Abstract
The literature on immigrant health has by and large focused on the relationship between acculturation (often measured by a shift in language use) and health outcomes, paying less attention to network processes and the implications of interethnic integration for long-term health. This study frames English-language use among immigrants in the United States as a reflection of bridging social capital that is indicative of social network diversity. Using longitudinal data on self-rated health and the incidence of chronic conditions from the New Immigrant Survey (2003, 2007), I examine the contemporaneous and longitudinal associations between interethnic social capital and health. The results show evidence for a positive long-term effect of linguistic integration on health status, but no cross-sectional associations were observed. Overall, these results highlight the possible role of network processes in linking English-language use with immigrant health and the time-dependent nature of the relationship between linguistic integration and health status.
As the demographic significance of immigration continues in the United States, immigrant health has received increasing scholarly attention. Prior research has examined various health outcomes for the immigrant population, such as self-rated health (e.g., Chiswick, Lee, and Miller 2008), healthcare utilization (Akresh 2009), obesity (e.g., Cawley, Han, and Norton 2009), cancer and coronary heart disease (e.g., Kasl and Berkman 1983), and the incidence of chronic health conditions (e.g., McDonald and Kennedy 2004). This literature suggests a healthy-immigrant effect and a postmigration decline in health that has mainly been attributed to acculturation, which is often measured using proxy indicators, such as length of residence in the United States and English-language use. Being the most widely used measure of acculturation, language acculturation in particular has received a lot of attention. However, the mechanisms that link acculturation, and in particular English-language use, and health outcomes are poorly understood.
While prior studies of the relationship between English-language use and health outcomes largely attribute this association to acculturative processes (Amaro et al. 1990; Deyo et al. 1985; Salinas and Sheffield 2011; Shelley et al. 2004; Trinidad et al. 2006), some suggest the possible role of social networks in mediating the link between English-language use and health behaviors (e.g., Unger et al. 2000). However, it is still unclear to what extent the association between English-language use and health outcomes can be attributed to social network mechanisms and language-based social capital, particularly the role of interethnic connectedness. In addition, most studies focus on English-language use in one social setting, namely English-language use at home, and restrict their analyses to subgroups within the immigrant population, which raises methodological and generalizability concerns. This study develops a comprehensive index of English-language use among immigrants in the United States and examines the implications of linguistic integration for immigrants’ health trajectories, arguing that English-language use is a source of social capital that is indicative of the composition of immigrants’ social networks. That is, among immigrants, English-language use with friends, colleagues, and family represents the extent of embeddedness in coethnic and interethnic social capital.
Despite the documented, multifaceted functions of social networks for new immigrants, very little is known about how they influence immigrant health and health behaviors. While ties to coethnics have generally been associated with the promotion of immigrants’ economic and social well-being (e.g., Bankston and Zhou 1995), the empirical implications of interethnic connectedness for immigrants’ outcomes, and particularly immigrant health, remain unclear. Migration research suggests that immigrants’ social capital extends beyond coethnic networks and increasingly distinguishes between the bonding and bridging roles of immigrants’ social networks (e.g., Lancee 2012), a typology that is important in identifying the various implications of social ties for immigrant health. While migration research also suggests that immigrants’ coethnic ties, including family and kinship ties, provide social capital that is protective against acculturative stress (Finch and Vega 2003), the role of ethnicity-bridging social capital is not well understood. The literature is also silent on whether or not the social capital of new immigrants has any implications for their long-term health outcomes.
This study uses panel data from the first (2003) and second (2007) waves of the New Immigrant Survey (NIS) to examine the short-term and long-term health consequences of new immigrants’ interethnic social capital. The NIS provides longitudinal data on various health outcomes, including self-rated health status and the incidence of diagnosed chronic health conditions. I construct a proxy indicator of interethnic connectedness (linguistic integration), based on English-language use at home, with friends, with colleagues, and with spouse, and examine its cross-sectional and lagged associations with health outcomes. In doing so, this study makes a threefold contribution to the immigrant health literature. First, it examines the implications of bridging social capital for immigrants’ health outcomes, and it extends this analysis beyond the global measure of self-rated health by examining changes in both self-rated health and the number of chronic conditions. Second, whereas prior studies focus on the health outcomes of Hispanic and Asian immigrants, this study uses national data of new legal immigrants to examine the health implications of network composition. Third, it contributes to our understanding of the language acculturation–health link by examining the long-term health effects of immigrants’ interethnic social capital and sheds light on the extent to which interethnic integration influences immigrants’ health trajectories.
