Impact of a Social-Emotional and Character Development Program on School-Level Indicators of Academic Achievement, Absenteeism, and Disciplinary Outcomes: A Matched-Pair, Cluster-Randomized, Controlled Trial

20 schools in 10 matched pairs across three Hawaiian islands, mean enrolment 544; 2002-03 to 2005-06 with a one-year post-trial follow-up in 2007

Hawaii, United States; deliberately selected low-achieving, low-income elementary schools.

The developer-led comparator that makes the independence gradient visible. Analysed entirely at SCHOOL level on archival administrative data with 20 clusters, giving 8 to 9 degrees of freedom. The standardized test changed mid-trial. The conflict-of-interest notice is explicit and is quoted here because it is the point - the research was done using the programme and the training and technical support of Positive Action Inc., and Dr Flay's spouse holds a significant financial interest in Positive Action Inc.

Large effects across the board at one year post-trial - standardized test mathematics effect size 0.52 (p = 0.291), mathematics proficiency 1.10 (p = 0.006), standardized test reading 0.54 (p = 0.028), reading proficiency 0.65 (p = 0.043), absenteeism -0.65 (p = 0.001), suspensions -0.87 (p = 0.028), retentions -1.08 (p = 0.210). The abstract summarises effect sizes as moderate to large, ranging from 0.5 to 1.1, for all examined outcomes.

Low as designed. The threats are 20 clusters, a mid-trial change of test instrument, and the declared financial conflict.

Failed under independent analysis. The same programme's Chicago trial was analysed by its developer team and by Mathematica; the developers report substantial effects, Mathematica reports none of 18 impacts significant with all effect sizes at or below 0.09.