Robust standard errors in brackets. Standard error clustered at the district level.

+ significant at 10 percent;

↵* significant at 5 percent; ** significant at 1 percent. All regressions are based on a panel of 57 schools. Each coefficient estimate is the result of a different regression. Each column represents the same specification, while each row indicates a different dependent variable. Demographic controls are log school enrollment, percent white, percent black, percent Hispanic, percent Asian, percent Native American, percent free lunch, percent LEP, and the log of mobile students. Specification 1 includes school controls and affected cohort size (Grade 12, Grade 11, lagged Grade 11 and lagged Grade 10 enrollment for AP outcomes, and Grade 12 and lagged Grade 11 enrollment for non-AP outcomes based on the cohort of high school graduates). Specification 2 includes all variables in Specification 1 and year-by-propensity score decile fixed effects (the preferred model). Specification 2* includes all variables in Specification 2 with the exception of 11th and Grade 12 enrollment for AP outcomes and omit Grade 12 enrollment for the non-AP outcomes based on the cohort of high school graduates. Specification 2** includes all variables in Specification 2 but excludes all grade enrollment variables. Placebo treatment uses the second lead of actual adoption as the treatment. This variable denotes two years before actual APIP adoption.