Background
Despite the economic opportunities that may come with immigration, ample evidence exists for its deleterious effect on immigrants’ health (Akresh 2008; Cunningham, Ruben, and Narayan 2008; Gordon 1957). In fact, a so-called “healthy immigrant effect” has long been documented in the migration literature, which suggests that immigrants overall report better health status than their native-born counterparts upon arrival and that this health advantage declines over time. As a result, the health status of the foreign-born tends to converge toward their native-born counterparts over time (Antecol and Bedard 2006; Jasso et al. 2004). While this immigrant health advantage has been attributed to various factors, including self-selection, health screening, or underreporting of health condition among immigrants (Akresh and Frank 2008; Chiswick et al. 2008; McDonald and Kennedy 2004), debate persists as to why immigration has a negative effect on health. Part of this decline in health could be attributed to the increased diagnosis of health conditions among immigrants resulting from higher utilization of health services after migration (McDonald and Kennedy 2004). Lower socioeconomic status, language barriers, and experiences of prejudice and discrimination may also be important sources of health disadvantages for the immigrant population (Derose, Escarce, and Lurie 2007; House and Williams 2000). However, explanations for the decline in immigrant health have largely focused on the effects of acculturation (Escarce, Morales, and Rumbaut 2006; Lara et al. 2005).
Acculturation may be defined as “the process by which immigrants adopt the attitudes, values, customs, beliefs and behaviors of a new culture” (Abraído-Lanza, Chao, and Flórez 2005:1244). While the concept of acculturation has gained interdisciplinary traction, its operationalization remains ambiguous, however, as it is often measured by proxy indicators, primarily the length of residence in the United States, English-language proficiency, and English-language use. English-language proficiency and use, in particular, have been the most widely used measures of acculturation, and findings generally associate linguistic acculturation with declining health and negative health behaviors. For instance, Unger et al. (2000) find an association between the degree of English-language use at home and cigarette smoking among Hispanic and Asian immigrants in California. They also find evidence for the mediating role of social norms, particularly peer influence in their friendship networks, in this association. Amaro et al. (1990) find a positive link between English-language use (vs. Spanish) and rates of drug use (marijuana and cocaine) among Mexicans and Puerto Ricans. English-language use at home was also found to positively influence the odds of alcohol use among Asian American adolescents (Hahm, Lahiff, and Guterman 2003). Further, Akresh (2007) finds that English-language use at work is positively associated with the level of dietary change among Hispanic immigrants. Overall, most of these studies focus on English-language use at home and attribute the link between language use and health outcomes to acculturative processes.
Despite the common use of English-language proficiency and English-language use in studies of acculturation and immigrant health, the extent to which English-language use is related to cultural adoption and the mechanisms linking language use and immigrant health are unclear. Further, while English-language proficiency and use are often highly correlated, which is why these two variables are often combined in indices of acculturation, studies often fail to distinguish the mechanisms that link English-language proficiency and use with health outcomes. Recent literature, however, highlights that language proficiency and use represent distinct concepts, and disaggregating them reveals different relationships with immigrant health. In particular, while English-language proficiency “relates to one’s skill with a tool (i.e., language) that may directly influence access to healthcare (e.g., communication between client and clinician) and potentially broader social determinants of illness (e.g., socioeconomic position),” English-language use may be more related to “one’s underlying cultural values, social networks, political ideology or construction of social identity” (Gee, Walsemann, and Takeuchi 2010:563). Social networks in particular are important because English-language use may be highly dependent upon immigrants’ contact with the native-born population and social network characteristics. Moreover, since social networks have been increasingly connected to health outcomes (Kawachi, Kennedy, and Glass 1999; Putnam 2000), it is important to consider the extent to which linguistic integration reflects structural changes in immigrants’ networks versus extant interpretations of English-language use as acculturation or cultural change.
Critics of the cultural perspective on immigrant health have often called for examining the role of structural factors, including social networks and social capital, in explaining links between acculturation and health. Social networks are the web of social relationships that surround individuals, and social capital refers to the resources embedded within social networks, which includes the various types of supportive (e.g., informational and instrumental) resources of individuals within ones’ networks. Social networks play a central role in immigrant communities, and migration itself could be viewed “as a special case of the development of social networks” (Eve 2010:1236) because the migration process is characterized by significant changes in migrants’ networks, which includes both the disruption of social relationships and the establishment of new ties. However, despite the centrality of network processes in migration, very little is known about the health consequences of changes in migrants’ networks after migration. This is an important gap in the literature to consider, given that social network and acculturative processes are inextricably linked and that some have actually found that social network processes mediate or explain the relationship between acculturation and health behaviors (Allen et al. 2008; Unger et al. 2000).
