| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
Gerontology Institute, University of Massachusetts, Boston.
Address correspondence to Jeffrey A. Burr, Department of Gerontology, University of Massachusetts Boston, 100 Morrissey Boulevard, Boston, MA 02125-3393. E-mail: jeffrey.burr{at}umb.edu
 |
Abstract |
|---|
 TOP Abstract Data and MethodsResultsDiscussionReferences |
|---|
Methods. Data from the 1990 U.S. Census of Population are used to evaluate multilevel multinomial logistic regression models of living arrangements among older Mexican immigrants.
Results. The authors found that strong English language skills increase the likelihood of living independently and increase the likelihood of being the head of a household. They also found that living in an area with a relatively high proportion of Hispanic persons increases the likelihood that older Mexican immigrants will live independently as compared with living with others without headship. Finally, interaction models were examined, demonstrating that living in a community with a relatively high proportion of Hispanics reduces the strength of the relationship between English language proficiency and living arrangement outcomes for older Mexican immigrants.
Discussion. The authors discuss the implications of their findings and point out possible avenues for further research.
Recent data indicate that approximately 670,000 persons age 60 and over entered the United States as legal immigrants between 1988 and 1998 (Immigration and Naturalization Service, 2000
). Partly as a result of these immigration patterns, a number of communities have experienced significant growth in their Hispanic elderly populations (Treas & Torrecilha, 1995). Concerns about the economic and social service impact of these demographic changes have prompted policy debates in recent years (Treas, 1997).
Household composition options provide migrants with one strategy for coping with the immigration process (Boyd, 1991; Glick & Van Hook, 2002). Research shows that living arrangements are shaped by resources facilitating independent living, such as economic affordability, as well as barriers to independent living, such as poor health. Cultural norms and preferences are also shown to be an important component (Angel, Angel, McClellan, & Markides, 1996; Burr & Mutchler, 1999). Studies documenting disproportionately high levels of coresidence among elderly Hispanics have identified English language proficiency and immigration history as being associated with household configuration (Burr & Mutchler 1992, 1993).
Studies of the living arrangements of older immigrants are still relatively rare, yet an increasing number of scholars are examining the issue (e.g., Angel, Angel, & Markides 2000; Blank, 1998; Glick, 2000; Glick, Bean, & Van Hook 1997; Kritz, Gurak, & Chen, 2000; Wilmoth, 2000). The purpose of this article is to examine the relationship between English language proficiency and living arrangements among older immigrant Mexicans in the United States. We also examine size of the local Hispanic community as a factor influencing household composition. Finally, we examine whether living in a community composed of a relatively large number of Hispanic persons modifies the relationship between language proficiency and residential outcomes.
Background and Literature Review
Language use and ability structures the individuals, groups, and organizations with which one can interact, especially when the linguistic orientation of a society is skewed toward a single dominant language. In the United States, a person who does not speak English well is potentially limited in his or her efforts to obtain schooling, employment, and social or health services. This limitation may be particularly significant when the linguistic community is small or dispersed.
More strongly and visibly than many other ethnic features, language use and ability acts as a vehicle for boundary maintenance, both supporting the unique cultural identity of its members and acting as a barrier to "out-group" interaction and infiltration. As a source and support of cultural identity, language use reinforces ethnic identity and group cohesion, promoting well-being in the process, while encouraging adherence to traditional norms and values that shape behavior. These norms and values include those relating to familism, gender roles, intergenerational obligation, and status of elders, among others.
Language use not only reinforces group membership identification, but also may increase distance between groups and act as a barrier to effective interaction among linguistic entities. Dependence on the linguistic group will be increased, with network affiliation constrained as a result. Lieberson (1981) argued that the retention of native language among immigrants forms a "shield against assimilation." This shield extends not only to issues of cultural identity, norms, and values, but also to characteristics having implications for socioeconomic success: A correlation between language use and socioeconomic status is observed, which has implications for behavior (Stevens, 1985).
