The Evidence on Teaching

Intensive Math Instruction and Educational Attainment: Long-Run Impacts of Double-Dose Algebra (with the PNAS 12-year follow-up)

Cortes, K. E., Goodman, J. S., & Nomi, T. (2015); Nomi, T., Raudenbush, S. W., & Smith, J. J. (2021, PNAS) · 2015

grade Bnatural-experimentindependentmixednumbers spot-checked
Sample
RD at the 50th-percentile cutoff; N=40,151 full / 9,700 RD sample; PNAS: 33,033
Population
CPS below-median 9th graders, 2003-04 cohorts (92% Black/Hispanic, ~90% low-income).
Design
Regression discontinuity; independent standardized outcomes (PLAN/ACT); placebo-clean. Double period of algebra (support class in place of an elective), problem-solving/verbal emphasis.
Key findings
The standout positive in the algebra literature: doubling instructional TIME at the right level for below-median freshmen produced real gains on independent tests AND huge attainment effects (+3.3pp BA on a 7% base, 12 years out) — largest for weak readers, at near-zero marginal cost (an elective period). But the 2004-cohort null shows the effect is implementation-fragile: ability-matched assignment and tolerable peer composition are decisive.
Genetic confound
LATE at the median; very-low-skill students benefit least (readiness floor). Effects survived assignment to much weaker peer groups.
Replication notes
Numbers verified. Within-district replication: works for 2003, fails for 2004 as implementation drifted. No out-of-district causal replication despite wide adoption.
DOI / URL
10.1353/jhr.2015.0004 ; 10.1073/pnas.2019030118

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Math achievement (fall-11th PLAN / ACT)SD+0.235 / +0.180 (attrition-robust bounds 0.08-0.14, all but one significant)standardized+2 to +2.5yrbusiness-as-usualover-2yrdomain-skill
High school graduationpp+9.8 (4-yr) / +12.1 (5-yr); college enrollment +10.8standardized+4-5yrbusiness-as-usualover-2yrattainment
BA attainment, 2003 cohort (PNAS)pp+3.3 on a 7.0% base (~47% relative) at 12 yearsstandardized+12yrbusiness-as-usualadulthoodattainment
2004 cohort (implementation drift)ppall college effects NULL — cut-score adherence fell 0.54->0.31 and peer sorting worsenedstandardized+12yrbusiness-as-usualadulthoodattainment

Cited by