Segmented assimilation: An approach to studying acculturation and obesity among Latino adults in the United States

Abstract

Segmented assimilation theory posits immigrants experience distinct paths of assimilation. Using cluster analysis and data from the National Latino and Asian American Survey, this study sought to apply this theory in relation to obesity among Latinos. Four clusters emerged: a “second generation classic”, a “third generation classic”, an “underclass”, and a “segmented assimilation” pattern. In analyzes controlling for sociodemographic confounders (e.g., age), second generation classic individuals had higher odds of obesity (OR= 2.70, 95% CI 1.47–4.93) relative to the segmented pattern. Similarly, third generation classic individuals had higher odds of obesity (OR=3.23, 95% CI 1.74–6.01) compared to segmented assimilation individuals.

INTRODUCTION

Over the past decade, there has been increasing interest in studying acculturation as an explanatory variable for shifting patterns of health among Latino populations in the United States (1). The growing popularity of acculturation has sparked a heated debate on the sociocultural determinants of health among Latinos (2–5). Measurement issues and conceptual simplification of the acculturation process have been key arguments in this debate. For example, some have cited reliance on “proxy” measures (e.g., length of time living in the United States) fails to acknowledge the role of social and other contextual factors that shape acculturation and the quality of “exposure” to the majority culture (2, 4, 6).

Theoretical paradigms within the social sciences may allow for a more critical examination and better measurement of acculturation. An example is the theory of segmented assimilation, a framework that underscores the interplay of background factors (e.g., human capital, family composition) and context (e.g., hostile labor markets, co-ethnic communities entrenched in low socioeconomic areas, discriminatory practices, and anti-immigrant governmental policies) in determining patterns of integration (7, 8). By noting features that may block immigrants and subsequent generations from full integration, the theory posits that immigrants assimilate into three patterns that are qualitatively distinct. Specifically, some immigrants experience the classic assimilation pattern characterized by the adoption of white middle-class values while simultaneously relinquishing ethnic values. Other immigrants follow an underclass pattern of assimilation, which is typified by poverty, low educational attainment and antagonistic attitudes towards middle-class values. Finally, others experience a selective or segmented path of assimilation characterized by rapid economic and educational advancement while intentionally maintaining ethnic values (7).

Despite its theoretical contribution and the enthusiasm it has received in the social sciences, little health research employs the theory of segmented assimilation. Of this handful of studies, only two focused on specific health outcomes among adults to yield mixed support for the theory. Johnson & Marchi operationalized segmented assimilation patterns by creating an interaction variable from acculturation proxy measures (e.g., nativity; language) and neighborhood characteristics (e.g., Latino immigrant neighborhood) (9). They found increased odds of infant low birth weight among English speakers residing in Latino immigrant neighborhoods when compared to English speakers in non-Latino neighborhoods. Elevated odds of low birth weight were also observed among English speakers residing in Latino immigrant neighborhoods when compared to Spanish speakers in the same neighborhoods. In contrast, Castro et al. used a growth mixture model to capture lifetime changes in socioeconomic status and acculturation to derive four different trajectory groups (10). The study provided some evidence consistent with segmented assimilation theory. Specifically, they observed an upward and downward trend in socioeconomic status, and an association between an upward trend and better health outcomes (e.g., life satisfaction).

Additional studies using segmented assimilation theory can drive forward the literature on health among Latino populations by better distinguishing between the distinct patterns of assimilation, while simultaneously addressing some of the main measurement and conceptual critiques of the acculturation concept. A more nuanced approach to the measurement and conceptualization of acculturation also will help to elucidate health outcomes that have been associated with acculturation. For example, there are striking patterns between obesity and acculturation that suggests a complex relationship, including one a non-linear relationship (11). Longer duration of residence and birth in (versus outside of) the U.S. (proxy indicators of acculturation) are associated with higher risk of obesity (12–17). The typology advanced in segmented assimilation theory could propel research on acculturation and obesity among Latinos as it offers a useful alternative to the linear, somewhat simplified model that dominates much existing research.

This study uses segmented assimilation theory as a guiding framework to conceptualize and measure acculturation and to examine its association with obesity among Latinos in the United States. We use a novel method, cluster analysis, to measure acculturation. We examined two specific hypotheses.

First, we examined whether the data would support the patterns of assimilation proposed by segmented assimilation theory. Hypothesis one: the data will reveal underclass, classic, and segmented assimilation patterns. Second, we sought to explore whether obesity varies with the patterns of assimilation proposed by segmentation assimilation theory. Hypothesis 2: obesity prevalence varies across the three groups, with the lowest obesity prevalence found in the segmented assimilation pattern and highest prevalence among the classic pattern of assimilation.