The Significance of Interethnic Network Diversity
Several competing mechanisms could be identified that link social network ethnic diversity with immigrant health. One is acculturative stress, which refers to the social and psychological strain that results from intercultural contact (Berry 2006). At least in the short run, diverse networks may be associated with higher levels of stressors that result from the process of adaptation to a new social and cultural environment. Typologies of acculturative stress suggest that it may have different sources or components (Rodriguez et al. 2002; Salas-Wright et al. 2015). Acculturative stressors include the lack of English-language proficiency, discrimination, family disruptions, pressure from the host culture to adapt, and pressure from the home culture to retain one’s ethnic identity (Rodriguez et al. 2002). To the extent that higher levels of contact with nonimmigrants increases the social and psychological pressures to adapt to the host culture, acculturative stress provides an important link between interethnic social networks and immigrant health outcomes. Particularly, acculturative stress has been linked to poor physical and mental health (Finch et al. 2001; Finch and Vega 2003; Mulvaney-Day, Alegrıá, and Sribney 2007; Noh and Kaspar 2003). Moreover, ties to nonimmigrants could be associated with faster or higher degrees of acculturation, which may have negative effects on health and health behaviors independent of the effects of acculturative stress.
On the other hand, interethnic social contact could indicate a higher level of social integration and a smaller “social distance” with the native population, which may have its own psychosocial benefits. Diverse social networks, in general, have been linked to positive health outcomes through their effects on the diffusion of health-related information (Erickson 2003), promotion of healthy norms and behaviors (Kawachi et al. 1999; Weitzman and Kawachi 2000), social influence, and social support (Berkman and Glass 2000). Social support is a multidimensional concept (House 1981) that includes not only instrumental (e.g., time and resources) and emotional (e.g., empathy and concern) support but also informational (e.g., advice and referral) and appraisal support (e.g., praise, feedback and affirmation). Interethnic networks could, therefore, play an important role in the provision of not only emotional and instrumental social support but also informational and appraisal social support, which is often less emphasized in the migration literature. Better social integration with members of the host society not only lessens the alienating and isolating effects of migration and the disruption in immigrants’ premigration networks but also provides the social capital that is requisite for successful interaction with the various social institutions of the host society. Nonimmigrant ties could also serve as conduits of useful health-related information, such as information about healthcare access, which may be unavailable in coethnic networks. Ethnicity- and nativity-bridging social capital may, therefore, have a positive implication for immigrant health in that they could add to immigrants’ stock of knowledge about the healthcare system and enhance their abilities to make better health decisions. Social ties with nonimmigrants may also influence and facilitate the adoption of health-related norms and behaviors in the host culture. While these may include negative health behaviors, such as smoking, drinking, and unhealthy dietary changes, they could also apply to salubrious behaviors, such as smoking cessation or involvement in sports and physical-fitness activities.
English-language Use as Social Capital
Migration scholars have increasingly recognized the significance of “linguistic social capital” for the economic integration of immigrants (e.g., Chiswick 1991), but it is only recently that its noneconomic and social significance has received attention (Nawyn et al. 2012). Even though research on acculturation and linguistic isolation has focused on English-language use at home, the use of English language in different social settings is often a key source of language-based social capital. For instance, immigrants’ ability to form nativity- and ethnicity-bridging social ties, which are often an important source of information that is instrumental in the economic and social adjustment process, is dependent upon both language proficiency and use. English-language proficiency is, in fact, often a necessary but not sufficient condition (Akresh, Massey, and Frank 2014) for English-language use in social interactions, and living in a community of coethnics reduces the odds of English-language use in various social settings.
Prior research on the determinants of English-language use has established that individual, demographic, and community characteristics are responsible for variations in patterns of language use (Stevens 1992). At the individual level, human capital, language proficiency, years in the United States, and age at migration determine patterns of English-language use. Country of origin is also an important determinant of English-language use, given the widespread use of the language around the world. In addition, immigrants’ social environments, particularly linguistic homogamy and residential segregation, are important demographic predictors of language-use patterns, highlighting the close link between immigrants’ social network characteristics and linguistic integration. As Stevens (1992:181) highlights, immigrants’ frequency of English-language use is dependent upon both “the resources and incentives encouraging or allowing them to use English [and] the demographic context that underlies the opportunities for [immigrants] to participate in social situations in which their minority language is a possible means of communication.”
