California's 1996 class-size reduction at scale (CSR Consortium; Jepsen & Rivkin)
CSR Research Consortium (2002); Jepsen, C., & Rivkin, S. (2009) · 2009
grade Bquasi-experimentindependentunreplicated
Sample
Statewide (>$1B/yr); K-3 classes ~29 -> ~20
Population
California K-3.
Design
Non-random overnight rollout; Jepsen-Rivkin exploit within-school enrollment shocks.
Key findings
Class size and teacher quality compete for the same dollars and the same labor pool: statewide CSR diluted teacher quality exactly where STAR predicted the biggest gains, flattening the equity benefit. Never lower class size by lowering the hiring bar.
Genetic confound
Within-school enrollment shocks plausibly exogenous.
Replication notes
The definitive scale-up cautionary tale.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Pure class-size channel | ES | real but modest gains in 3rd-grade math/reading... | standardized | annual | business-as-usual | end-of-treatment | domain-skill |
| Equilibrium teacher-quality response | offset | ...largely offset by forced mass hiring of inexperienced/uncertified teachers, CONCENTRATED in poor/minority schools | standardized | annual | business-as-usual | end-of-treatment | domain-skill |
Cited by
- Class-size reductionmixedconf: highgc: low