METHODS

Sample Design and Data Collection

We used data from the 2002–2003 National Latino and Asian American Survey (NLAAS)–one of the largest population-based surveys of Latinos and Asian Americans ever conducted in the United States. (18) Among other advantages of the NLASS study are the myriad of sociocultural variables not currently available in many large epidemiological datasets (e.g., NHANES, NIH). The NLAAS is based on a stratified multistage area probability sample of the English-speaking household population of the continental United States (19). The survey was conducted either face-to-face or via telephone by a fully bilingual lay interviewer from 2002 through 2003 and had a 75.5% response rate for the Latino sample (20). This study analyzes data from the 2,481 Latino respondents from four distinct ethnic subgroups: 577 Cubans; 495 Puerto Rican; 868 Mexican; and 614 “Other” Latinos (i.e., peoples of South America and the Caribbean). Given that this study involved only secondary data analyzes of de-identified existing data, the project was exempt from Human Subjects Review.

Measures

Patterns of assimilation

An explorative cluster analysis was performed to assess whether different patterns of assimilation postulated by segmented assimilation theory would emerge from this data. Prior to performing the cluster analysis, each variable was standardized to remove potentially large effects due to arbitrary differences in the standard deviations or means of the variables (21). Because the measures were on different scales, all continuous variables were standardized into z-scores (M=0; SD=1). Variables were selected using the segmented assimilation framework, in which socioeconomic position, social acculturation, and generation status are essential characteristics of the distinct patterns of assimilation proposed by the theory. Therefore, cluster variables were derived using measures tapping into socioeconomic position, sociocultural constructs, and generation. Variables that tapped socioeconomic position included: annual household income (continuous variable with possible responses ranging from $0 to $1,000,000 or more), education (continuous; possible responses 0–17 years or more), and subjective social status (measured as a continuous variable with the MacArthur Scale of Subjective Social Status, which ranged from 1 to 10 with higher scores reflecting greater status (22). Individual variables that tapped sociocultural indices included English proficiency (sum of three items; higher scores represented greater English proficiency); ethnic affiliation (sum of three items; e.g., “how closely do you identify with other people who are of the same racial and ethnic descent as yourself?” using a 4-point scale, 1=Very; 2=Somewhat; 3=Not very; 4=Not at all); generation status. The latter was a 3-level variable in which first generation was defined as a respondent born outside the U.S., second generation as a respondent born in the U.S. with at least 1 foreign-born parent, and third generation as a respondent born in the U.S. with two U.S.-born parents. These individual variables were used to create different assimilation patters using a two-step procedure in cluster analysis. Once the clusters were created, the segmented assimilation pattern served as a reference category because we hypothesized this group would have the lowest level of obesity.

Outcome

BMI was calculated from self-reported height and weight, using the standard formula of weight (kg) divided by height squared (m2). We defined obesity as BMI> 30kg/m² in accordance with current federal guidelines for the measurement of obesity among U.S. adults (23).

Analyses

Statistical analyses were performed using STATA/SE version 10. Individuals were excluded from the analysis if they (1) were missing self-reported weight and height N=88 or (2) had a questionably large BMI (>65 kg/m²) N=5. Thus, a sample of 2,481 respondents was included in the analysis. Key demographic variables known to be associated with obesity were included as covariates, such as age, ethnicity, time in the U.S, marriage and employment. T-tests and chi-squares were used to examine differences across patterns of assimilations among these same demographic variables Then, to determine whether patterns of assimilation were significantly associated with obesity while controlling for covariates, a logistic regression model was fitted. Obesity was regressed on dummy-coded categorical variables representing the different patterns of assimilation, controlling for age, self-rated physical health (i.e., poor, fair, good, very good, excellent), ethnicity (i.e., Latino subgroup), length of time in the U.S., gender and activity limitations (i.e., having a physical condition that substantially limits the person’s physical activity). To account for sample design effects, the SVYLOGISTIC procedure was used as well as sample weights to adjust for probability of selection and non-response. Statistical significance was set at a p value of ≤.05.

RESULTS

Hypothesis 1: Assimilation Patterns and Sample Characteristics

Four distinct clusters emerged (Table 1). Cluster 1 was defined as “second generation classic” because the SES and acculturation profile corresponds to the classic pattern in segmented assimilation theory: that is, seamless assimilation into American society and relinquishment of ethnic culture. This cluster exhibits higher than average socioeconomic indicators (i.e., education, income, social status), alluding to a higher degree of integration into American society (results not shown). This cluster also had high English ability but low Latino ethnic affiliation, suggesting low attachment to ethnic culture (results not shown). The profile of Cluster 2, “third generation classic”, was very similar to that of cluster 1 except for one important element, generation status: 100% were third generation. Cluster 3, the “underclass pattern” was consisted of first generation individuals (100%), as well as Cluster 4, “segmented assimilation” (100%) Table 1 illustrates mean values of income, education, social status, Language (English proficiency), and ethnic affiliation.