Immigrants’ English-language use, therefore, constitutes a type of social capital because it is not only an indicator of the level of language and cultural assimilation (Alba 1990) but also a key indicator of whom immigrants interact with and the type of social network in which they are embedded (Akresh 2007). English-language use represents bridging (interethnic) social capital since host-language usage among immigrants is influenced more by contacts with natives and coethnics than it is by language proficiency itself (Vervoort, Dagevos, and Flap 2012), and social network composition might actually explain the link between language use and health outcomes (Allen et al. 2008). Therefore, the degree to which immigrants use their native language indicates the extent to which they are connected to their respective ethnic community (Alba 1990). That is, given similar levels of proficiency, immigrants who indicate that they do not speak English at home or with their friends and/or colleagues are more likely to have personal and professional networks composed primarily of coethnic contacts than immigrants who say they do speak English in these social settings. Holding language proficiency constant, therefore, variations in the extent of English-language use in social interactions capture network ethnic diversity.
Analytical Strategy
This study uses a longitudinal approach to examine the relationship between immigrants’ interethnic social capital—as reflected by the extent of linguistic integration—and measures of health status. I employ a lagged dependent variable (LDV) approach, which addresses several methodological concerns in the aforementioned literatures. First, an important concern in migration research is the extent to which variations in immigrant outcomes are determined by variations in unmeasured premigration characteristics. Second, a criticism of social-capital research on health has to do with the issue of causality (Kawachi, Subramanian and Kim 2008), and some have generally questioned the causal status of social capital, indicating its potential endogeneity in statistical models (e.g., Chen 2011; Mouw 2003, 2006). Third, while self-rated health has been used extensively in the literature (e.g., Kawachi et al. 1999), primarily because studies have shown it to be a reliable predictor of mortality (Idler and Benyamini 1997), its predictive ability among immigrants and its efficacy in cross-ethnic comparisons of health status have been questioned (Finch et al. 2002). Recent studies have suggested that change in self-rated health is a better indicator of general health status than self-rated health measured at one point (Ferraro and Kelley-Moore 2001; Han et al. 2005).
The lagged model controls for baseline characteristics and allows me to measure differences in health outcomes between two time points after migration. The lagged dependent variable, therefore, serves as a control for pre-immigration confounders that may be associated with both linguistic integration and health outcomes. It also allows me to rule out reverse causality, which is a concern in health research generally. Looking at the relationship between baseline interethnic social capital and health outcomes in both 2003 and 2007, the LDV model allows me to establish the appropriate temporal order between the two variables. While temporal ordering alone does not allow me to make a causal connection between linguistic integration and health outcomes, it does address concerns about reverse causality between social capital and health.
Data and Measures
Data
Panel data from the adult sample of the NIS were used to analyze the effects of social capital on various health outcomes. The NIS is a longitudinal survey of recent immigrants age 18 or older who obtained legal permanent residency between May and November 2003 (i.e., as soon as the immigrant was granted permanent residency). The sample includes both new arrivals and adjustees—immigrants who changed their visa status after some time in the United States. The sample, however, does not include undocumented immigrants and immigrants who may be in the process of status adjustment. A sample of 12,500 new adult immigrants was selected from this sampling frame, and a total of 8,573 interviews were completed for a response rate of 69% in the first wave. The second round of interviews was conducted with 8,456 respondents from the first round between 2007 and 2009 and includes 4,363 (3,902 fully completed and 461 partially completed) adult interviews, yielding a response rate of 46.1%. That is, from the 8,456 adult respondents in the first round, only 4,363 completed round 2 interviews (48.4% attrition rate). The survey also provides probability weights to correct for the oversampling of migrants with employment visas. Respondents with missing data on the dependent variables, self-rated health and the number of chronic conditions, were excluded from the analysis. However, one limitation of using these data is the high attrition rate and other missing data on key covariates, particularly household income. Multiple imputation by chained equations (with M = 50 imputations) was used to impute missing data on all covariates, which provided an analytic sample size of 3,885.
Dependent Variables
The dependent variables were two measures of health status: self-rated health and the number of chronic diseases. Below is a description of each of the outcome variables. See Table 1 for descriptive statistics on these variables.
Descriptive Statistics for Outcome Variables, New Immigrant Survey (NIS) Panel (2003–2007).
Note: Data are weighted with sampling weights. Poor/fair self-rated health is a dichotomous variable indicating fair or poor health (1) versus good, very good, or excellent (0). The number of chronic diseases ranges from 0 to10 in 2003 and 0 to 11 in 2007.
Self-rated health
The NIS includes a measure of self-reported health with five response options (“excellent,” “very good,” “good,” “fair,” or “poor”). Respondents were specifically asked, “Would you say your health is excellent, very good, good, fair, or poor?” The variable was dichotomized into fair or poor health (1) versus good, very good, or excellent (0).
