2. Materials and Methods
2.1. Study Design
This secondary data analysis uses baseline data from a randomized controlled effectiveness evaluation of a parenting education program, Educación Inicial, operating in collaboration with a national conditional cash transfer (CCT) program (Prospera) in rural communities in three states (Puebla, Oaxaca, and Chiapas) in Mexico. Details of the results from the Randomized controlled trial (RCT) and parenting program can be found in the impact evaluation, [
39] and mediation analysis [
40]. This study was approved by the institutional review boards at the University of California, Berkeley and the National Institute of Public Health in Cuernavaca, Mexico (protocol ID: 2010-05-1528).
2.2. Sampling and Data Collection
A sample of 102 indigenous and 102 non-indigenous communities were selected for randomization according to the following study eligibility criteria: (1) rural location (population < 2500) in the Mexican states of Chiapas, Puebla, or Oaxaca; (2) at least 15 families with children ages 0–2 years; (3) at least 70% of families in the community eligible to receive CCT benefits; and (4) no current or prior operation (within the past five years) of Educación Inicial.
All CCT program beneficiary families in the eligible indigenous and non-indigenous communities were invited to participate in the study. Baseline data were collected on 2472 children ages 4–18 months in 2008, and 1929 children (78%) were eligible for a developmental assessment. Of the eligible children, one was lost due to missing the HOME Inventory, five were lost due to missing the Extended Ages and Stages Questionnaire, and 34 were missing the Ages and Stages Socio-Emotional Questionnaire. In several cases, children were missing more than one measure, which resulted in an analysis sample of 1893 children. We retained the original RCT evaluation sample for this secondary data analysis.
2.3. Measurement of Parenting Practices
Stimulating parenting practices were measured using the Infant/Toddler version of The Home Observation for Measurement of the Environment (HOME) Inventory [
41]. The Infant/Toddler version of the HOME is applied to children ages 0–3 years, is composed of 45 items scored as yes or no, and is conducted through an in-home observation and interview of the parent (in this study, the mother was interviewed). A higher score on the HOME indicates more supportive and stimulating parenting practices, although the instrument does not set any screening cutoff points. The HOME has six subscales: (1) parent responsiveness to child’s behavior (Responsivity subscale); (2) parent avoidance of inappropriate restriction and punishment (Acceptance subscale); (3) family routines and safety and predictability of the home’s physical environment (Organization subscale); (4) play and learning materials in the home for development (Learning Materials subscale); (5) parental active engagement in the child’s development (Involvement subscale); and (6) parental involvement of a variety of people and experiences in the child’s daily life (Variety subscale) [
42,
43]. The items that comprise the instrument were selected based on empirical evidence, and then validated through testing [
43]. The instrument has been well validated in the U.S. and used worldwide, including in several Latin American countries [
44,
45]. We conducted an exploratory factor analysis that supported the use of the six subscales in our population; the individual items loaded onto six factors that cumulatively explained the variance in HOME subscale scores. Conservative imputation of zero was applied to retain a score for the total HOME for children missing 1–3 questions (74 missing 1 item, 8 missing 2 items, and 6 missing 3 items). We chose a more conservative imputation since this is our primary outcome of interest. Less than 5% of the analysis sample (
n = 88) was missing between 1 and 3 questions on the HOME; only one child was missing more than three questions on the HOME, and was therefore excluded from analysis. The internal consistency of the scale was satisfactory (Cronbach α = 0.8227).
2.4. Measurement of Child Development
Child’s baseline cognitive development was assessed using the Extended Ages and Stages Questionnaire (EASQ), a developmental screening tool used globally. The EASQ was adapted for use in Mexico by researchers at the Instituto Nacional de Perinatología in Mexico City, and administered to parents as an interview. The EASQ is designed for children 4–60 months old and measures three developmental domains: gross motor function, personal–social ability, and communication; individual domain scores are summed to produce a global score [
46]. The Ages and Stages Questionnaire: Socio-Emotional (ASQ:SE) is a screener for children at risk of emotional or social problems [
47]. Missing scores were imputed using the mean of the child’s domain score for any child missing fewer than 3 questions in a given domain. EASQ and ASQ:SE scores were standardized using two-month age intervals with a mean of 100 and standard deviation of 15.
2.5. Measurement of Household, Parent, and Child Characteristics
Data were collected on child age (in months) and sex (boy, girl), parent education (kindergarten or less, completed primary school, completed secondary school or higher), whether the father was present in the household, family self-identification as indigenous, household crowding (number of people in the house divided by the number of rooms), state of residence (Chiapas, Oaxaca, or Puebla), and household wealth. Household wealth was measured using a household asset index, that was consolidated using principal components analysis [
39,
40], and then divided into quartiles, to identify differences between income groups. Fewer than 1% of follow-up observations were missing data for mother’s education, while 4.8% of observations were missing data for father’s education, and 5.2% for father’s presence. No data were missing for child age or sex, household wealth or crowding, family self-identification as indigenous, or state of residence. Where demographic data were missing, values were imputed using the community mean. In addition to the above measures, household composition (number of adults and children in the house) and the availability of electricity and piped water in the home were used to describe baseline differences between indigenous and non-indigenous communities, but were not included in the analyses.