The challenges and opportunities posed by multilingualism among immigrant elderly persons have not been systematically addressed. Indeed, English language proficiency levels among elderly persons have seldom been documented, much less studied in depth (see Mutchler & Brallier, 1999). Nevertheless, some research shows limited English proficiency forms a barrier to obtaining needed social and medical services (Hart, Gallagher-Thompson, Davies, DiMinno, & Lessin, 1996; Tanjasiri, Wallace, & Shibata, 1995), providing an impetus to provide services on a bilingual or multilingual basis. These pressures may be most evident in areas where the non-English-speaking elderly population is large, growing rapidly, and/or particularly needy of formal services.
The concrete ways language ability affects elderly immigrants' lives depend to a large degree on the social and cultural environment within which the older person lives. For the older person with few needs for formal assistance and limited required contact with formal institutions, most of which provide services in English, lack of proficient English skills may not pose a particular difficulty. Insofar as language retention reflects cultural identity, cultural values regarding familism and reverence for the aged may be retained. In short, the non-English-speaking older person with few needs for formal services that would draw him or her outside of the linguistic community may experience limited hardship associated with poor English skills.
For other elderly individuals, poor English skills may form a substantial barrier to independence and general well-being. Linguistic isolation may effectively bar participation of the older individual in the larger society (Zsembik, 1996), possibly serving as an obstacle to constructive interaction with service providers. For non-English-speaking elderly persons, effective social support networks may be limited to members of the ethnic group. In short, an older immigrant with assistance needs (health care, economic) may be dependent on informal support networks—which may be narrow in scope in areas characterized by relatively small numbers of coethnics—as a means to offset these impediments to well-being.
Older individuals living in the United States with significant needs for assistance—such as those with very low incomes or extremely poor health or with smaller social support networks as is often the case for recent migrants—are particularly disadvantaged if they also have limited English skills. Although elderly individuals of all groups rely to some extent on members of their personal and familial networks to provide informal assistance and serve as a conduit to formal services (Litwak, 1985), older persons with poor English skills may be especially dependent on this network. Bilingual members of the older person's social network—such as adult offspring—may form critical bridges between the elderly individual and the larger society, where financial, medical, and other support is often obtained. Although network support does not require coresidence, it often does, especially given cultural norms surrounding age, kinship, and coresidence in traditional societies. On the basis of the arguments presented above and findings from earlier research (Burr & Mutchler, 1993; Wilmoth, 1997; Zsembik, 1996), we hypothesize that Mexican immigrants with poor English language skills will be more likely than those with adequate skills to coreside with others as opposed to living independently.
Few studies have considered the impact of contextual factors on living arrangement outcomes for different race and ethnic groups. This is an important gap in the literature in part because the geographic distribution of elderly persons of different race and ethnic groups in the United States is uneven (Rogerson, 1998), due to historical patterns of internal migration and immigration as well as economic and political forces (Frey, 1995; Funkhouser, 2000). Because older members of immigrant groups are geographically clustered in unique ways, we anticipate that this clustering may be important to living arrangement options, in part because geographic clustering may lead to the development of linguistic communities. Although there has been some research on contextual factors and household structure, this remains an underdeveloped research area, especially for immigrant populations (e.g., Mutchler & Krivo, 1989).
Living in a community with relatively large numbers of like-group members provides a greater opportunity to be embedded in culturally consistent social networks. This is especially significant for immigrants, who are often embedded in transnational social networks and who use these networks to help ease some of the difficulties with transitioning from one culture to another (e.g., Massey, Alarcon, Durand, & Gonzalez, 1987). Thus, living in a community with a high level of coethnic group concentration may facilitate the formation of relatively dense social networks that promote the establishment of culturally consistent formal and informal social support systems for older persons. Older immigrants with poor language skills, low levels of education, and few economic resources may be more likely to receive informal assistance from networks of friends and neighbors, within the context of a community that shares a common language.
