High school start times and student achievement: Looking beyond test scores

https://doi.org/10.1016/j.econedurev.2020.101975Get rights and content

Abstract

The American Academy of Pediatrics recommends that U.S. secondary schools begin after 8:30 a.m. to better align with the circadian rhythms of adolescents. Yet due to economic and logistic considerations, the vast majority of high schools begin the school day considerably earlier. We leverage a quasi-natural experiment in which five comprehensive high schools in one of the nation’s largest school systems moved start times forty minutes earlier to better coordinate with earlier-start high schools. Here, disruption effects should exacerbate any harmful consequences. We report on the effect of earlier start times on a broad range of outcomes, including mandatory ACT test scores, absenteeism, on-time progress in high school, and college-going. While we fail to find evidence of harmful effects on test scores, we do see a rise in absenteeism and tardiness rates, as well as higher rates of dropping out of high school. These results suggest that the harmful effects of early start times may not be well captured by considering test scores alone.

Introduction

School districts in the United States typically set school start times based on financial and logistical constraints related to transportation, athletics, and the labor market dynamics of both parents and students. On average, high and middle school students nationwide start school just after 8:00 a.m., while elementary school students start 10–15 minutes later (USDOE, 2016). However, earlier start times may be harmful to adolescents, whose circadian rhythms differ from those of younger children. Recognizing this, in 2014 the American Academy of Pediatrics (AAP) recommended that schools set bell schedules no earlier than 8:30 a.m. to improve the “physical and mental health, safety, and academic achievement” of older students (Owens, Au, Carskadon, Millman, and Wolfson, 2014b, p. 646). In response to this recommendation and new evidence linking later start times to better health and academic outcomes (e.g., Bowers, Moyer, 2017, Marx, Tanner-Smith, Davison, Ufholz, Freeman, Shankar, Newton, Brown, Parpia, Cozma, Hendrikx, 2017), some school districts have implemented later start times (Owens, Drobnich, Baylor, & Lewin, 2014a). In the most notable start time policy change to date since the AAP’s recommendation, California became the first state in the nation to require high schools to start no earlier than 8:30 a.m. (Luna, 2019).1

A growing body of evidence suggests that later start times indeed translate into more sleep for adolescent students (e.g., Bowers and Moyer, 2017), but it remains unclear whether delaying the bell impacts student performance. It is difficult to estimate the causal effects of school start times because the logistical and financial factors determining start times may be correlated with family background or socioeconomic status of students. Schools consider the social needs of the populations they serve (e.g., popular after-school activities or early morning childcare for working parents) and often schedule around morning commutes to avoid congestion on roads or public transportation systems. All else equal, moving a school starting time later in the morning has clear costs in terms of logistics, but measuring the net expected benefit of doing this is complicated by confounding factors.

In this study, we report the effects of a significant policy change in a large school district that provides plausibly exogenous variation in school starting times. In the 2012-13 school year, the Wake County Public School System (“Wake County”), the 15th largest school system in the U.S. (NCES, 2018), shifted start times forty minutes earlier in five large, comprehensive high schools to better coordinate with high schools already starting at 7:25 a.m. The decision to move the bell earlier was motivated by logistical concerns, but such a move could plausibly reduce students’ sleep duration2 and result in decreased school performance. In our setting, the “late” start time was 8:05 a.m., which is 25 minutes earlier than the AAP recommendation. While moving start times 40 minutes earlier is predicted to have harmful consequences for students, it may understate the true impacts of early start times as measured against the 8:30 a.m. standard. We take advantage of this exogenous policy shock to measure the impact of this large start-time shift on both cognitive and non-cognitive outcomes in the short- and medium-term using as our control group those high schools that did not shift start times.

Our results leverage detailed, longitudinal administrative data that tracks multiple cohorts of students over time to estimate the effects of these earlier start times. We identify our main estimates in a difference-in-differences framework. We include a range of robustness tests, including models with student fixed effects, and we explore heterogeneity within the results. To preview our results, we fail to find evidence that variation in start times affected scores on the ACT exam, a college-readiness test required for all students in the district. However, using several measures of student engagement, we do find evidence that absenteeism, tardiness, and dropout rates all increased under earlier start times. Across prior achievement terciles, we find that student engagement effects are significant for those in the bottom two terciles of baseline achievement. Our study suggests not only that earlier start times can increase absenteeism rates and reduce student engagement, but that studies focusing solely on test scores have the potential to underestimate the non-cognitive consequences of early start times.

This study makes several contributions to the literature on start times and adolescent well-being. First, in contrast to much of the existing work that relies on indirect identification, our study is among the few that examines the effects of policy-relevant bell schedule changes.3 Second, our rich dataset allows us to estimate impacts on both cognitive and non-cognitive outcomes. Recent work has demonstrated the importance of non-cognitive skill development on the longer-term outcomes of children (e.g., Jackson, 2018, Petek, Pope, 2018).4 Our findings suggest that student engagement and non-cognitive skill formation might be harmed by earlier start times, even when test scores do not fall appreciably. Third, while existing evidence typically reports test-score based achievement, we also estimate start time effects on high school graduation and college enrollment. Fourth, our unique setting in which school begins even earlier allows us to estimate “worst-case scenario” effects, whereby any disruption due to bell schedule changes will exacerbate the negative effects of earlier start times. Finally, our large sample size allows us to explore heterogeneity. With roughly 10,000 students enrolled in each of the eight cohorts, we investigate heterogeneous impacts by demographic characteristics and prior achievement terciles.

This paper proceeds as follows. Section 2 describes the related literature on school start times. Section 3 details start times in Wake County and the policy context around the shift. Section 4 summarizes our data and method. Section 5 describes our results, heterogeneous effects, and robustness checks. Section 6 reports longer-term outcomes. Section 7 concludes.

