School start time delays and high school educational outcomes: Evidence from the START/LEARN study

2,153 adolescents (51% male, 49% female; mean age 15 at baseline) across Minneapolis-St Paul high schools

US public high school students; policy-change schools delayed start by 50-65 min, comparison schools held early start times

Difference-in-differences on administrative school records (late arrivals, absences, behaviour referrals, GPA) over 1 baseline year and 2 follow-up years, in the same natural experiment that produced the Widome 2020 actigraphy result. This is the paper that tests achievement specifically, and it is the strongest achievement-side design in the literature. Limits: only 5 schools, GPA is teacher-assigned rather than an independent standardized test, and school-level clustering is coarse.

A 50-65 min delay produced three fewer late arrivals, one fewer absence, a 14% lower probability of a behaviour referral, and a 0.07-0.17 higher GPA in policy-change versus comparison schools. Effects were larger in the second follow-up year than the first, and the absence and GPA differences emerged only in year two — the opposite of the fadeout pattern that dominates the rest of this archive. Note the outcome classes: the large, immediate, robust effects are attendance and behaviour (non-cognitive); the achievement effect is a GPA shift of at most 0.17 grade points on a teacher-assigned measure, with no standardized test result reported.

Low for the design, medium for the interpretation. Bell time is imposed at school level, so allocation is orthogonal to student genotype. But GPA is assigned by teachers who knew the district had changed start times, and the comparison schools are non-randomly different districts; differential migration of families out of or into policy-change schools over three years is not fully ruled out.