Segmented Assimilation

1
Segmented Assimilation Latino Childrens
Health Behavior

• Jinsook Kim, PhD, Northern Illinois University,
• Anne R. Pebley, PhD, University of California Los
  Angeles, Noreen Goldman, DSc, Princeton
  University
• November 07, 2006
• APHA 134th Annual Meeting, Boston, MA

2
Significance of the Issue

• Growing proportions of second generation
  immigrant children in US (20 of population
  ageslt18 in 2000)
• Potentially significant impacts of immigrant
  childrens health behavior on the future US
  health profile
• Criticism on acculturation frameworks
• Limitations of acculturation frameworks for
  immigrant children

3
Segmented Assimilation Theory

• Alternative model of immigrant adaptation
• Attention to immigrant profile changes
  divergent adaptation outcomes
• Emphases on the importance of social contexts in
  adaptation of racial/ethnic minority immigrants
• Decreased mobility chances for less-educated/low-s
  killed immigrants
• Less favorable reception than for European
  immigrants of early 1900s

4
Segmented Assimilation Theory

• Three Possible Adaptation Patterns
• Straight-line upward mobility with time ?
  acculturation and integration into the middle
  class
• Economic advancement with continued attachment to
  ethnic culture and group identity
• Downward mobility with integration into the
  underclass less-educated racial/ethnic minority
  immigrants

5
Segmented Assimilation Theory

• Assumptions
• Minority immigrants higher likelihood of living
  in disadvantaged neighborhoods due to residential
  segregation in US
• Higher prevalence of risky behaviors in poor
  neighborhoods due to adversarial subculture of
  the neighborhoods and lower aspiration

6
Specific Aims

• Test two assumptions of the segmented
  assimilation theory using empirical data
• Examine how neighborhood characteristics are
  associated with health behavior of Latino
  immigrant children and youths
• Evaluate differentials in health behaviors among
  1st, 2nd, and 3rd/higher generation Latino
  immigrant children and youths

7
Significance of the Study

• Apply the segmented assimilation framework to
  examine health behavioral aspects of immigrant
  adaptation
• Test main assumptions of the segmented
  assimilation theory
• Use multilevel modeling to account for
  interdependence of observations within clusters
  and to examine cross-level interactions

8
Hypotheses

• Residential Segregation Hypothesis
• Latino immigrant families and African American
  families are more likely to live in economically
  disadvantaged neighborhoods compared to Whites
  with similar family characteristics.

9
Hypotheses

• Neighborhood Effect Hypotheses
• Lifetime smoking, drug use, and having sex among
  children and youths in poor neighborhoods are
  more prevalent than in non-poor neighborhoods.
• Lifetime drinking among children in poor
  neighborhoods is less prevalent than in non-poor
  neighborhoods.

10
Hypotheses

• Behavioral Assimilation Hypothesis
• The prevalence of risky health behaviors among
  US-born Latino children is similar to that of
  African American counterparts, controlling for
  demographic and family characteristics.
• The prevalence of risky health behaviors are
  highest among 3rd/higher generation Latino
  children, lowest among first generation, and
  second generation in between, controlling for
  demographic and family characteristics.

11
Data Sources

• Wave I 2000-2001 Los Angeles Family and
  Neighborhood Survey (L.A. FANS) data
• Part of a longitudinal study of a representative
  sample of Los Angeles County individuals
• Information on social background, household
  socioeconomic status, family life, neighborhood
  life, health status, etc.
• Multi-stage sampling design
• 65 census tracts from 3 poverty strata
  (non-poor, poor, very poor)
• Blocks sampled dwelling units listed
• Households sampled (40 to 50 households per
  tract interviewed)

12
Data Sources (contd)

• In households with children, one child (agelt18)
  chosen at random ? a sibling selected at random
• Sampled children age 9 or older interviewed about
  school, behavior, and family relations
• Children aged 12-17 answered a full set of
  behavioral questions including sexual behavior
  drug use
• Additional information about a child provided by
  the childs primary care giver
• Young adult sample (age 18 to 20) with smoking
  information

13
Data Sources (contd)

• Neighborhood-level information from the 2000
  Census tract-level data
• linked to individuals and families of the
  L.A.FANS data
• neighborhood quality information, including SES,
  residential mobility, and ethnic composition

14
Data Structure

A B
Individuals
15
Variables

• For Residential Segregation Hypothesis
• Outcome variables
• Living in poor (poor or very poor) neighborhood
  or not (binary)
• Median household income of the tract (continuous)
• Predictor variable Ethnicity-generation variable
  with 5 categories (White, African American, 3rd
  or higher generation Latino, 2nd generation
  Latino, 1st generation Latino)
• Control variables family characteristics

16
Variables

• For Neighborhood Effect Hypothesis
• Outcome variables
• Ever smoking
• Ever drinking
• Ever using drugs
• Ever having sex
• Predictor variable
• Neighborhood poverty in 3 categories (non-poor,
  poor, very poor)
• Median household income of the tract

17
Variables

• For Behavioral Assimilation Hypothesis
• Outcome variables
• Ever smoking
• Ever drinking
• Ever using drugs
• Ever having sex
• Predictor variable Ethnicity-generation variable
  with 5 categories (White, African American, 3rd
  or higher generation Latino, 2nd generation
  Latino, 1st generation Latino)
• Control variables Individual family
  characteristics