Chronic diseases
For chronic diseases, respondents were asked if they were ever diagnosed with each of the following conditions: high blood pressure, diabetes or high blood sugar, cancer, heart disease, chronic lung disease, heart attack, angina, stroke, psychiatric problems, arthritis, asthma, poor eyesight, poor hearing, trouble with pain, and depression. For each respondent, the total number of diagnosed chronic diseases was calculated.
Independent Variable
Interethnic connectedness
While language usage has been widely used as an indicator of cultural assimilation, it also reflects the kinds of personal and professional networks in which immigrants are embedded (Akresh 2007). In this study, an index of interethnic connectedness was constructed as an ordinal measure from questions about language usage with friends, at work, and at home. Specifically, respondents were asked to check from a list (“check all that apply”) the languages they use at home, at work, with friends, and with spouse. Four variables were constructed for English-language usage with friends, at work, at home, and with spouse. Each of those items were coded into three possible responses: “no English” (0), “some English” (1), and “English only” (2). The final interethnic connectedness index is the row total of English-language usage with friends, at work, at home, and with spouse, and it ranges from 0 to 8. These values represent increasing levels of linguistic integration. Therefore, holding English-language proficiency constant, higher levels of linguistic integration as measured by English-language usage indicate increasing levels of interethnic social interaction in the immigrant’s social network, and nonusage is highly indicative that his or her social network is composed primarily of coethnics.
Covariates
The analyses controlled for other predictors of health, such as human capital, sociodemographic characteristics, and household income. Human-capital measures included both education in years and the level of English-language proficiency. Respondents self-rated their English-language ability on a scale of 1 to 4 (1 being “not at all” and 4 being “very well”). Controls for sociodemographic characteristics included age (in years), gender (1 = female), race, and marital status. Marital status was dichotomized to distinguish between the married and no-married (i.e., single, cohabiting, divorced, or widowed). Since the linguistic integration/interethnic connectedness index included speaking English at work, and speaking English at work is dependent upon employment, the models controlled for employment status (a dummy variable indicating whether or not the respondent was currently employed). Similarly, since those married to someone who does not speak their native language would be more likely to speak English with their spouse and/or at home, the models also controlled for whether or not the respondent had a native-born spouse (i.e., spouse born in the United States).
Being married to a native-born person is also an important control because intermarriage is considered to be an indicator of high level of acculturation. In addition, the models included years since arrival in the United States, commonly used as a proxy measure of acculturation. This was calculated as time since migration (or years in the United States). Finally, given the widespread use of the English language around the world, country of origin influences language use and preference. Therefore, the analyses controlled for whether or not the respondent was from an English-speaking country. 1 Binary indicators of church membership and attendance were used as controls for religious involvement, which is an important source of social networks for new immigrants (Bankston 2014). Finally, the models also controlled for household income and household size. Household income was coded as four categories: less than $25,000 (reference category), $25,000 to $49,000, $50,000 to $75,000, and more than $75,000. See Table 2 for summary statistics on the full set of covariates, which were all measured in 2003.
Descriptive Statistics for the New Immigrant Survey Panel (2003–2007), N = 3,885.
Note: Data are weighted with sampling weights. Values are based on multiply imputed data (M = 50).
In addition, the models accounted for acculturative stress and changes in health behaviors. Unfortunately, the NIS did not include direct measures of acculturative stress. It is important to note, however, that acculturative stress is often conceptualized in terms of experiences of discrimination, negative reception in the host country, and lack of language and cultural competences (e.g., Orjiako and So 2014). The models already included key correlates of acculturative stress, particularly language proficiency, English-speaking country of origin, and race-ethnicity, which are often the basis of discrimination against immigrants and contribute to group differences in the context of reception (Portes and Rumbaut 2014). To adequately control for cultural adoption, the models also included indicators of lifestyle changes related to health behaviors, which may be confounders. Health behaviors included cigarette smoking, alcohol drinking, diet, and the level of physical activity. Smoking frequency and drinking frequency measured the number of cigarettes the respondent smoked per day (0 for nonsmokers) and the number of days the respondent consumed alcohol in the last three months (0 for nondrinkers), respectively. In addition, respondents were asked to rate the degree (on a scale of 1 to 10) to which their current diet was different from their diet in their home country. A physical activity measure was constructed based on how many times per week respondents engaged in light and vigorous physical activities.
