Our outcome variable, child health, was dichotomized as “excellent/very good/good” versus “fair/poor,” the latter serving as the reference. In the original NSCH study, participants had been offered the responses “excellent,” “very good,” “good,” “fair,” and “poor.” Our explanatory variables made up seven groups: social disorganization factors, social structural factors, social relationships and social support, parental health, medical insurance, acculturation factors, and demographic characteristics.

Our social disorganization factors comprised two dichotomous variables and one continuous variable. Rundown neighborhood indicated that a parent had (yes) or had not (no) reported his/her neighborhood (a) to have “litter or garbage on street or sidewalk,” (b) to feature “poorly kept or rundown housing,” or (c) to feature “vandalism such as broken windows or graffiti.” Racial discrimination (yes/no) measured whether his/her parent had reported a participating child to have ever been treated or judged solely on race/ethnicity. Safe neighborhood described how much a participating child’s parent agreed that the family’s neighborhood was safe for the child, using the offered responses 4 (definitely agree), 3 (somewhat agree), 2 (somewhat disagree), and 1 (definitely disagree).

Our social structural factors included variables measuring parents’ educational attainment as well as family income. Parent educational attainment gave the highest level of study completed, using offered responses as follows: 1 (8th grade or below), 2 (9th–12th grade), 3 (graduated high school or GED), 4 (vocational school), 5 (some college), 6 (associate degree), 7 (undergraduate degree), 8 (master’s degree), 9 (doctoral or professional degree). Employed parent (yes/no) described parents who had been paid employees during 50 of the 52 weeks preceding NSCH interviews. Family income-to-poverty ratio gave the percentage of federal poverty level that a family’s income represented, figures provided in the NSCH data set. Finally, participation in public assistance programs was measured via two variables, receipt of TANF and receipt of SNAP, describing families’ receipt of associated benefits during the 12 months preceding interview.

We used six explanatory variables to measure social relationships and social support. Single mother (yes/no) described parents who were single female parents. Next, a response scale was used to measure family cohesiveness and involved two survey items. Parents were asked whether their families drew on strengths that family members possessed, and they were asked whether their families talked together about problems they faced. The response scale comprised 1 (none of the time), 2 (some of the time), 3 (most of the time), and 4 (all the time). Scores for the two items were summed to obtain a total score for each parent, with a higher total score indicating stronger family cohesiveness. The measure yielded a Cronbach’s alpha of 0.85.

In addition, we used the dichotomous variable family support to indicate whether a spouse/partner, other family members, or friends were providing a parent with emotional support encouraging his/her parenting efforts. Similarly, the dichotomous variable professional support indicated whether a counselor or other healthcare provider was supplying a surveyed parent with emotional support, and another dichotomous variable peer/religious group support indicated whether a parent had joined a support group of peers or a religious group for the purpose of obtaining emotional support. Finally, neighbor support was measured via a total score from survey items asking parents how much they agreed that adults in the neighborhood (a) know where to get help, (b) watched out for each other’s children, and (c) provided help to other parents when requested to. A relatively high total score implied a relatively strong network of supportive neighbors. For all three items, offered responses were 4 (definitely agree), 3 (somewhat agree), 2 (somewhat disagree), and 1 (definitely disagree). The three items yielded a Cronbach’s alpha of 0.79.

Resembling our child health outcome variable, our explanatory variable parent health was dichotomized as “excellent/very good/good” versus “fair/poor”, the latter serving as the reference. Parent health was a self-reported measure. Our study also considered some variables describing families’ participation in public or private medical insurance or assistance programs. Preliminary analysis of a variable indicating private insurance participation, however, suggested the dichotomous variable was vulnerable to singularity, due to our preliminary modeling’s inability to estimate the variable’s coefficient, or odds ratio. The coefficient could not be estimated because we encountered no cases in which a parent holding private insurance reported a child to be in “fair/poor” health. In light of the real possibility of singularity, our final analysis employed a single dichotomous variable, insured, to indicate a child’s coverage by either public or private medical insurance.

We used three dichotomous explanatory variables to measure a family’s acculturation: parent born in U.S., child born in U.S., and speaks English at home. The latter variable stated whether or not English was the language largely used in the family’s home. We did not include parent’s U.S. residence less than five years as a variable because doing so generated singularity in preliminary analysis. Finally, we used three demographic variables as controls in our modeling. The three were parent age (in years), child age (in years), and girl (boy providing the reference).