Further, community and religious organizations in these areas are likely to include persons of the same ethnic group and thus be more sensitive to the specific cultural and social customs and values of members of these groups. These organizations (e.g., health clinics, transportation services, in-home personal and health services) are more likely to have employees and volunteers who speak the elderly person's native language. Under these conditions, older immigrants may be more likely to establish and maintain residential independence. Thus, our second hypothesis is that older Mexican immigrants are more likely to live independently than to coreside with other adults if they live in communities with relatively large concentrations of Hispanic persons.
Finally, the conceptual arguments presented above suggest that living in an ethnically dense community will modify the relationship between English language proficiency and the likelihood of coresidence. Although we expect that poor English language skills increase dependence on other persons in the elderly Mexican immigrant's social network, promoting a greater likelihood of coresidence, it is likely that the impact of language deficits are reduced when living in an area with larger numbers of Hispanic persons, where a shared reliance on the Spanish language is more likely. English language proficiency also becomes less important when an older immigrant lives in an area where social services are provided in a culturally sensitive manner and where informal social networks are likely to be dense. Our third hypothesis is that the positive relationship between English language proficiency and the likelihood of coresidence among older Mexican immigrants will be modified (weakened) when living in areas with a relatively high concentration of Hispanic persons.
|
Data and Methods |
|---|
|
TOPAbstractData and MethodsResultsDiscussionReferences |
|---|
County data from the 1990 Census of Population and Housing (STF-3C) are merged with individual-level data for our multilevel analyses. The smallest geographic unit identifiable in the PUMS files is the PUMA (public-use micro-data area). FIPS codes on the micro-data file identify PUMAs that correspond to places, counties, or county groups including 100,000 or more persons. We examine a subset of all PUMAs in the continental United States. Included are PUMAs composed of a single county with 100,000 or more population or places smaller than one county possessing at least 100,000 persons. PUMAs identified by place are combined with other places within the same county to form the county unit. We limit our analysis to these large urban areas because the remaining PUMAs are composed of multiple counties that typically have large land areas with relatively low population density and such areas do not have what can realistically be considered local communities with distinct economies and social and political characteristics.
For PUMAs composed of multiple places of 100,000 or more population, covering only a portion of a county, county-level data are assigned to each component PUMA. Data manipulation is required to establish correspondence between the geographic unit provided on the PUMS file (PUMA) and that provided on the STF file (county). In each case, a one-to-one correspondence exists whenever the PUMA is composed of a single county. When a county contains several PUMAs (multiple places of 100,000 or more population), the county-level information is used for all component PUMAs, understanding that in these cases, which are primarily portions of large urban areas, important subcounty neighborhood differences are unobservable. Thus, the smallest geographic unit for which ethnic group characteristics are described is the county, even when the county is composed of multiple PUMAs. We refer to our contextual units as counties.
We evaluate the role of language and ethnic community context separately for married and unmarried persons. Married persons have spouses who typically provide many levels of material and emotional support, as well as providing on-site caregiving in times of acute or chronic illness. By definition, an unmarried person does not have this built-in level of support through marriage. Also, because number of children born is not reported in the Census for men, we further separate our analysis by gender. This means we have three study samples: unmarried women, unmarried men, and married couples. To evaluate married couples, we focus on married men as the unit of analysis and include information about their spouses in the analysis. The sample size for each marital-status/gender group is as follows: unmarried women, 7,669; unmarried men, 2,973; and married couples, 6,570. The county sample size for each marital-status/gender group is as follows: unmarried women, 202; unmarried men, 159; and married couples, 198. County sample sizes are not the same for each group due to variation in the presence of members from each group across counties from which an 8% sample has been drawn.
