The Evidence on Teaching

Achievement Effects of Four Early Elementary Math Curricula (IES cluster RCT)

Agodini, R., Harris, B., Seftor, N., Remillard, J., & Thomas, M. (Mathematica for IES) · 2013

grade Brctindependentmixednumbers spot-checked
Sample
111 schools randomized; 58 schools (2,045 students) in the 2-year analysis
Population
Grades 1-2, 12 districts, 10 states; Title I-heavy volunteer sample.
Design
The only multi-curriculum RCT in US elementary math: schools randomized to Saxon (scripted), Math Expressions (blended), Investigations (constructivist), or SFAW/enVision (basal). Independent ECLS-K outcome. Caveats: 2-year result from the 58 voluntarily-continuing schools; Saxon grade-2 teachers spent ~1 extra hr/week on math.
Key findings
The cleanest causal evidence in the math wars: the only stable loser across two years is the most strongly constructivist, minimal-guidance curriculum (Investigations), trailing ALL three rivals by 0.22 SD on an independent test. The winner set is heterogeneous (scripted, blended, conventional) — the result indicts the low-guidance pole rather than crowning scripting.
Genetic confound
Random assignment removes selection; curriculum effects on domain skills.
Replication notes
Numbers verified from NCEE 2013-4019. Direction converges with Indiana matching (Bhatt-Koedel); post-CCSS multi-state data (Blazar 2020) show curriculum differences largely vanished — era-dependent.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
After 2 years: each rival vs InvestigationsSD+0.22 (69.4 vs 65.5, SD=18; ~9 percentiles); winners indistinguishable from each otherstandardizedend of year 2active-alternativeend-of-treatmentdomain-skill
Year 1: Math Expressions/Saxon vs InvestigationsSD~+0.15-0.18 (first cohort up to +0.24-0.30)standardizedend of year 1active-alternativeend-of-treatmentdomain-skill
Curriculum switching (SFAW year1 -> Saxon year2)SD+0.35 (significant); other switches nullstandardizedend of year 2active-alternativeend-of-treatmentdomain-skill

Cited by