2.6. Measurement of Community Characteristics
A community was defined as indigenous or non-indigenous based on the prominence of the indigenous population. Indigenous communities were classified as such, by National Institute of Statistics and Geography (INEGI), if at least 70% of the population speaks an indigenous language. The community marginalization index, developed by the National Population Council, was used as a measure of community level SES to account for differing SES levels among indigenous and non-indigenous communities. It is a composite measure of 8 indicators (e.g., percent of the population younger than 15 years old that is illiterate or has not completed primary school; average number of occupants per room in households; percent of households with dirt floors; and percent of households without: a toilet, electricity, piped water) that are consolidated using a principal components analysis [
48].
2.7. Statistical Analysis
For our first objective, we tested differences in the HOME subscale and total HOME scores between indigenous and non-indigenous communities using three techniques determined by type of outcome variable: ordinary least squares regression for the total HOME score; Poisson regressions for the Acceptance, Organization, Involvement, and Variety scales; and negative binomial regressions for the Responsivity and Learning scales. We examined unadjusted means and then adjusted for child age and sex, parent education, father residence in the household, family self-identification as indigenous, household crowding and wealth, and community marginalization. We used Bonferroni corrections to account for multiple tests (Per Comparison Error Rate: 0.05/7 = p < 0.007).
To address our second objective, we stratified analyses of predictive characteristics of total HOME score by community indigenous status. We first examined the unadjusted associations between each of the characteristics and total HOME score. We then analyzed an adjusted model that included demographic characteristics (family self-identification as indigenous, child age and sex, parent education, father’s presence, household wealth and crowding, and community marginalization).
For our third objective, we examined modifiers of the association between parenting practices and living in an indigenous community, including self-identification as indigenous, parent education, and household wealth. We examined these factors in separate models to further understand how they may moderate the relation between living in an indigenous community and total HOME score. The modification models were adjusted for demographic characteristics. In all of our analyses, we included indicator variables for state of residence (Chiapas, Oaxaca, and Puebla) and clustered standard errors at the community level. We conducted our statistical analyses using STATA 14 (STATA Corporation, College Station, TX, USA).
4. Discussion
To our knowledge, this is the first large-scale study to examine the relationship between parenting (as measured rigorously by both interview and observations) and indigenousness in a diverse sample of indigenous and non-indigenous families living in a range of communities with varying proportions of indigenous families. The key study findings were that, while living in an indigenous community and family self-identification were independently associated with less stimulating parenting practices, for indigenous families, living in an indigenous community where they were a part of the majority population was associated with more stimulating parenting behaviors compared to living in a non-indigenous community.
Our first objective was to test differences in Mothers’ stimulating parenting behaviors between indigenous and non-indigenous communities. Our findings supported our hypothesis; the negative association between living in an indigenous community and less stimulating parenting was robust to the inclusion of controls for community marginalization, household poverty, and other demographic characteristics. HOME sub-scale scores for all but the Acceptance scale were lower in majority indigenous communities than in non-indigenous communities. The Acceptance subscale of the HOME is a set of questions around parents’ tolerance of non-optimal behaviors from their child, and their avoidance of harsh punishment or excessive restriction of their child [
41]. An example is parents’ avoidance of expressing overt annoyance with or hostility toward the child. Greater parental acceptance in indigenous communities may reflect the findings of Mosier and Rogoff (2003), in how Mayan parents perceive and tolerate the behaviors of their children under the age of six years. That scores differed between indigenous and non-indigenous communities, after controlling for sociodemographic characteristics, indicates that there were factors beyond parental education and poverty that were differentially associated with parenting, in which ethnicity may play a role. Exemplifying this finding, family self-identification as indigenous modified the association between indigenous community and stimulating parenting.
The second objective was to explore demographic characteristics associated with mothers’ parenting practices within indigenous and non-indigenous communities, and how demographic and SES factors may modify the relation between living in an indigenous community and parenting practices. We were able to conduct this analysis based on the fact that “indigenous community” was defined according to the proportion of the population that spoke or understood an indigenous language, and all eligible communities were stratified by indigenous status prior to study assignment. Demographic characteristics associated with stimulating parenting were similar in indigenous communities to those observed in other studies in LMICs, such as household wealth and maternal education [
44]. Wealth was a strong predictor of stimulating parenting practices, particularly in households with wealth greater than the median, but was more strongly associated with stimulating parenting practices in non-indigenous communities among households above the 75th percentile. These findings support the second study hypothesis, that parenting practices in non-indigenous communities were more sensitive to household SES, if defined as household wealth, than in indigenous communities. The difference in the relation between household wealth and stimulating parenting practices may lie in the fact that indigenous households had much lower wealth on average (mean = 0.59 vs. 0.96 index scores) than non-indigenous households, as well as the fact that the economic difference between wealth quartiles may be greater in non-indigenous communities than indigenous communities. This is to say that money matters less in indigenous communities than non-indigenous communities. Potential explanations could lie in a lower income inequality in indigenous communities, or that majority indigenous communities may have a different type of social structure or hierarchy than non-indigenous communities. However, if household SES is defined as parent education, the association is less consistent. For example, maternal education was not significantly associated with stimulating parenting in non-indigenous communities. The findings in the non-indigenous communities were consistent with a gradient effect of wealth on parenting. The findings in indigenous communities did not display a clear gradient, but nevertheless supported the findings of previous literature linking stimulating parenting practices to SES in Mexico [
30] and added nuance of the relative importance of family wealth between indigenous and non-indigenous communities. In contrast, community marginalization (controlling for community indigenous status) was not associated with less stimulating parenting practices, suggesting that community-level poverty was not a good indicator of individual-level outcomes, and variations in parenting practices were determined to a greater extent by individual and household characteristics.