Table 1.

Means (and standard deviations) for patterns of assimilation

	2nd Generation Classic (N=487)	3rd Generation Classic (N=364)	Underclass (N=936)	Segmented (N=592)
Income	53,531(50396)	56,428(49730)	23,430(19871)	70,315(54390)
Education	12.60(2.6)	12.17(2.6)	9.11(3.4)	13.90(2.2)
Social Status	5.83(1.7)	5.59(1.8)	4.70(2.1)	6.40(1.5)
Language	10.28(2.3)	10.38(2.2)	3.90(1.4)	8.89(2.4)
Ethnic Affiliation	9.86(1.7)	9.91(1.6)	10.53(1.7)	10.03(1.8)

As for additional sample characteristics, mean age across assimilation patterns differed, with second generation classic group being the youngest and the underclass group being the oldest (Table 2). There were no differences across gender, employment, and BMI; however, the ethnic composition differed across the distinct patterns (Table 2).

Table 2.

Demographic characteristics by pattern of assimilation

	2nd Classic	3rd Classic	Underclass	Segmented
Age (mean years)	35.76*	36.23	40.12	36.61
Sex (%)
Female	48.41	46.19	45.66	46.23
Male	51.59	53.81	54.34	53.77
Ethnicity (%)
Cubans	3.17	.17	6.36	8.76
Puerto Ricans	16.27	10.62	5.72	12.50
Mexicans	56.89	58.09	62.36*	36.13
Other	23.67	31.13	25.57	42.61
Marriage (%)
Yes	53.55	57.38	71.78	68.13
No	46.45	42.62	28.22	31.87
Employment (%)
Employed	61.07	66.54	62.46	71.63
Unemployed	9.81	8.08	6.19	7.26
Not in Labor Force	29.12	25.38	31.34	21.11
BMI (mean)	28.01	28.69	27.77	26.81

^*p<0.05

Hypothesis 2: Odds of obesity by assimilation patterns

Table 3 presents crude (Model 1) and adjusted (Model 2) logistic regression analyses of the association between assimilation patterns and obesity. Model 1, included patterns of assimilation only, and Model 2 (the adjusted analyses), included age, ethnicity, time in the U.S., gender, activity limitation, and health status as covariates.

Table 3.

Logistic regression of obesity on patterns of assimilation

	Model 1 (OR, 95% CI)	Model 2 (OR, 95% CI)
Patterns of Assimilation
Segmented	Reference	Reference
2nd Generation Classic	1.73 (1.29, 2.52)	2.70 (1.47, 4.93)
3rd Generation Classic	2.01 (1.38, 3.26)	3.23 (1.74, 6.01)
Underclass	1.42 (.93, 2.26)	1.13 (.72, 1.87)
Age		1.0 (.99, 1.01)
Ethnicity
Mexican		Reference
Cuban		.66(.49, .89)
Puerto Rican		.82 (.59, 1.25)
Other		.74 (.57, .95)
Time in the US*
Less than 5 years		Reference
5–10 years		2.27 (1.22, 4.23)
11–20 years		1.51 (.88, 2.68)
21+ years		2.34 (1.38, 4.39)
Gender
Female		Reference
Male		1.20 (.98, 1.57)
Activity limitation
No		Reference
Yes		1.56 (.99, 2.13)
Health status
Good		Reference
Poor		2.33 (1.32, 4.11)
Fair		1.75 (1.27, 4.11)
Very Good		.72 (.53, .99)
Excellent		.49 (.35, .70)

Model 1 shows that the second and third generation patterns had 1.73 and 2.01 respectively, greater odds of being obese compared to segmented assimilation individuals. The odds of obesity for the underclass pattern compared to segmented assimilation individuals were not significant. After adjusting for the effects of age, ethnicity, time in the United States, gender, activity limitation and health status, the association between assimilation patterns and obesity increased. That is, the second generation classic group had 2.70 higher odds of obesity whereas third generation classic individuals had 3.23 higher odds of obesity compared to the segmented assimilation group. This confirms the study’s hypotheses that the classic pattern has the highest odds of obesity compared to the segmented assimilation pattern.