18
Analysis

• Residential Segregation Hypothesis
• Two-level logistic regression and linear
  regression, adjusting for clustering of
  individuals in households
• N f (ethnicity-generation group, X)
• N (neighborhood quality) living in poor (i.e.,
  poor or very poor) neighborhood median
  household income
• X (family characteristics) family SES (earning,
  assets, household heads education level), age of
  household head, number of household members

19
Analysis

• Neighborhood Effect Hypothesis
• Chi-square tests of equal distribution of each
  health behavior outcome across neighborhoods
  (non-poor, poor, very poor)
• Simple logistic regression with each neighborhood
  quality outcome (poverty category median
  household income) as a predictor and one of 4
  health behaviors as an outcome

20
Analysis

• Behavioral Assimilation Hypothesis
• Three-level logistic regression, adjusting for
  clustering of individuals in households and
  neighborhoods
• HB f (ethnicity-generation group, X)
• HB (health behavior) ever smoking, ever
  drinking, ever using drugs, ever having sex
• X (individual family characteristics) age and
  gender of an individual, household heads
  education level

21
Individual Characteristics (n1,453)
Characteristic Characteristic Mean (std.dev) or percent
Age (years) Age (years) 13.5 (3.2)
Gender Male 49
Female 51
Race/ethnicity White 21.5
African American 10.8
Latino, 1st generation 17.8
Latino, 2nd generation 38.3
Latino, 3rd generation 11.7
Living in Very poor neighborhood 32.3
Poor neighborhood 32.9
Non-poor neighborhood 34.8

22
Household Characteristics (n1,089)
Characteristic Mean (std.dev)
Number of household members 5.0 (1.9)
Log family earnings 8.85 (3.40)
Log family assets 7.56 (4.25)
Household heads education (years) 11.3 (4.6)
Household heads age (years) 40.6 (9.7)

23
Neighborhood Characteristics (n65)

Characteristic Characteristic Mean (std.dev) or percent
Median household income () Median household income () 44,859 (27,563)
Poverty category Very poor (top 10) 30.8
Poor (between 10-40) 30.8
Non-poor (bottom 60) 38.5
Racial/ethnic High Asian Pacific Islander 10.8
Predominantly White 13.8
Predominantly Latino/African American 12.3
Predominantly Latino 38.3
White Other 11.7
24
Results
Health Behaviors () by Race/ethnicity-generation
Race/ethnicity-generation Ever Smoking Ever Drinking Ever Using drug Ever Having sex
n 1,445 777 777 773
White 24.8 48.9 23.1 11.3
African American 14.9 31.9 17.6 25.3
Latino, 3rd generation 21.3 42.7 17.7 21.1
Latino, 2nd generation 15.3 36.9 16.2 14.7
Latino, 1st generation 22.7 35.4 10.4 12.0
Total 19.4 39.6 17.1 15.4
Significant (plt0.05) based on a Chi square test
of equal distribution across race/ethnicity-genera
tion groups.

25
ResultsOdds Ratio of Living in Poor
Neighborhoodsa
Odds Ratiob (95 CI) P-value
White (reference category)
African American 5.3 (3.1, 9.3) lt0.001
Latino, 3rd generation 1.9 (1.1, 3.3) 0.017
Latino, 2nd generation 8.0 (4.7, 13.6) lt0.001
Latino, 1st generation 9.0 (4.8, 17.0) lt0.001
a Living in a census tract with the top 40 of
poverty distribution (poor or very poor) b
Controlled for household SES (asset, earning,
household heads education level), family size,
and household heads age, and adjusted for
clustering of individuals in household
26
ResultsCoefficients of Median Household Income
Coefficienta (robust SE) P-value
White (reference category)
African American -2.7 (0.2) lt0.001
Latino, 3rd generation -1.7 (0.2) lt0.001
Latino, 2nd generation -2.4 (0.2) lt0.001
Latino, 1st generation -2.6 (0.2) lt0.001
a Controlled for household SES (asset, earning,
household heads education level), family size,
and household heads age, and adjusted for
clustering of individuals in household
27
Results
Health Behaviors () by Neighborhood Poverty
Neighborhood poverty Ever Smoking Ever drinking Ever Using drug Ever Having sex
Non-poora 20.3 40.3 18.8 12.4
Poorb 19.9 40.8 17.2 15.4
Very poorc 17.8 37.5 14.8 19.4
a census tracts with the bottom 60 of poverty
distribution b census tracts with poverty
distribution between 10 and 40 c census tracts
with the top 10 of poverty distribution
28
Results
Results (Odds Ratioa) of Multivariate Multilevel
Models
Race/ethnicity-generation Ever Smoking Ever Drinking Ever Using drug Ever Having sex
African American 0.61 0.51 0.60 2.23
Latino, 3rd generation 0.94 0.84 0.45 1.75
Latino, 2nd generation 0.61 0.69 0.42 0.93
Latino, 1st generation 0.48 0.60 0.19 0.58
n of individuals 1,404 764 764 760
n of households 1,054 644 643 641
a Controlled for age, gender, and household
heads education level, and adjusted for
clustering of individuals in households and
neighborhoods plt0.05

29
Key Findings

• Assumptions of the segmented assimilation theory
  partially supported
• Residential segregation hypothesis supported
• Neighborhood effect hypothesis partially
  supported (sexual behavior)
• Behavioral assimilation hypothesis supported
  more risky behaviors with generational increment
  among Latino children
• No inter-level interaction