage to develop and estimate meaningful statistical models. Questionnaire items addressing these issues should be retained in future research. It would be useful, however, to have corresponding data on parents' transnational practices, the nuclear and/or extended family members remaining in the communities and countries of origin, and the possible effects of divorce or separation followed by the migration of one parent with children on the frequency of trips back to the home country (e.g., as in transnational custody issues). It will be important as well to have more detailed information on employment and selfemployment, including explicit prompts to capture the possibility of informal employment as a possible source of remittances.
Likewise the purposes of visits back to the country of origin were not queried and additional international experience for business, tourism, or the maintenance of multi-local familial ties remain uncharted territory. Information on the transnational activities and involvements of the organizations in which respondents participate, for instance their churches, would also be desirable, pointing to the need for a multi-level analysis of transnational migration. The fact that we have had to offer speculative interpretations of some counterintuitive findings points to loose ends that cannot be resolved with the data in hand. Particularly important is the need to investigate further the positive relationship of unemployment and criminal history with remittance behaviour and what remittances actually mean in terms of identity formation within key segments of the population of young adults comprising today's second generation.
1. As explained by Portes and Rumbaut (2001) 'the survey team in each city [which included several members who were long-term permanent residents] combined its own knowledge of the area with the cooperation of the respective school districts to target schools serving the principal immigrant nationalities, those containing students from smaller immigrant groups, and a control sample of schools where students of native parentage were dominant'. 2. Tests for overdispersion in the Poisson regression models (available on request) revealed that these counts did not follow a true Poisson distribution, confirming that negative binomial regression is appropriate. 3. One such example is presented and discussed in another article in this issue. See Fernández-Kelly and Konczal (this issue). 4. Some observations from the field experience of both the quantitative data collection and the qualitative component of the third round of the CILS in Miami seem pertinent. One is that the transnationalism literature typically assumes foreign experience will be with the country of origin and is silent on the possibilities of transnationalism and foreign travel experience involving additional countries. Some of the Cuban respondents we spoke with had extensive travel experience in various Latin American countries, and we had news of respondents having emigrated as far as Australia and questionnaires and inquiries from verified respondents coming in from Spain and England (neither of which was a parental country of origin).
Likewise, against what appeared to be a largely consistent and conventially mainstream attitude of believing in God but having little or no use for a Church , there also ran an undercurrent of alternative, if not exotic, religious beliefs. Examples include Sufism, Rastafarianism, Santeria and Lukumi, and even belief in Wicca. Some of these were explicit and consciously chosen apart from parental beliefs, others came part and parcel with the religious heritage brought by the parents.
This points to an aspect of identity formation among a segment of the second generation formed by relatively footloose explorers of the world, its societies and their cultures and beliefs. Lacking the kind of historical depth that builds over the course of generations in some locales, and the regular social obligations and reciprocities that come bundled with that depth, some among the second generation in the United States have much greater latitude than most people to take advantage of their additional linguistic and cultural resources to explore the world and its various customs and belief systems.