Association Between Parental Educational Attainment and Youth Outcomes and Role of Race/Ethnicity

IMPORTANCE The concept of minorities’ diminished returns refers to the smaller protective effects of educational attainment for racial and ethnic minority groups compared with those for majority groups. OBJECTIVE To explore racial and ethnic differences in the associations between parental educational attainment and youth outcomes among US adolescents. DESIGN, SETTING, AND PARTICIPANTS A cross-sectional study was performed of 10619 youth aged 12 to 17 years who were participants at wave 1 of the Population Assessment of Tobacco and Health (PATH) study, a nationally representative survey, in 2013. Data analysis was performed from August to October 2019.


Tobacco Dependence
Tobacco dependence was calculated using the following 3 items: time to wanting to use first tobacco product after waking, ever having strong cravings to use tobacco, and ever feeling that one really needed to use tobacco. Responses to these items ranged from 1 (low) to 5 (high). We operationalized tobacco dependence as a dichotomous outcome (1, any level of dependence vs 0, no dependence).
This measure is a standard measure and is used to model tobacco dependence in youth. 48

Aggressive Behavior
Two items were used to measure aggressive behaviors. Participants were asked "When was the last time you did the following 2 or more times: were a bully or threatened other people and started physical fights with other people?" Items were positively correlated (r = 0.30). Item responses ranged from 1 to 4: 1, never; 2, over a year ago; 3, 2 to 12 months ago; and 4, past month. We calculated a sum score that ranged from 2 to 8, with a higher score indicating more aggressive behaviors. These items are taken from the externalizing behaviors component of the Global Appraisal of Individual Needs-Short Screener. 49 This measure predicts future risk of arrest or incarceration within the last 12 months. This variable has been previously used as a categorical variable reflecting high-risk individuals, with very good predictive validity. 50 Studies have shown that the Global Appraisal of Individual Needs-Short Screener instrument has adequate internal consistency, as well as overall and subscale construct validity. 49

Low Grade Point Average
Participants were asked "What is your current overall GPA?" The answers were "mostly A's," "A's and B's," "mostly B's," "B's and C's," "mostly C's," "C's and D's," "mostly D's," "D's and F's," and "mostly F's." Low GPA was treated as a binary outcome (0, mostly As vs 1, any other status). Self-reported GPA and grades are well-established outcomes and are commonly used to measure school performance.
Validation studies have shown that, overall, the self-reported GPA has very low absolute over-and underreporting, which emphasizes its validity across grade levels and subject areas. 51

Psychological Distress
Psychological distress was measured using the following 4 items. "When was the last time you had significant problems with the following: feeling very trapped, lonely, sad, blue, depressed, or hopeless about the future; sleep trouble, such as bad dreams, sleeping restlessly, or falling asleep during the day; feeling very anxious, nervous, tense, scared, panicked, or like something bad was going to happen; and becoming very distressed and upset when something reminded you of the past?" Item responses ranged from 1 to 4 as follows: 1, never; 2, over a year ago; 3, two to twelve months ago; and 4, past month. 52 We calculated a sum score that ranged from 4 to 16, with a higher score indicating more psychological distress (Cronbach α = 0.82). These items are taken from the internalizing behavioral component of the Global Appraisal of Individual Needs-Short Screener instrument, 49 which was previously used to categorize youth as low vs high risk. This simplified classification has good predictive strength, 50 as well as internal consistency and overall and subscale construct validity. 49

Chronic Medical Conditions
To measure CMCs, parents were asked whether any physician has ever told them that their child has any of the following 7 CMCs: hypertension, high cholesterol, asthma, bronchitis or pneumonia, attention-deficit/hyperactivity disorder, dental problem, and diabetes. Parent reports provide valid information regarding CMCs, although some bias due to self-report measurement is expected. We treated CMCs as a binary outcome, where 1 equals 1 or more CMCs regardless of their type, and 0 equals no CMCs.

Statistical Analysis
We analyzed the data using SPSS statistical software version 23.0 (IBM). To analyze the PATH data, survey weights that are due to the sampling design (clustered stratified sampling) were applied.
Because survey weights were considered, the results are representative of the US youth population.
Taylor series linearization was used to re-estimate the variance of the variables. For our data analysis, we first examined the distribution of our variables. We did not perform linear regression because 3 of our variables were not linearly distributed (ie, tobacco dependence, aggressive behaviors, and number of CMCs). We also tested the assumption of homoscedasticity (eg, random distribution of error terms), which failed for tobacco dependence, aggressive behaviors, and number of CMCs.
Because we had 5 outcomes with different distributions, because it was essential for us to use the very same multivariable modeling approach to all outcomes (for comparability of MDRs across outcomes), and because we could not perform linear regression for 3 of our outcomes, we chose to operationalize all our outcomes as dichotomous. Thus, for multivariable analysis, we applied binary logistic regressions to all our outcomes (ie, tobacco dependence, aggression, low GPA, psychological distress, and CMCs). We ran 2 logistic regression models per outcome, all in the pooled sample that included white, black, Hispanic, and non-Hispanic participants. The first model did not have any interaction terms. The second model included 4 interaction terms between race and ethnicity with parental educational attainment (high school and college graduation). From our logistic regression models, we reported odds ratios (ORs), 95% CIs, and P values (2-sided χ 2 tests). P < .05 was considered statistically significant. Data analysis was performed from August to October 2019.    outcomes, higher parental educational attainment was negatively associated with the undesired youth outcomes (tobacco dependence, aggression, and low GPA), and for 2 outcomes, high parental educational attainment was positively associated with undesired outcome (psychological distress and CMCs).  ; P = .003) and CMCs (OR, 1.56 [95% CI, 1.14-2.14]; P = .005). All these findings suggest that the protective effect of high parental educational attainment on youth development might be systemically smaller for Hispanic and black than for non-Hispanic youth. These findings suggested that high parental educational attainment has a less significant association with positive youth outcomes for black and Hispanic families than for non-Hispanic families.

