Accountability and Flexibility in Public Schools: Evidence from Boston's Charters and Pilots

Boston charter and pilot school admission lotteries, middle and high school

Boston charter-school applicants (lottery-identified), plus applicants to Boston's in-district 'pilot' schools as a contrast.

Oversubscription lotteries remove applicant selection entirely; estimates are LATEs for applicants to oversubscribed schools, which is the population a founder actually faces. Pilot schools (union-staffed, in-district, autonomous) are included as a same-city comparison, which is what isolates the model from the autonomy.

Boston's 'No Excuses' charters produce the largest replicated school effects in US data (~0.4 SD/yr in middle-school math). The pilot-school null in the same city is the load-bearing contrast: it is the instructional model, not governance autonomy, that moves scores.

Lotteries are clean within applicants. Raw charter-versus-district gaps are mostly selection; these estimates are not.

The most replicated large causal school effect in US data; extended by Angrist, Pathak & Walters 2013 to the whole of Massachusetts.