Choices for studying choice: Assessing charter school effectiveness using two quasi-experimental methods

Two quasi-experimental methods – fixed effects (FE) and virtual control records (VCR) – were used to measure charter schooling in 14 states and two districts. The new VCR method uses all available observable charter student characteristics and prior performance to create a composite comparison record. A head-to-head comparison of the FE and VCR methods used the same charter students to test the FE control (e.g., the charter student's own traditional public school experience) and the VCR for equivalence. The comparison produced highly similar estimates; charter coefficients were identical in sign and significance and of the same general magnitudes. In an analysis of the sampling fractions included in each method using all available tested charter students, the VCR method was found to produce more generalizable results. In the policy analysis, charter school quality was found to be demographically and geographically uneven with only 19 percent of charter schools outperforming their local markets.

► Fixed effects and virtual control records used to measure charter schooling.
► Direct comparison of methods with same charter students yields similar estimates.
► Virtual controls minimizes selection bias, more generalizable than fixed effects.
► Charter school quality is uneven across student demography and geography.
► Only 19 percent of charter schools outperform local markets in math and reading.