The dependent variable measures whether an older Mexican immigrant: (a) lives alone, if unmarried, or if married, lives in a couple-only household (with no other adults); (b) lives with other adults as head of household; or (c) lives with other adults without headship. For each gender/marital-status group, we examine three living arrangement contrasts: Model 1, living with others as head of household versus living independently (reference group); Model 2, living with others without headship versus living independently (reference group); and Model 3, living with others without headship versus living with others as head of household (reference group).
We include headship status in our dependent variable as an indicator of social, economic, and functional independence. Analyses (not shown here) indicate that older Mexican immigrants who are heads of household are younger, have stronger English language skills, have higher personal incomes, have higher incomes relative to the total household income, have fewer functional limitations, and are less likely to be recent immigrants.
The central individual-level variable for this analysis is English language proficiency. English language proficiency is based on self-reports of whether English is spoken at home and, if so, how well it is spoken. A dummy variable is created for English language proficiency; 1 = speaks English not well or not at all, 0 = speaks English only, very well, or well.
Other individual variables include whether the older person is a recent immigrant (1 = immigrated between 1980 and 1990, 0 = immigrated prior to 1980) or has a functional limitation (1 = a limitation associated with mobility, with personal care, or both, 0 = no limitations). In addition, economic status is measured by annual (1989) personal income for unmarried persons. For married persons, annual incomes of spouses are combined. The income measures are transformed by the natural log to account for a right skew in the distribution of the data. Education level is measured as 1 = 9 or more years of school and 0 = less than 9 years of school completed. Marital status is included in the regression models for unmarried persons as 1 = ever married and 0 = never married. Age in years is also included.
The Census reports number of children born for women only. For unmarried women, we include a continuous indicator of number of children born. For unmarried men, no indicator of number of children born is available. For married men (couples), we include a continuous indicator of children ever born for the current spouse. We include a measure of current spouse's functional limitation status (measured as above) and English language proficiency (measured as above). We also include an indicator of whether the spouse is under 60 years of age (1 = under 60 years old, 0 = 60 years old or older). This measure is a proxy for the increased likelihood of having younger adult children in the household who have not yet left home or who have returned home. In the regression models, continuous indicators are centered around their grand means.
Conceptually, geographic areas that contain a relatively large number of coethnic group members are more likely to support linguistic communities, increasing the likelihood that larger informal social networks are formed and that culturally appropriate social services will be available. For example, in areas where many individuals speak Spanish, social service agencies are more likely to advertise their services in Spanish, hire Spanish-speaking employees, and be sensitive to group norms and preferences. Operationally, we assess ethnic group concentration using a location quotient (LQ) measure (Smaje, 1995). LQs are calculated according to the following equation: LQ = (m/M)/(t/T), where m represents size of the Hispanic population in the county, M represents size of the Hispanic population in the United States. t represents the population size for all individuals in the county, and T represents the population size for all individuals in the United States. LQs are interpreted as reflecting the geographic concentration of a Hispanics in a county relative to concentration of Hispanics in the United States. A value of 1.0 occurs when the representation of Hispanics in a county is equivalent to the share of Hispanics in the U.S. population. Values of less than 1.0 indicate a relatively low proportion of Hispanics in the county, and values greater than 1.0 indicate a relatively high proportion of Hispanics in the county, relative to the U.S. Hispanic population.
Qualitative research indicates that members of otherwise distinct Hispanic groups who live in geographic proximity often live cooperatively in more general social and economic networks (e.g., Chavez, 1990). This cooperation is facilitated in part by the fact that most members of these groups speak the Spanish language. Thus, LQs are generated for the Hispanic population as a whole instead of for the Mexican-origin population only. This strategy does not imply that these groups are necessarily similar along other important cultural, social, and economic dimensions.
Choosing to coreside with others may also be influenced by characteristics of the local housing market, especially cost. We include in our models a control variable for cost of housing in the local area. This concept is measured as the median rental cost of housing in each county.
