Early to rise? The effect of daily start times on academic performance
Edwards F · 2012
grade Bquasi-experimentindependentunreplicated
Sample
Wake County (NC) Public School System administrative data 2000-2006, 146,687 students district-wide; analysis focuses on middle school; 14 middle schools changed start times, 7 by 30+ minutes
Population
US middle school students (grades 6-8), Wake County North Carolina, with follow-up into grade 10
Design
Three nested identification strategies of increasing credibility: (1) cross-school comparison with controls, (2) within-school over time as start times changed, (3) within-STUDENT across years when the student's school changed its bell time. The student fixed-effects specification is the credible one and it is what the headline estimate rests on. Bus-scheduling logistics, not school characteristics, drive the start-time changes, which is what makes the shock plausibly exogenous. NOTE: the journal full text is paywalled; effect sizes here are taken from the author's own summary of the same study in Education Next (https://www.educationnext.org/do-schools-begin-too-early/), fetched directly.
Key findings
A one-hour later start time raises math scores by 1.8-2.2 percentile points and reading scores by 1.0-1.5 percentile points, with 1.8 and 1.0 the most conservative (student-fixed-effect) estimates. Edwards reports percentile points, not standard deviations; converting near the median gives roughly 0.05-0.07 SD, which is our arithmetic and not his. Effects persist into 10th grade and increase with student age, consistent with the adolescent-circadian mechanism. The effect is more than twice as large for students in the bottom third of the test-score distribution as for the top third. Mechanism evidence: students starting an hour later watched 12 fewer minutes of TV per day, spent 9 more minutes per week on homework, and had 1.3 fewer absences.
Genetic confound
Low. The within-student specification compares the same child to himself before and after a bus-driven bell-time change, which nets out every stable genetic and family influence by construction. This is the within-person design the archive privileges. Residual risk is time-varying shocks correlated with the scheduling change and differential attrition from the district.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Math test scores (achievement, standardized) | percentile points per one-hour later start, student fixed effects | +1.8 to +2.2 percentile points (conservative estimate 1.8) | standardized | same school year | business-as-usual | end-of-treatment | domain-skill |
| Reading test scores (achievement, standardized) | percentile points per one-hour later start | +1.0 to +1.5 percentile points (conservative estimate 1.0) | standardized | same school year | business-as-usual | end-of-treatment | domain-skill |
| Persistence of achievement effect | follow-up regression | effects persisted into 10th grade | standardized | several years after exposure | business-as-usual | over-2yr | domain-skill |
| Heterogeneity by prior achievement | subgroup estimates | effect more than 2x larger for bottom-third than top-third students | standardized | same school year | business-as-usual | end-of-treatment | domain-skill |
| Time use and attendance (non-cognitive/behavioural) | per one-hour later start | -12 min/day TV; +9 min/week homework; 1.3 fewer absences | mixed | same school year | business-as-usual | end-of-treatment | behaviour |
Cited by
- Do later school start times increase adolescent sleep, and does that raise achievement?strong supportconf: mediumgc: low