Statistical Methods
Data from the first (2003) and second (2007) waves of the NIS were used to estimate three separate models for each health outcome. The first model (Model 1) was a cross-sectional model using data from the first wave of the NIS and estimated the cross-sectional associations between linguistic integration and health outcomes. The second model (Model 2) included the same set of independent variables and covariates to estimate their effects on Time 2 health outcomes (i.e., lagged independent variables and covariates). The final model (Model 3) included the lagged dependent variable to estimate the effect of including the lagged variable on the association between social network composition and health. 2 Whereas Model 1 allowed me to examine if immigrants with higher network diversity had better health outcomes upon migration, Models 2 and 3 enabled me to estimate the consequence of this network diversity for their health outcomes several years later. A comparison of Model 2 with Model 3 allowed me to examine the statistical effect of including the LDV. The lagged model controlled for potential spuriousness and reciprocal effects (Kawachi et al. 2008) and also reduced omitted variable bias due to unobserved characteristics, such as premigration determinants of health status in the United States. Logistic regression models were used for the analysis of the dichotomous self-rated health measure. Negative binomial regression models were used for the incidence of chronic diseases, which was a count variable.
Results
Table 1 presents descriptive statistics for the outcome variables. For the sample overall, there was a decline in health between 2003 and 2007, which is consistent with prior findings that suggest declining immigrant health over time. For self-rated health, approximately 10% of the sample reported poor/fair health in 2003, whereas 21% of the sample reported fair/poor health in 2007. For chronic conditions, the average respondent in 2003 had .81 chronic conditions, whereas the average in 2007 was 1.25. Both of these differences in health status were statistically significant.
Results from logistic regression models of self-rated health are in Table 3. While there is no association between interethnic connectedness and self-rated health in the cross-sectional model, the results suggest that interethnic connectedness, as measured by English-language use with others, has a lagged effect on self-rated health (in 2007), as shown in Model 2. That is, immigrants who were embedded in a more ethnically diverse network in 2003 reported higher levels of self-rated health in 2007 compared to immigrants in less diverse networks. More specifically, a one-unit increase in interethnic connectedness is associated with a 9% decrease in the odds of reporting poor or fair health versus the combined categories of good, very good, or excellent health. This is net of other predictors of self-rated health, including English-language proficiency, which itself has an independent positive effect on self-rated health (a one-unit increase in English-language proficiency is associated with a 25% decrease in the odds of fair or poor health versus the combined categories of good/very good/excellent health). This protective effect of interethnic social capital on self-rated health in 2007 withstands the inclusion of the lagged self-rated health variable, which controls for unmeasured predictors of self-rated health that could bias the results.
Logistic Regression Models of Self-rated Health (N = 3,885), New Immigrant Survey (2003–2007).
Note: OR = odds ratio; CI = confidence interval. Model 1 is a cross-sectional model of self-rated health at Time 1. Models 2 and 3 regress self-rated health at Time 2 on all independent variables at Time 1. Estimates are based on multiply imputed data (M = 50).
p < .10, *p < .05, **p < .01 (two-tailed significance tests).
Negative binomial regression models of the number of diagnosed chronic conditions are in Table 4. Net of all covariates, there was no association between interethnic connectedness and the number of chronic conditions in 2003 (Model 1). Controlling for baseline differences in the number of chronic conditions, interethnic connectedness in 2003 is negatively associated with chronic conditions in 2007. More specifically, a one-unit increase in the interethnic connectedness index in 2003 is associated with a 3% reduction in the incidence of chronic conditions in 2007. However, English-language proficiency is not associated with the incidence of chronic conditions in any of these models, which is contrary to the findings for self-rated health. This points to a potentially important difference between the two measures of health status (for instance, self-rated health measures perceived health status, whereas chronic conditions measure the number of actual/diagnosed diseases) or the different mechanisms through which English-language proficiency and use influence health outcomes.
Negative Binomial Regression Models of the Incidence of Chronic Conditions (N = 3,885), New Immigrant Survey (2003–2007).
Note: IRR = incidence rate ratio; CI = confidence interval. Model 1 is a cross-sectional model of self-rated health at Time 1. Models 2 and 3 regress self-rated health at Time 2 on all independent variables at Time 1. Estimates are based on multiply imputed data (M = 50).
p < .10, *p < .05, **p < .01 (two-tailed significance tests).
