School timing, sleep, and adolescent behavior: Evidence from lottery-assigned start times

Double-shift school systems assign some adolescents to early-morning and others to afternoon schooling, yet causal evidence on the consequences of such large timing differences remains limited. I study a selective public middle school in Argentina, where the outcome of a public lottery determines a 7:30 a.m. or 1:15 p.m. start time for up to three years. Unlike the relatively modest schedule changes studied in most prior work, the lottery shifts the school day by nearly six hours while holding curriculum, facilities, and school-day length fixed, identifying reduced-form effects of school timing. Afternoon schooling increases self-reported night sleep by nearly an hour and reduces the share sleeping at most seven hours by 34 percentage points from a 65% baseline. Boys also gain total daily sleep, whereas morning-shift girls offset their shorter nights with longer naps, leaving total daily sleep unchanged. These differences largely disappear before non-school days. Behavioral responses diverge sharply by gender: afternoon assignment reduces boys’ aggressive or delinquent behavior, especially vandalism, but increases girls’ substance use initiation. Together with changes in extracurricular and social activities, these contrasting responses caution against interpreting school timing through a single channel such as sleep. In double-shift systems, school timing is thus a behavioral design choice, not merely a capacity tool.