Less Sleep, More Problems? Evidence from Lottery-Assigned School Start Times

Argentine middle school students lottery-assigned to morning or afternoon shifts for up to three years (SSRN working paper; sample size not stated in the deposited abstract)

Middle school students in Argentina's double-shift school system, assigned by lottery to a 7:30 a.m. or a 1:15 p.m. start; described as an academically selected population

The only true lottery design in this literature. Argentine double-shift schools randomly assign admitted students to the morning or afternoon shift, producing a six-hour exogenous difference in start time sustained for up to three years. This is exactly the design class the archive privileges — allocation is random, so chronotype, sleep-need, and family background are balanced across arms by construction. Graded B rather than A because it is an unpublished working paper that has not been peer-reviewed or replicated, and because only the deposited abstract was accessible (SSRN returned HTTP 403; metadata and abstract retrieved via the Crossref API). The reported outcomes are sleep and behaviour, NOT test scores.

Boys assigned to the afternoon shift report sleeping about one hour more than boys assigned to the morning shift, and the prevalence of short sleep falls sharply. The sleep difference DISAPPEARS on nights before non-school days — the internal falsification test, confirming institutional timing is the binding constraint rather than any difference between the groups. Girls largely offset the early schedule through earlier bedtimes and napping, leaving total daily sleep roughly unchanged. Afternoon assignment reduced boys' aggressive or delinquent behaviour by about 30 percentage points; girls' behavioural responses did not mirror this. The paper explicitly frames its contribution as consequences "beyond test scores" and does not supply an achievement estimate.

Lowest in the topic. Lottery assignment makes genotype, chronotype distribution, and family background identical in expectation across arms; the placebo on non-school nights closes the remaining gap. This is the design that shows a bell time is a causal constraint on adolescent sleep rather than a marker of the families who choose it. Caveat on external validity, not on identification: the population is academically selected and the contrast is six hours, far larger than any US bell-time policy.