TABLE 1.

Descriptive statistics on central covariates before and after the reform.

	G9		G8
	n = 2.405		n = 2.413
Variable	M	SD	M	SD	ES	p
Immigration background (1 = yes)	0.23	0.42	0.22	0.41	1%	0.349
Books at home	4.72	1.24	4.73	1.25	–0.01	0.858
HISEI	58.16	15.42	58.41	15.50	–0.02	0.653
Perceptual speed	65.32	11.41	64.98	11.96	0.03	0.660
Reasoning	10.80	1.26	10.71	1.27	0.07	0.023
Class repeater (1 = yes)	0.10	0.30	0.06	0.24	4%	<0.001

Descriptive statistics were estimated using full information maximum likelihood estimation, cluster-robust standard errors, and survey weights. HISEI = highest international socioeconomic index in the family. ES = Effect size. We used Cohen’s d (Cohen, 1988) for continuous variables, which was estimated as M_G9-M_G8 divided by the pooled SD, and differences in percentage points for dichotomous variables. Please also see Hübner et al. (2017a) and Quis (2018) for additional tests of potential selectivity and representativeness and comparisons of differences on covariates across different cohorts. Additional information on the estimation of the survey weights can be found in Schönberger and Aßmann (2014).