Discussion
The current study showed significant interactions between race and ethnicity with parental educational attainment on youth outcomes. Overall, the associations between high parental educational attainment and youth outcomes were found to be less significant for ethnic minority families than for non-Hispanic white families. However, this pattern was heterogeneous and inconsistent across ethnic groups and level of education that showed the interaction.
Because MDRs could be observed across outcomes and ethnic minorities (ie, blacks and Hispanics), we hypothesize that the probable causes for the observed MDRs of parental educational attainment are some upstream social forces. This is because proximal or downstream determinants of tobacco dependence, aggression, low GPA, psychological distress, and CMCs vary to a large degree. These outcomes, however, overlap in terms of upstream distal social determinants. [53][54][55]  Abbreviations: NA, not applicable; OR, odds ratio. a Interaction term was not entered into the model.

JAMA Network Open | Public Health
Association Between Parental Educational Attainment and Youth Outcomes and Role of Race/Ethnicity This study confirmed what was previously found for black adults, 41-43 black youth, 27-29 and Hispanic adults. [44][45][46] Compared with non-Hispanic white youth, black and Hispanic youth remain at high risk of asthma, 36 attention-deficit/hyperactivity disorder, 37 mental health problems, 25 depression, 30,31 obesity, 29,35 dental health problems, 56 poor health care use, 26 low GPA, 33 poor school attainment, 24 poor school bonding, 34 impulse control, 27 and cigarette smoking, 38 all of which are disproportionate to their high parental education.
These patterns are not limited to youth but carry over into adulthood. These patterns also are not limited to educational attainment 25 and are seen for income, 42 employment, 57 and marital status. 32 Minorities' diminished returns are well described for obesity, 29,35 substance use, 44 depression, 42 happiness, 58 affect, 58 anxiety, 32 self-rated health, 59 CMCs, 37 oral health, 26 and mortality. 57 This study suggests that the same patterns also apply to Hispanic youth.
Because of the existing MDRs, we should observe worse-than-expected outcomes in Hispanic youth from middle-class families. Thus, ethnic health disparities are not limited to the low SES sections of the society because other social mechanisms are at work to diminish the health return of parental education for ethnic minority families.
In this study, Hispanic youth showed less aggression than white youth. This seemingly counterintuitive result could be understood in the context of the Hispanic health paradox 60,61 and the healthy immigrant effect. 62 Similarly, black youth had less tobacco dependence than white youth. Some better mental health outcomes of blacks compared with whites have been attributed to resilience, which might be a function of adversity. 63 As mentioned already, despite the general trend, some inconsistencies were observed. We found differences in MDRs based on race and ethnic minority group, educational level, and the outcome. We need to conduct more studies to understand the circumstances in which a particular educational level does promote an outcome for one but not another ethnic group. Despite these heterogeneities and nuances, black and Hispanic youth tended to be at an overall disadvantage compared with non-Hispanic white youth.

Future Research
There

Policy Implications
These results suggest that specific types of policies and public health programs may be needed to reduce ethnic health disparities. Because disparities are shaped by the differential outcomes of SES, there is a need to go beyond policies focusing exclusively on health to policies addressing the broader social processes placing ethnic minorities at a relative disadvantage. There is a need for innovative national and local policies that reduce disparities due to differential SES and also those that happen at identical SES levels (ie, MDRs). 20,26,27,29,32,41,44,56,67 There is a need to discover policy and program solutions that equalize a minority population's abilities to leverage its available educational attainment. 19,20 Future research should explore the role of discrimination in the labor market, 41

Limitations
This study had some methodological limitations. The cross-sectional design prevented us from making any causal inferences. The sample size was imbalanced across racial and ethnic groups. We only included black, Hispanic, and non-Hispanic white youth. Other ethnic minorities such as Asian populations should be included in future studies. We only studied the differential outcome of educational attainment. Other family SES indicators, such as income, employment, and wealth, were not available. These important components of SES are not accounted for in our analyses, and they may confound the associations between education, ethnicity, and youth outcomes. The degree to which disparities remain beyond SES should be determined in the studies that can control for a wide range of SES components. We did not have data on area-level SES factors. This study did not include variation by zip code or geographic location. Despite these limitations, we believe this study still contributes to extending the existing literature on MDRs among US youth. A large sample size, a random sampling, generalizable results to the United States, and the multiple outcomes studied are among the strengths of this study.

Conclusions
In the United States, black and Hispanic youth are at a relative disadvantage compared with their non-Hispanic white counterparts regarding the magnitude of the association between educational attainment of their parents and youth outcomes.