Using Social-Emotional and Character Development to Improve Academic Outcomes: A Matched-Pair, Cluster-Randomized Controlled Trial in Low-Income, Urban Schools

14 Chicago public schools in 7 matched pairs (6 pairs retained for state-test analyses), grades 3-8, 1,170 students in the analytic sample; of 624 grade-3 baseline students only 131 (21%) remained at grade 8

Chicago, United States; low-income urban elementary and middle schools.

The single most instructive independence contrast in this archive. THESE ARE THE SAME 14 SCHOOLS as the Illinois site of the federal Social and Character Development trial - same randomisation, same cohort, same years - analysed by two different teams. Many tests here are reported ONE-TAILED, attrition is severe, and the conflict-of-interest notice states that the research used the programme, training and technical support of Positive Action Inc. in which Dr Flay's spouse holds a significant financial interest.

On the Illinois state test, reading effect size 0.22 (p = 0.16 one-tailed, NOT significant) and mathematics 0.38 (p = 0.07 one-tailed, marginal). Subgroup results include reading for African-American boys 1.50 (p = 0.02 one-tailed) and mathematics for girls 0.41 and for low-income students 0.42 (both p under 0.10, one-tailed). School-level absenteeism -0.78 (p = 0.015 one-tailed). Teacher-rated academic motivation 0.39 (significant), teacher-rated academic ability 0.14 (not significant), self-reported disaffection with learning -0.19 (significant) and self-reported grades 0.02 (not significant). Set against this, Mathematica's independent analysis of the SAME schools found none of 18 growth-curve impacts significant with all effect sizes at or below 0.09.

Low as designed. The threats are one-tailed testing, 21% cohort retention, and the declared financial conflict.

Failed under independent analysis of the identical schools. This pair - developer team and Mathematica on the same randomisation - is the clearest single demonstration in the archive that analyst identity can determine the published conclusion.