Growth-Mindset Interventions at Scale: Experimental Evidence From Argentina

202 public secondary schools (102 treatment, 100 control) in Salta, Argentina; analytic sample of roughly 8,800-9,400 Grade 12 students

Argentina; Grade 12 students in public secondary schools, a predominantly low-achieving population.

School-level randomisation stratified by urban/semi-urban status and school type, with a no-treatment control. The intervention is a single-session adaptation of Blackwell et al. 2007 - students read a passage on the malleability of the brain, write a letter to a classmate, and post it in the classroom - delivered by ministry field staff. Implementation was verified. The outcome is the Grade-12 national assessment, IRT-scaled and administered by the government two months after the intervention, with 93-99% response rates: a fully independent standardised achievement measure. FDR q-values are reported for multiple comparisons. Cost USD 2.82 per student. The author flags that he used the developers materials without the context-specific adaptation the developers normally perform.

Precisely estimated zero. Mathematics treatment coefficient -0.020 (SE 0.049) unadjusted and +0.015 (SE 0.035) adjusted for prior-year school achievement, FDR q = 0.853. Reading -0.051 (SE 0.063) unadjusted and -0.008 (SE 0.043) adjusted, q = 0.853. The author states he can rule out effects larger than 0.07 SD in both subjects. School performance outcomes (pass, fail, repeat, dropout) null. Student effort null, ruling out effects above 4 percentage points, with the one marginal result being treated students 3.7pp MORE likely to be chronically absent. Self-beliefs null on average with NEGATIVE signs in disadvantaged subgroups — increased perceived difficulty of schoolwork among girls and lower self-efficacy among girls and low-income students. No significant moderation by sex, socioeconomic status or prior grade repetition.

Low. School-level randomisation.

A direct failure of the National Study of Learning Mindsets' central claim. Yeager et al. reserve the effect for lower-achieving students in medium-achieving schools; Ganimian tested exactly that population at scale and found nothing, with no moderation by any of the individual characteristics the theory nominates.