Our final objective was to examine modifiers of the association between parenting practices and living in an indigenous community. We found evidence for differential patterns of parenting among indigenous families depending on where they lived, supporting our hypothesis. Indigenous parents living in non-indigenous communities used less stimulating parenting practices (based on home observation and interview), but indigenous parents living in majority-indigenous communities demonstrated more stimulating parenting practices. There are two possible directions of this association. The first, is that there may be an effect of community on parenting, the second, is that there may be something unique about indigenous families that move out of indigenous communities. Indigenous families that are minorities in their communities may face additional stressors and challenges that can undermine their parenting. It is also possible that there may be greater economic opportunity outside of indigenous communities, leading parenting responsibilities to fall on other providers or household members. Conversely, indigenous families with less stimulating parenting may be selecting to move out of indigenous communities. Less stimulating parenting among indigenous families that live in non-indigenous communities could be a reflection of less social capital. Future studies may examine factors that help explain the direction and causes of this pattern.
The extensive use of the HOME Inventory—an in-depth measure used here—across several decades enables us to compare scores for our sample with those from other contexts. The mean total HOME score for this population (mean = 25.3, SD = 6.1) was lower than the original HOME sample from Arkansas in 1979 (60% African American, 40% European American) with a mean of 30.9 and SD of 7.6 [
49], as well as a mixed sample from multiple U.S. sites (54% European American, 18% African American, 28% Mexican-American) with a mean of 32.7 and SD of 7.1 [
45]. Prior studies in Latin American countries have generally supported use of the HOME Inventory [
23,
50,
51,
52], and scores for this sample were similar to those found in other Central and South American countries, with similar income levels at their respective times of study [
53]. The mean total score of this study was similar to that of some lower income countries, such as Paraguay [
23], and some upper middle-income countries, such as Argentina, Brazil and Uruguay [
44,
50,
54]. However, the mean score for this Mexican sample was lower than a sample in Costa Rica, a lower-income country (at the time of study) [
22], and a sample in Chile, another upper middle-income country [
55]. HOME scores may be similar across Latin American countries with similar collectivist values. Differences may be due to the particular population studied within Mexico: rural, poor, and marginalized indigenous and non-indigenous communities.
Strengths of the present study included having a very large sample of both indigenous and non-indigenous families living in a range of majority-indigenous and non-indigenous communities—with excellent assessment of family environment and resources. This design enabled us to examine the parenting patterns of indigenous- and non-indigenous-identifying families living in majority or minority indigenous communities. There are several additional potential design and methodological limitations to this study. Notably, as this is a cross-sectional study, we cannot make claims regarding the directionality of some of the associations found. In addition, the HOME Inventory—while extensive and considered the highest assessment standard for the field—does not capture every aspect of parenting that could potentially influence child development [
44]. The HOME was designed to capture key elements of parenting and the home learning environment that are associated with child development; through its use around the world, the HOME has identified similarities and differences in parenting practices between cultures and ethnicities [
56]. However, as detailed previously, these associations with child development have not been examined among indigenous populations. There is also a potential for measurement bias of the HOME in indigenous communities, or the HOME may favor a majority culture or ethnicity. However, there was considerable variability in scores, and indigenous families in indigenous communities had higher HOME scores compared to indigenous families in non-indigenous communities. Presumably, indigenous families living in indigenous communities would be less acculturated to the majority culture or ethnicity, countering the notion of measurement bias of the instrument. Nevertheless, there may be selection factors present in the study that are not accounted for. Indigenous families that remain in indigenous communities rather than migrating might be those with stronger economic and social supports, while those families that migrated may be those more likely to have less stimulating parenting practices for reasons not captured in this study. Indigenous families living in non-indigenous communities may also face stigma or other social marginalization not able to be captured in this study.
Despite these limitations, this study makes a unique contribution to the literature on indigenous marginalization, demonstrating that there are concerns in indigenous communities beyond material wealth and education. Families living in indigenous communities on average had lower parenting scores than families living in non-indigenous communities, suggesting that indigenous communities could be targeted for parenting intervention. However, indigenous families living in indigenous communities had more stimulating parenting practices than indigenous families living in non-indigenous communities. These associations speak to complex relations between ethnicity and place that warrant further investigation.