Military Connectedness and Children's Individual Adaptation: Analyzing Differences on the School Success Profile with Propensity Score Matching

Background

Military connected (MC) children are a vulnerable population at increased risk for decreased individual adaptation (IA) (Lester & Flake, 2013). Little research exists regarding how varying types of MC status may influence outcomes and what specific outcomes may be affected. This study advanced research on MC via two interconnected analytic techniques. First, we attended to nuances of MC by measuring connections for adults, siblings, and adult and siblings. Second, we used propensity score methods to derive more accurate estimates of the effects of MC group status.

Methods

Sample. Data were collected from 2011 to 2015 from a sample of 9,536 students in 66 middle and high schools across seven states using the School Success Profile (SSP). The SSP contains 263 questions assessing students’ beliefs about their social environment and IA. The SSP is both well validated and widely-used in schools (G. L. Bowen & Swick, 2013). The sample used in the analysis was 50.5% male and 47.2% White, 26.6% African American, and 19.5% Hispanic.

Measures. “Military Connectedness” (MC) was measured with two items assessing whether any “adults” or “brothers and sisters” were currently serving on active duty. From these items, a single composite variable was created that partitioned students into four exclusive MC groups: None (n = 8,163), Adult (n = 479), Sibling (n = 333), and Adult + Sibling (n = 141). Six IA factors from the SSP were modeled: “Academic Achievement” (one item assessing grades), “Physical Health” (nine items assessing health problems), “Positive Adjustment” (six items assessing negative feelings), “Self-Confidence” (five items assessing feelings of confidence), “Trouble Avoidance” (11 items assessing school-based positive behaviors), and “Success Orientation” (12 items assessing positive life outlook). All six IA variables were coded positively on a standardized 0.0–1.0 scale.

Analysis. The main analysis was a multinomial extension of the propensity score framework developed to balance treatment groups and, thus, approximate the benefits of randomization in observational studies (Guo & Fraser, 2015). First, we estimated generalized propensity scores in Stata using a multinomial logit model that included key covariates. Second, the inverse of the propensity scores were used to define three MC sampling weights. Third, the sampling weights plus covariates were included in linear regressions to predict children’s IA outcomes with a correction for clustering within schools. The non-MC students served as the reference group.

Results

MC demonstrated deleterious effects on students’ IA with significant differences across MC groups, indicating that the three groupings represent distinct conditions. All 18 MC regression coefficients (3 groups × 6 outcomes) were algebraically negative and confirmed research suggesting MC is a risk factor for children’s IA. Moreover, analyses revealed statistically significant effects for 11 MC coefficients ranging from B = −.031 to B = −.089. The negative effects from an Adult+Sibling MC were strongest on four of six outcomes.

Conclusions

Overall, this study suggests that future MC analyses and interventions should consider conceptualizing children’s MC status not as binary but as multivariate. Failure to do so may miss important nuances on how MC affects children.