Snooze or lose: High school start times and academic achievement

Nationally representative sample from the Child Development Supplement to the Panel Study of Income Dynamics (PSID-CDS), including time-diary data

US high school students, nationally representative

First study of the start-time/achievement relationship on a nationally representative US sample, with time diaries allowing direct measurement of the sleep and time-allocation mechanism. But the identification is longitudinal-with-controls, not a policy shock: start times vary across the schools students happen to attend, and school choice is correlated with family background. Graded C rather than B for that reason, despite the strong data.

Female students attending schools with later start times get more sleep and score higher on reading tests. Male students do NOT get more sleep when their schools start later, and their test scores do not change. The sex asymmetry is the finding worth carrying: the achievement effect appears exactly where the sleep effect appears and is absent exactly where the sleep effect is absent, which is what a genuine sleep-mediated mechanism should look like — but it also means the average effect over a mixed-sex population is roughly half of the female-only estimate.

Medium-high. Students are not assigned to start times by any exogenous process here; families sort into districts and schools, and the same parental characteristics that produce early- or late-starting schools also produce heritable ability differences. The time-diary mediation is informative about mechanism but does not fix identification.