Table 4.

OLS and Change Models of the Effects of Out-of-Home Placement on Subsequent Cognitive Skills and Behavior Problems, Matched Sample

	Vocabulary	Matrices	Internalizing	Externalizing
	B (SE B)	B (SE B)	B (SE B)	B (SE B)
Model 3: OLS model with full set of controls
Out-of-home placement	-0.55 (0.71)	-0.03 (0.52)	1.25 (0.71)	1.19 (1.04)
Model 4: Residualized change model
Out-of-home placement	-0.33 (0.61)	0.03 (0.50)	1.26 (0.63)	1.22 (0.89)
Model 5: Simple change model
Out-of-home placement	-0.25 (0.66)	0.09 (0.59)	1.27 (0.77)	1.28 (1.04)
Model 6: Difference-in-difference model
Out-of-home placement*Follow-up	-0.66 (0.73)	-0.05 (0.59)	1.20 (0.79)	1.54 (1.09)
Model 7: Fixed effects model
Out-of-home placement	-0.60 (0.73)	0.11 (0.56)	1.36 (0.77)	1.59 (1.08)

Note: 616 matched observations. Regression coefficients (B) and standard errors (SE B) presented. All models control for the number of months between the baseline and follow-up observations, child age, caregiver age, family income-to-poverty ratio, child gender, child race and ethnicity, caregiver marital status, grandparent present, caregiver not US born, caregiver education, family risk score, maltreatment types at initial investigation, whether the initial report was substantiated, and whether the child experienced an out-of-home placement prior to his or her baseline assessment. Note, however, that time invariant measures (child gender, child race and ethnicity, caregiver not US born, family risk score, maltreatment types at initial investigation, whether the initial report was substantiated, whether the child experienced an out-of-home placement prior to his or her baseline assessment, and number of months between the baseline and follow-up assessments) were differenced out of the fixed effects models (Model 7) such that these effects were not directly estimated under this specification. Standard errors from the difference-in-difference models (Model 6) were corrected for intra-cluster correlation due to multiple observations of each child. Treatment and comparison group children were matched on all covariates including the baseline measures of cognitive skills and behavior problems.

^*p<0.05;

^**p<0.01;

^***p<0.001.