Aiming High and Falling Short: California's Eighth-Grade Algebra-for-All Effort

222 CA unified districts (balanced panel), 2003-04 to 2009-10; ~280,000-300,000 8th graders per year (~2M student-years)

California 8th graders in unified school districts, 2003-2010; outcome is district-mean 10th-grade CAHSEE math two years later.

District-level panel. Two specifications: district fixed effects with year fixed effects and demographic controls (adj R2=.40, 220 districts), and a random-growth model adding district-specific linear time trends (adj R2=.04, 213 districts). Identification is off within-district year-to-year variation in the share of 8th graders taking the Algebra I or higher end-of-course exam — district uptake of the state's algebra-for-all push, which was uneven and never fully mandated (the courts postponed implementation in spring 2010 and California then adopted Common Core, which recommends pre-algebra in grade 8). Not a clean shock and not a discontinuity, so grade C rather than B is right: the authors themselves note internal validity "hinges on the assumption that the only time-varying within-district characteristics that systematically covary with both eighth-grade Algebra placement patterns and CAHSEE math scores are captured by our controls." Robustness: results are unchanged excluding Los Angeles Unified or any other single large district, are linear in the size of the enrollment change, and show no significant pre/post-mandate interaction.

System-level evidence that pushing more 8th graders into algebra lowered downstream achievement — negative on all five CAHSEE content domains including basic number sense, so it is not an artifact of the test missing advanced content. But "across the board" applies to content, not to districts: the effect is concentrated entirely in large districts (>=850 8th graders) and is a precise zero in small and mid-sized ones. The authors' explanation is organizational, not pedagogical — reshuffling course placements is a bureaucratic shock that large districts absorb badly. The marginal-student story is therefore about system capacity, and the randomized evidence (Heppen 2012) for individually well-prepared students points the other way.

District-level design with district fixed effects, so student selection into algebra cannot drive the estimate; the estimand is a system-level placement-policy effect, not a student-level course effect. Residual threat is time-varying within-district confounders, which the random-growth specification is designed to absorb and which shrinks the estimate from -0.07 to -0.05.

Sign converges with Clotfelter et al. (2012) IV estimates in North Carolina (negative) and with Allensworth et al. (2009) in Chicago (no improvement in achievement, graduation or college-going from universal 9th-grade algebra), though Chicago's "double dose" helped low achievers (Nomi & Allensworth 2009) while hurting high achievers (Nomi 2012). It does NOT converge with Heppen et al. (2012), a randomized trial in rural New England middle schools that found positive effects of enrolling relatively high-achieving students in 8th-grade algebra — a contradiction the authors address directly by arguing the system-level and student-level estimands differ. California's mandate was never fully implemented: the courts postponed it in spring 2010 and state policy no longer requires accelerated algebra.