Sleep Tight, Act Right? The Effect of Sleep on Student Behavior

• In Event: Implementing School-Based Mental Health Services to Support Students

Abstract

Ample evidence underscores the role of sleep in supporting children’s cognitive, behavioral, and emotional development. Insufficient sleep is consistently linked to poorer academic performance, increased behavioral problems, and compromised emotional regulation for children and youth (e.g., Astill et al., 2012; Dewald et al., 2010; Heissel & Norris, 2018; Sadeh et al., 2002). Several studies have documented the negative impact of less sleep due to exogenous variation in sunset times (Jagnani, 2024), changes due to daylight savings time (DST; Gaski & Sagarin, 2011), or variation in school start times (Carrell et al., 2011; Groen & Pabilonia, 2019) on student outcomes, either outside the US or in specific states with time zone variation. Although DST and its impact on several outcomes have been studied extensively, few studies have estimated the effect of contextual factors related to sleep on student behavior outcomes both nationally and across seasons. This gap limits our understanding of the broader and potentially compounding effects of time policies on youth outcomes.

We provide the first national-level estimates regarding the impact of sleep on student behavior, as proxied by in- and out-of-school suspension rates. These estimates are relevant to understanding how structures impacting student sleep (e.g., school start times, DST, time zone borders (TZBs)) may affect students’ likelihood of suspension, which has downstream, negative impacts across a juvenile’s life course (Bacher-Hicks et al., 2025; Sorensen et al., 2022; Davison et al., 2021; Welsh & Little, 2018). 

To conduct our estimation, we use a sharp border regression discontinuity design, leveraging variation in sunrise/sunset and clock times around TZBs across the United States. Our key variable is the distance from a school to a TZB, with our key comparison being schools that are within a specified distance (~100km) on each side of a TZB. We find a sharp discontinuity in our first stage — being just to the east of a TZB means sunset time is approximately 1 hour later than schools just to the west of the border. Other research that uses this identification strategy finds that this translates to a decrease in 19 minutes of sleep for adults (Giuntella & Mazzonna, 2019). 

We merge sunset times with bi-annual school-level suspension data from the Civil Rights Data Collection. Results show that schools just east of the border (and, thus, where students likely sleep less) have, on average, a 40%-off-the-baseline increase in out-of-school suspensions, driven by male and middle grade students. Results are robust to modeling changes in both magnitude and direction after inclusion of state fixed effects (which restricts our identifying variation to 11 states within multiple time zones) and other sensitivity analyses (e.g., by estimating the impact of a shifted time zone boundary, and re-estimating models after shifting bandwidth selection and leveraging different optimization routines) — though they often become less precise. These analyses suggest our primary results are driven by the time/sleep variation induced by the TZB. Our findings reveal another avenue in which external, yet malleable, context characteristics beyond students' control impact their educational outcomes.

Authors

• Berta Bartoli, New York University
  Presenting Author

• Jo R. King, Boston University
  Non-Presenting Co-Author