Algebra timing — acceleration mandates, delay mandates, and readiness-based placement

Verdict

The algebra-timing literature is the purest demonstration in the database that sequencing mandates fail in both directions because they ignore the ability distribution — and that the constructive alternative is measurement plus dosage.

Universal acceleration harms. The best-identified study (Clotfelter/Ladd/Vigdor): pushing median students into 8th-grade algebra cut Algebra I scores 0.32 SD and on-time geometry passage 10pp; low performers pushed into 9th-grade algebra lost 0.5 SD. Never positive at any ability level; verified in a second district; the district reversed itself. California's algebra-for-all converges system-wide (Domina); Chicago's 1997 college-prep-for-all raised failures and moved nothing (Allensworth) — and cost high-skill students, who lost ground as abolishing remedial tracks mixed the classrooms. The OLS-vs-IV contrast is the textbook selection lesson: early algebra-takers look better (+0.18) because able students select in; the causal effect is negative.

Universal delay also fails. San Francisco's delay-of-algebra-for-all (Huffaker/Novicoff/Dee): AP-math participation down ~15%, ethnoracial gaps untouched, affluent families bought workarounds, the board reversed course. No achievement gains ever demonstrated.

What works — two things:

1. Targeted acceleration by objective readiness scores (Wake County RD): no test-score harm, +30pp taking-and-passing 9th-grade geometry for marginal students — and objective criteria eliminated race/income placement gaps conditional on skill (the blind-audition effect).
2. Double-dose algebra — more time at the right level for below-median students (Cortes/Goodman/Nomi): +0.18–0.24 SD on independent ACT-series tests and, 12 years later, +3.3pp BA attainment on a 7% base — one of the largest credential effects of any school-level intervention, at the cost of one elective period. But the 2004-cohort null (when cut-score adherence slipped) shows it is implementation-fragile. The short-run evaluation (Nomi & Allensworth 2009) found B-or-better rates +9.4pp with the smallest gains for the very lowest-skill students — a readiness floor even here — and the mechanism paper (Nomi & Allensworth 2013) shows why the policy simultaneously raised test scores and course failures: assigning below-median students a second period silently regrouped classrooms, dropping average peer skill ~0.16 SD below the cutoff. Any targeted-support policy that groups by need is running both mechanisms at once.

Hereditarian-lens assessment

Risk: low (natural experiments and RD throughout). This is the lens's showcase topic: every finding is a prior-ability × treatment interaction. Mandates fail because course labels can't override readiness; objective ability-based placement improves both efficiency and equity (measuring ability better beat ignoring it); intensification works but hits a readiness floor (the very-low-skilled benefit least even from double time). And the attainment result cuts against fatalism: instruction never moved g, but it durably moved credentials and life outcomes — test- score effects understated the long-run value ~4×.

Practical guidance

• Placement by objective measured readiness, reviewed each year — no blanket grade-level rules, no subjective teacher gatekeeping (which reintroduces demographic gaps).
• Accelerate students above the readiness bar (see also the grouping/acceleration structure topic); expect lower grades in harder courses but no achievement harm.
• Double-dose, don't decelerate, students below the bar: a second scheduled period of the same course with problem-solving/verbal emphasis, in place of an elective.
• Guard the implementation details: honor the cut score, and don't let support classes become bottom-tracked ghettos (peer-composition drift killed the 2004 cohort's gains).