Urban Charter School Study Report on 41 Regions

41 urban regions across 22 states

Urban charter-school students matched to traditional-public-school students.

Matched-observational 'virtual control record' design: each charter student is compared to a synthetic twin built from demographically similar students in the schools charter students left. Not a lottery. Graded C: the matching is on observed demographics and prior scores only, so selection on unobserved (and partly heritable) student and family traits is not addressed. Widely criticised on exactly this point. No DOI.

The broadest description of the urban charter sector: on average urban charters outscore matched traditional-school peers, most for disadvantaged subgroups. Its value here is coverage (41 regions) rather than identification — where lotteries exist, use them.

HIGH for this design. Matching on demographics and prior scores does not remove selection on family traits that are both heritable and predictive of choosing a charter. Read as descriptive coverage of the sector, not as a causal estimate — the lottery studies are what carry causal weight here.

Directionally consistent with the urban lottery estimates, which is the main reason to keep it; the magnitudes are not independently trustworthy.