Estimating the effects of individual and community characteristics on the living arrangements of older Mexican immigrants is confounded by the fact that immigrants are clustered into geographic areas in systematic ways. Individual effects are not independent of these grouping characteristics (Bryk & Raudenbush, 1992). Statistical analyses that incorporate grouping characteristic variables directly into regression models as if they were individual-level variables fail to account for this fact. Correlated errors among the individual-level variables within the counties may yield misleading results and the estimation of standard errors of coefficients are likely to be inappropriate (typically, the standard errors are artificially low).
To overcome these statistical and substantive issues and to increase the robustness of the parameters, we estimate a set of nonlinear (multinomial logistic regression) hierarchical models of living arrangements, including both contextual and individual characteristics, taking into consideration the complex error structure represented in this class of multilevel models. We represent with a single equation the combined individual- and community-level parts of the model, comparing two categories of our living arrangements variable:
|
|
00, b0, and ok are parameters to be estimated. The set of individual-level variables, b, for person i in county j is represented by Xibj. The set of county-level variables, k, in county j is represented by Ckj. The error term has two components: 
oj representing the random errors associated with the county-level equations for the Level-1 (individual) intercepts and rij representing the error associated with the individual-level parameters.
In addition to testing for the statistical significance of the fixed effects of the individual-level and county-level variables on individual-level living arrangement behavior in the models (the significance of the ok coefficients), we test for the presence of additional variability across counties by estimating random intercepts models. A statistically significant result provides evidence that the average likelihood of independent living varies from county to county. We do not include in our models the full range of contextual variables likely to influence variability in living arrangements across areas. Introducing a random intercept term allows us to account for these unobserved effects. In the regression tables, we report the variance components for the random intercepts, which represent the residuals at the county level for each living arrangement contrast. We also estimate separate cross-level interaction models to examine whether Hispanic concentration modifies the relationship between language proficiency and living arrangements. The parameters are estimated using HLM V.5 (Raudenbush, Bryk, Cheong, & Congdon, Jr., 2000).
|
Results |
|---|
|
TOPAbstractData and MethodsResultsDiscussionReferences |
|---|
|
Next, we use multinomial logistic regression techniques to evaluate multilevel models of household composition among older Mexican immigrants who are age 60 and older. We first present the results from the main effects models for each marital-status/gender group and then present results of an analysis of the cross-level interaction between English language proficiency and Hispanic group concentration. Log-odds coefficients and standard errors are reported.
Table 2 contains regression results from the main effects model for unmarried women. The results indicate that, after controlling for other personal and community characteristics, older Mexican immigrants who have poor English language skills are more likely to live with others than to live independently, either as head of household or without headship. This result is consistent with findings noted in other research (Wilmoth, 2000). Persons with limited or no English language skills are less likely to be heads of households; that is, unmarried Mexican immigrant women with poor language skills are more likely to live with other adults without headship than to maintain headship when living with other adults.
|
The impact of education level and income status works in the way that previous research would suggest (Burr & Mutchler, 1992; Kritz et al., 2000). Persons with more education and those who have more economic resources are more likely to live independent of other adults, irrespective of headship status. On the basis of this analysis, education level has no statistically significant impact on whether an older Mexican immigrant maintains headship when living in a coresidential situation. However, having few economic resources does increase the likelihood that, when living with other adults, older Mexican immigrants are less likely to be head of household.
With respect to the effect of coethnic concentration, the results indicate that living in a county with higher Hispanic group concentration increases the likelihood of living independently as compared with living with other adults without headship among unmarried Mexican immigrant females. This supports the hypothesis that independence is promoted when living in an area with relatively large numbers of coethnics. Our analyses also indicate that living in an area of relatively high Hispanic concentration increases the likelihood (log odds) that headship will be maintained when living with other adults as compared with living with other adults without headship. Living in areas with higher rental costs increases the likelihood of living in more complex households.
