Confounding in Non-Experimental Studies Linking Caregiver Warmth to Youth Adjustment: The Problem and Potential Solutions

Objective: The literature linking more caregiver warmth to better youth adjustment is almost entirely non-experimental, leaving estimates vulnerable to confounding bias. We evaluated the extent and potential impact of confounding in these non-experimental designs.Method: Reviewing the literature, we identified potential confounders of the link between caregiver warmth and three indices of youth adjustment: externalizing, internalizing, and prosocial behavior. We then examined the associations between these confounders, warmth, and adjustment in longitudinal data from the ABCD Study: 11,880 youth ages 9- to 15-years-old (48% female, 20% Hispanic, 64% white, 16% Black, 2% Asian).Results: Across analyses, 64-70 of the 80 potentially confounding variables we examined held significant associations with both caregiver warmth and youth adjustment, always in directions that exaggerated the apparent beneficial effect of warmth. When the full set of potential confounding variables were covaried, the apparent effect of caregiver warmth on youth adjustment shrunk by 36%-100% (median estimate = 72%). In follow-up analyses, we could successfully eliminate most of the bias introduced by the full set of 80 confounding variables by controlling for a carefully chosen subset of 2-9 “top confounders” (e.g., prior measurements of warmth and adjustment.Conclusions: There are many confounders of caregiver warmth and youth adjustment in non-experimental data. The magnitude of the resulting bias is quite large and would plausibly change scientific conclusions in a given study. Regularly measuring and adjusting for “top confounders” would help, as would incorporating more quasi-experimental and experimental designs.