Table 2.

Descriptive statistics for Calibration Simulations

Covariate	Sample Mean		|Standardized Difference|
	Experimental	Target
Cohort†	0.76	0.50	0.52
Age	12.34	13.43	1.51
Number of schools attended since 1st grade	3.02	3.31	0.16
Female	0.45	0.51	0.14
Number of siblings†	2.87	2.79	0.05
Educational aspirations	2.63	2.71	0.13
Educational expectations	2.30	2.37	0.10
Sure of graduating from high school	0.60	0.68	0.16
Homework	0.84	0.72	0.26
Talk about school with parents	0.95	0.96	0.08
Sibling dropped out	0.29	0.22	0.17
Ever skip school	0.18	0.12	0.18
Ever late for school	0.54	0.53	0.02
Extracurricular activities†	0.70	0.74	0.10
Reading more than 2 hrs/wk	0.32	0.32	0.00
TV more than 2 hrs/day†	0.74	0.73	0.01
Mother employed	0.53	0.63	0.22
Mother has BA or higher	0.29	0.21	0.21
Father employed†	0.71	0.81	0.25
Father has BA or higher	0.28	0.25	0.08
Not living with both parents†	0.37	0.44	0.16
Family receives public assistance	0.58	0.35	0.48
Primary language not English	0.07	0.08	0.03
Overage for grade	0.83	0.46	0.73
Low grades	0.15	0.23	0.18
Discipline problems at school	0.68	0.52	0.32
External locus of control	0.65	0.45	0.40
Self-esteem	−0.08	0.01	0.14
School climate†	0.14	0.14	0.00
Baseline math test score†	28.16	39.17	0.58
Baseline reading test score†	26.06	37.25	0.63
% days absent (log)	−1.78	−1.87	0.27
Black	0.02	0.17	0.40
White	0.34	0.20	0.34
Hispanic	0.48	0.44	0.08

Note: The table shows experimental sample and target population means for all 35 covariates used in the Calibration Simulations. The last column shows absolute standardized differences, i.e., the absolute value of the difference between experimental and target means divided by the standard deviation in the target population.

^†denotes covariates that moderate the treatment effect at at least 3 out of 5 sites (see text).