Model 2 in Table 3 also shows that the covariates of age, gender and activity limitation had non-significant associations with obesity. In contrast, ethnicity, time in the United States, and health status were all significantly associated with obesity. Being Cuban (OR=.65) had a significant protective effect on obesity compared to being Mexican. Similarly, Other Latinos relative to Mexicans, OR=.74, had lower odds of obesity. Relative to less than 5 years, living in the United States for five-to-ten years or for more than 21 years resulted in higher odds of obesity. Interestingly, individuals living in the United States between 11 and 20 years only had 1.51 odds in obesity. Lastly, having poor or fair health status was associated with 2.33 and 1.75, respectively, greater odds of obesity compared to good health. Having very good health (OR=.72) and excellent health (OR=.49) was associated with lower odds of obesity compared to those individuals reporting good health.

DISCUSSION

Using the theory of segmented assimilation and a novel cluster analytic approach, this study explored whether obesity varied among the clusters representing varying patterns of assimilation. The crude logistic regression analyses showed that the second and third generation patterns were 1.73 and 2.01 times, respectively, more likely to be obese compared to the segmented assimilation group. This is consistent with findings that greater time in the United States is associated with statistically significant increases in BMI and/or obesity/overweight (12–15, 24–26).

Other covariates such as age, gender, and activity limitations had non-significant effects on obesity. In contrast, health status and ethnicity were significantly associated with obesity. Specifically, being in poor and fair health was associated with greater odds of being obese compared with reporting good health. This finding replicates prior observations on the association between obesity and health (27). Lower odds of obesity among Cubans and “other” Latinos may reflect differences in dietary intake and dietary quality previously identified across Latino ethnic groups (28, 29).

This study has some limitations that should be noted. First, BMI was based on self-reported height and weight, which tends to underestimate the prevalence of obesity. However, weight classifications based on self-reported weight and height have been shown not to differ between immigrants and U.S.-born Mexicans, except for those who are underweight. Second, the theory of segmented assimilation was specifically developed in relation to the assimilation patterns that second generation youth experience. Thus, family dynamics warrant further exploration in health research. Specifically, the family structure (or lack thereof) of immigrant families is of particular interest to the segmented assimilation framework as it can affect the kind assimilation path that the second generation experiences. Family structure is also intimately tied to the assimilation process as it has a major impact on the cultural and economic resources of families (e.g., two parent families have higher incomes than one-parent households) (8). These resources can, in turn, affect the rate at which parents and their children acculturate. This is of major importance because these inter-generational patterns position second generation children into certain patterns of assimilation (8). For example, a pattern of “dissonance” (whereby children reject the values and language of parents) is associated with a range of negative outcomes, while “consonance” (the joint accommodation of children and parents to the values and language of the new culture) or “selectivity (preservation of key elements of the parental culture while learning language and values of host culture) were associated with positive educational outcomes (8).

Further, the theory of segmented assimilation emphasizes that complete Americanization is not necessarily beneficial; however, the protective effects of limited assimilation also come with some disadvantages (30). For example, lack of social ties outside the ethnic community may hinder an individual’s ability to experience the full range of available opportunities. Limited ties to other groups or communities may also place a tremendous amount of pressure and excessive obligations towards relative and other co-ethnics. Further, the theory has been criticized for contending that the challenges faced by today’s non-white second generation are more severe than those faced by earlier European immigrants (8). However, this study was not able to test these theoretical possibilities. The cross-sectional nature of the data presents another limitation. Finally, the age of the dataset is an additional limitation. Unfortunately, there continues to be a dearth of large, longitudinal datasets that include underrepresented ethnic minorities. This type of data is needed to move the field forward.

Despite these limitations, this study utilized the concept of assimilation patterns formulated by the theory of segmented assimilation. Using a novel method of cluster analysis, results provided evidence to support, in part, the assimilation typologies among a representative sample of Latinos living in the United States. Further, an association between patterns of assimilation and obesity was observed. Using a more complex theoretical framework than the models dominating the field of public health to date, the findings of this study are significant in moving forward the study of acculturation and health among Latinos in the United States. These findings also have practical applications in that they can inform strategies among Latinos across generations and assimilation patterns such as designing community or school programs that address the particular needs of second and third generation Latino children and the obesogenic environments that they may be exposed to. It also seems important to understand the mechanism that promotes or hiders the shift in diet throughout the acculturation process for Latinos. For example, changing social networks across assimilation patterns might influence dietary change since robust evidence suggests that eating with others affects individual dietary patterns, and obesity is transmissible through social ties. (31) Future interventions that target the specific social network features that influence obesity-related behaviors among Latinos may be a fruitful way to reduce obesity-related disparities across the acculturation continuum.