Table 3 presents the results of similar models (without the number of children born variable) for unmarried male Mexican immigrants age 60 and over. The pattern of results for the living arrangement contrasts and the key theoretical variables in our models, English language proficiency and relative Hispanic population concentration, is similar to that found for unmarried women. Having poor English proficiency skills reduces the likelihood of living alone and increases the chances of living with others without headship. Also, as relative Hispanic concentration increases, the log odds of living alone or living as head of a complex household increases, compared with living with other adults without headship. As rental costs increase, the likelihood of living with others increases (with or without headship).
|
|
|
|
Discussion |
|---|
|
TOPAbstractData and MethodsResultsDiscussionReferences |
|---|
Three hypotheses were developed and each received some support. First, we found that persons who possess poor English language skills are more likely to live with others compared with those with better English language proficiency. Second, living in an area with relatively high proportions of Hispanic persons does not impact the likelihood of living with other adults with headship versus living independently, but it does increase the likelihood of living with others without headship as compared with living alone. This is arguably the contrast representing the greatest extremes between independence and nonindependence in living arrangements. Third, Hispanic group concentration modifies the relationship between language ability and living with others without headship versus living alone for unmarried persons. The strength of the English language ability effect is reduced. This implies that independence may be more likely when an older unmarried Mexican immigrant lives in an area where he/she is embedded in Spanish speaking social networks and also has access to service environments that are more sensitive to her/his cultural background. The same interaction effect does not surface for married couples. One possible explanation is that married persons have social networks that are more likely to include adult children, adult grandchildren, and other relatives, and as a result, more opportunity (or demand) for coresidence. It may also be that the need for assistance from the informal network or from social service agencies is reduced because married persons have spouses, who act as caregivers in times of need and who may serve as a liaison with the world outside the home.
The results from the analysis of our multilevel models provides evidence that poor English language proficiency acts as a barrier to independence in living arrangements. Consequently, organizations tasked with providing social services (e.g., home delivered meals, home health care, transportation) to Latino seniors should make efforts to hire Spanish-speaking persons who can communicate with older Mexican immigrants. These organizations should also be sure that public service announcements and other forms of public outreach are provided in Spanish.
The effect of Hispanic concentration also has implications for local governments and nongovernmental organizations tasked with providing services to older immigrants. Our results suggest that older Mexican immigrants living in communities with relatively low concentrations of Hispanics are more likely to live in more dependent circumstances. Social, political, and economic power resources that come with a large group presence in the community may not be available in local areas with low Hispanic population concentration. Thus, older Mexican immigrants in such areas may be at a disadvantage when it comes to living arrangement choices. Policy makers and service providers in these communities need to be encouraged to provide culturally sensitive support under these circumstances.
The study design has some limitations that may be overcome by appeals to alternative data sets or through the collection of new data. First, the nature of census data does not allow us to examine the impact of social network variables that lie outside the household. In addition, census data do not contain information on preferences for living arrangements or detailed information on health status. Second, it is likely that counties are not the only geographic social units that are important for helping us understand the decisions that immigrants make regarding their living arrangements. Information collected at the neighborhood level and at levels higher than the county may be helpful. Unfortunately, census data do not allow for analysis of geographic units smaller than the ones used here. Mixed method analysis using qualitative and quantitative techniques represent one possibility for overcoming the shortcomings of survey data like those from the census. Third, the strategy of attaching county data to PUMAs means we are essentially examining large urban areas. Given this design feature, we cannot examine with these data whether the effects found herein also apply to small urban and rural areas, limiting the generalizability of our findings to large urban areas only. Finally, with the high rates of immigration that occurred in the 1990s, it should be informative to replicate this study with data from the 2000 census and possibly develop over time analyses, when the data become available.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication May 24, 2002. Accepted for publication October 24, 2002.
|
References |
|---|
|
TOPAbstractData and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. L. Angel and R. J. Angel Minority Group Status and Healthful Aging: Social Structure Still Matters Am J Public Health, July 1, 2006; 96(7): 1152 - 1159. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
