Once we're into correlations with socioeconomic status, etc., we need Judea-Pearl-style causal analysis to suss out how determinative biological differences are. (Genotypes are presumably occasionally causes and never effects, but that won't apply to phenotypes.)