The Evidence on Teaching

Explaining Charter School Effectiveness

Angrist JD, Pathak PA, Walters CR · 2013

grade Anatural-experimentindependentunreplicated
Sample
Massachusetts charter lotteries, urban and non-urban schools, pooled with observational estimates for non-lottery schools
Population
Massachusetts charter applicants.
Design
Same lottery machinery as the Boston study, extended statewide so that urban and non-urban charters can be compared under one design. This is what turns 'charters work' into a statement about intake and model rather than about the sector.
Key findings
The same sector, the same state and the same design give +0.34 SD/yr in urban schools and -0.16 in non-urban ones. For an advantaged intake, grade-A evidence for the No-Excuses model is absent to negative. Match the model to the intake; math always moves 2-5x more than reading.
Genetic confound
Lottery-identified within applicants. Urban gains are partly recovery from a weak counterfactual school rather than absolute instructional superiority.
Replication notes
The sharpest demonstration of intake-dependence in the charter literature.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Urban Massachusetts chartersSD/yr+0.34 mathstandardizedper year of attendancebusiness-as-usualend-of-treatmentdomain-skill
Non-urban Massachusetts charters (same design, advantaged intakes)SD/yr-0.16 math — NEGATIVEstandardizedper year of attendancebusiness-as-usualend-of-treatmentdomain-skill
What explains the urban/non-urban gapdecompositionadoption of No-Excuses practices and the achievement level of the intake; gains are larger for LOWER-baseline students and are not peer-drivenstandardizedn/aunclearuncleardomain-skill

Cited by