Sensitivity Analysis
It should be noted that English-language use and linguistic integration are often highly correlated. Therefore, an auxiliary analysis (available upon request) was conducted, where English-language use and linguistic integration were included in these models separately in order to assess the extent to which the findings on linguistic integration may have been influenced by collinearity. The only substantive difference with regard to the relationship between linguistic integration and health was that the effect of linguistic integration on self-rated health in 2003 was negative and marginally significant (i.e., lower odds of reporting poor/fair health). The statistical significance of the lagged coefficient for interethnic connectedness also improves in these models. However, I caution against interpreting the effects of English-language use without controlling for English-language proficiency, as such models may be reflecting the effects of English-language proficiency and not linguistic integration. While some use a combined index of language proficiency and language use given their correlation, others have cautioned that a combined index may be conflating different mechanisms (Gee et al. 2010). Further, it should also be noted that the variance inflation factors were low across these models, indicating that multicollinearity did not inflate the estimated coefficients.
The analysis also examined whether or not English-language use in different domains has the same effect on health outcomes. Regression models of self-rated health and chronic conditions were estimated, where each of the language-use variables (domains) were included as separate variables (instead of as one composite index). Table 5 shows the results from this analysis, which further highlights the potential importance of network mechanisms vis-à-vis acculturative processes. The results for the cross-sectional models are as one would expect based on the analysis of the linguistic integration index (that is, none of the language-use variables are associated with self-rated health or the number of chronic conditions). The results from the lagged models highlight the particular importance of English-language use with friends and English-language use at work. Particularly, English-language use with friends and English-language use at work are negatively associated with the odds of fair/poor health and the number of chronic conditions, respectively. None of the other language-use variables were individually associated with health outcomes.
Coefficients from Regression Models of Health Status on English-language Proficiency and Use (N = 3,885), New Immigrant Survey Panel (2003–2007).
Note: OR = odds ratio; IRR = incidence rate ratio. Model 1 is the cross-sectional model of health at Time 1. Model 3 is the lagged dependent variable model of health at Time 2. All models control for the full set of covariates listed in Table 2. Estimates are based on multiply imputed data (M = 50).
p < .10, *p < .05, **p < .01 (two-tailed significance tests).
These results suggest that, holding English-language proficiency and English-language use at home constant, English-language use outside the home has positive implications for immigrant health. While linguistic integration in general might be important because it indicates social network diversity, language use at home and outside the home may be related to network processes to varying degrees. This is an important finding because English-language use at home is commonly used as an indicator of acculturation and has been negatively associated with health and health behaviors (e.g., Unger et al. 2000). This study finds no such effect of language use at home on immigrant health. The particular importance of language use outside the home, on the other hand, is better understood as the effect of social networks, as opposed to the effects of acculturation, on immigrant health. To the extent that language use outside the home corresponds to access to weak ties and better social integration in the host society, this is further evidence that social network processes might be particularly important in the link between English-language use and immigrant health. 3
Unfortunately, the NIS does not include direct measures of social network characteristics that would allow cross-validation of the interethnic connectedness index. However, I was able to examine the correlations between the index and other proxy indicators of social network characteristics found in the NIS (results available on request). One such indicator is whether or not the respondent found employment with the help of a relative, which may be an indicator of reliance on coethnic networks (Sanders, Nee, and Sernau 2002; Tegegne 2015). A second indicator is the percentage coethnic and percentage native-born at the religious organization for respondents who reported religious involvement. This analysis shows that the interethnic connectedness proxy index is negatively correlated with percentage coethnic at the religious organization (–.25) and reliance on relatives in the job market (–.19) and positively correlated with percentage native-born at the religious organization (.24). While these measures are only proxy indicators and the correlations are, therefore, relatively weak, they are statistically significant at all levels (p < .001) and are consistent with the argument that linguistic integration is associated with social network ethnic diversity.
A few other robustness checks were conducted. First, given that the lagged models are estimating the effects of covariates on change in health status, I examined if the findings were sensitive to extreme cases of change in health status. Regression diagnostics were conducted to identify potentially influential cases, and the lagged models were rerun excluding “extreme” cases of change in health status. The results are unchanged and confirm that the findings are not due to extreme cases of illness in some respondents between 2003 and 2007. Second, I considered the possibility of bias in the results due to attrition (for instance, due to declining health between the two waves), particularly because there is a relatively high rate of attrition in the NIS panel data. I examined differences on baseline characteristics between attrited and panel respondents. Controlling for sociodemographic characteristics, there was no association between any of the health outcomes and the probability of attrition, indicating no systematic differences between attrited and panel respondents in health status and health behaviors. That is, there is no relationship between Time 1 health status (both self-rated health and chronic conditions) and the likelihood of attrition, and there is no evidence in the data to suggest that attrited individuals may have been those whose health is more likely to have deteriorated between 2003 and 2007.
Discussion
This study focused on the role of language-based social capital in immigrants’ health status. Using longitudinal data, it examined the implications of language-based social capital on long-term health trajectories. The focus was on the potentially ethnicity-bridging role of English-language use among immigrants and the implications of this kind of social capital for self-rated health and the incidence of chronic conditions. The results highlight the time-contingent relationship between of language-based social capital and health status. Overall, whereas interethnic connectedness is not associated with current health status, it appears that immigrants who were embedded in a more ethnically diverse network in 2003 experienced less of a decline in self-rated health and lower levels of incidence in chronic diseases over time.
These results have implications for how the significance of linguistic integration among immigrants is understood. To the extent that one understands English-language proficiency and linguistic integration to largely indicate acculturative processes, then it means that acculturation has a negligible effect on health status in the short term. However, the results would also imply that acculturative processes lead to long-term improvements in self-rated health and lower rates of incidence of chronic health conditions, which is contrary to extant findings in the migration literature—particularly studies that have attributed the healthy immigrant effect to acculturation (e.g., House, Kessler, and Herzog 1990; Stephen et al. 1994). I argue that linguistic integration is better understood as bridging social capital that represents the presence of cross-cutting ties, which is to say a higher level of social interaction with the nonimmigrant population. While social influence through bridging ties may possibly lead to increases in negative health behaviors among immigrants, the long-term effect of interethnic connectedness on health status may very well be protective due to the benefits associated with better social integration with the native population.
Overall, the findings cast doubt on interpretations of immigrants’ English-language use as primarily representing acculturative processes. Prior literature suggests that the acculturation–health link is primarily driven by changes in immigrants’ lifestyle that have negative consequences for health, particularly changes in smoking (Unger et al. 2000), drinking (Hahm et al. 2003), and diet (Akresh 2007). However, this study suggests that there are no associations between linguistic integration and health status in the short term and that the effects of English-language use in the long term may very well be positive. Moreover, this relationship between linguistic integration and health status does not seem to be influenced by changes in health behaviors, as controlling for health behaviors did not influence the association between linguistic integration and health status. Instead, the results lend support for a social capital interpretation of linguistic integration among immigrants, which suggests that English-language use represents structural integration among immigrants that may provide social and psychological resources that are beneficial for health.
A few considerations and limitations should be noted with regard to the data and methodology used in this study. First, a comparison of the cross-sectional and lagged models highlights important methodological concerns in cross-sectional research on immigrant health. These models highlight the presence of lagged effects of interethnic social capital, in the absence of cross-sectional associations. This suggests that longitudinal designs may be required to identify the potentially lagged effects of immigrant integration on health status. Second, in the absence of direct measures of interethnic network diversity, linguistic integration potentially serves as a proxy measure. However, given that language use has been used as a proxy for acculturation in the literature, it is impossible to empirically distinguish the effects of acculturation from the effects of interethnic network integration. Third, the NIS sampling frame includes only newly admitted legal immigrants, and it does not include undocumented immigrants and other legal immigrants who may be in the process of status adjustment. Therefore, these findings do not say much about the health implications of linguistic integration for these groups of immigrants, particularly for undocumented immigrants, who may be very different in terms of their human capital, social capital, and sociodemographic characteristics. Undocumented immigrants, for instance, may have lower levels of linguistic or interethnic integration than other immigrant groups, and additional research would be required to identify the short-term and long-term health implications of linguistic isolation. Finally, since chronic disease status is self-reported and is dependent upon the level of access to quality healthcare prior to and after migration, it is unclear to what extent this may bias the findings. It should be noted, however, that the models control for socioeconomic status, which would be a correlate of access to healthcare or quality of healthcare, and that the findings with regard to linguistic integration are consistent across the two health measures.
To conclude, to the best of the author’s knowledge, this is the first study that uses a longitudinal design to examine the lagged effects of linguistic integration on immigrant health. In doing so, it provides important empirical insight into the time-dependent implications of linguistic integration and highlights important methodological concerns for future research on immigrant health. While one of the strengths of this study is that it uses data from the NIS, which provides a large data set that exclusively sampled new immigrants in the United States and includes data on various measures of health and health behaviors, it also relies on proxy measures of social capital due to the lack of direct social network measures. Future research on immigrant health would benefit from further examining the role of linguistic integration as an important form of social capital by using more direct measures of immigrants’ network characteristics, such as the size, quality, and ethnic diversity of immigrants’ networks (e.g., percentage coethnic in one’s friendship network), to identify the extent to which the effects of linguistic integration are driven by structural changes in immigrants’ social networks.
Footnotes
Acknowledgements
The author thanks Jennifer Glanville and Anthony Paik for their useful feedback on an earlier draft of this article.