Section snippets

Background

Despite strong predictions about the benefits of delaying school start times to enable sufficient sleep,5

Start times in Wake County

School start times in Wake County and elsewhere largely depend on transportation logistics (Edwards, 2012, Fügenschuh, 2009, Hafner, Stepanek, Troxel, 2017, Jacob, Rockoff, 2011). Secondary schools—high schools, in particular—typically begin at the earliest tier so that school buses can be parked near the high school and fan out in relative concentric patterns throughout the morning in order to transport younger students. By using a small number of drivers to accomplish this, districts can

Data

We construct our panel using Wake County administrative data spanning the period from 2008-09 to 2018-19.9 We group students into cohorts based on their projected year of graduation (i.e., the class of 2014 would enter 9th grade in 2010-11). We define the school a student attends

Cognitive outcomes

Table 2 presents estimates of Eq. (1) where ACT test scores are the outcome of interest. The estimates in Column (1) regress the ACT combined score on treatment, while Columns (2)-(5) regress subject-level ACT test scores on treatment. The difference-in-differences estimates, β3 and β5 in Eq. (1), are presented at the top of the table. The interaction term “Treated  ×  GradYr2014-2016,” β3, captures the difference between the treatment and control groups when the treatment group was

High school graduation and postsecondary outcomes

Finally, we consider whether declines in non-cognitive outcomes translate into measurable differences in longer-term outcomes, namely on-time graduation and college enrollment. Table 7 reports impacts on high school attainment and a range of college-going outcomes. The cohorts are parallel to Table 3, Column (5), for dropout rates, except that now the class of 2013 is grouped with cohorts experiencing the bell schedule change (“GradYr2013-2016”) to account for the bell schedule change during

Conclusions

We show that in one of the largest school systems in the nation, moving high school start times earlier had no effect on cognitive outcomes, but did adversely impact important measures of student engagement. These findings contribute to a growing literature that suggests non-test score outcomes may yield important insights into students’ experiences at school (Jackson, 2018, Petek, Pope, 2018). Skills such as executive function, grit, and perseverance may not be well-reflected in test scores

CRediT authorship contribution statement

Matthew Lenard: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration. Melinda Sandler Morrill: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration. John Westall: Project administration,

References (66)

  • ACT

    Understanding your scores - The ACT test

    (2020)
  • A. Adan et al.

    Gender differences in Morningness-Eveningness preference

    Chronobiology International

    (2002)
  • E.P. Bettinger et al.

    Improving college performance and retention the easy way: Unpacking the ACT exam

    American Economic Journal: Economic Policy

    (2013)
  • H.S. Bloom

    Minimum detectable effects: A simple way to report the statistical power of experimental designs

    Evaluation Review

    (1995)
  • BLS

    Usual weekly earnings of wage and salary workers second quarter 2017

    Technical Report

    (2017)
  • J. Boergers et al.

    Later school start time is associated with improved sleep and daytime functioning in adolescents:

    Journal of Developmental & Behavioral Pediatrics

    (2014)
  • L. Bonner

    ACT test scores are stagnant in NC

    News and Observer

    (2015)
  • J.M. Bowers et al.

    Effects of school start time on students’ sleep duration, daytime sleepiness, and attendance: A meta-analysis

    Sleep Health

    (2017)
  • D. Carlson et al.

    The effect of english language learner reclassification on student ACT scores, high school graduation, and postsecondary enrollment: Regression discontinuity evidence from Wisconsin

    Journal of Policy Analysis and Management

    (2016)
  • S.E. Carrell et al.

    A’s from Zzzz’s? The causal effect of school start time on the academic achievement of adolescents

    American Economic Journal: Economic Policy

    (2011)
  • N.Y. Chan et al.

    Impact of a modest delay in school start time in Hong Kong school adolescents

    Sleep Medicine

    (2017)
  • F. Danner et al.

    Adolescent sleep, school start times, and teen motor vehicle crashes

    Journal of Clinical Sleep Medicine

    (2008)
  • T.S. Dee et al.

    The impact of no Child Left Behind on student achievement

    Journal of Policy Analysis and Management

    (2011)
  • J.F. Dewald et al.

    The influence of sleep quality, sleep duration and sleepiness on school performance in children and adolescents: A meta-analytic review

    Sleep Medicine Reviews

    (2010)
  • A.K. Dills et al.

    Course scheduling and academic performance

    Economics of Education Review

    (2008)
  • F. Edwards

    Early to rise? The effect of daily start times on academic performance

    Economics of Education Review

    (2012)
  • J.A. Groen et al.

    Snooze or lose: High school start times and academic achievement

    Economics of Education Review

    (2019)
  • M. Hafner et al.

    Later school start times in the U.S.: An economic analysis

    (2017)
  • R. Halvorsen et al.

    The interpretation of dummy variables in semilogarithmic equations

    The American Economic Review

    (1980)
  • M. Hansen et al.

    The impact of school daily schedule on adolescent sleep

    Pediatrics

    (2005)
  • J.A. Heissel et al.

    Rise and shine: The effect of school start times on academic performance from childhood through puberty

    Journal of Human Resources

    (2018)
  • P. Hinrichs

    When the bell tolls: The effects of school starting times on academic achievement

    Education Finance and Policy

    (2011)
  • C.K. Jackson

    What do test scores miss? The importance of teacher effects on non–test score outcomes

    Journal of Political Economy

    (2018)
  • Cited by (14)

    • School start times and academic achievement - A systematic review on grades and test scores

      2022, Sleep Medicine Reviews
      Citation Excerpt :

      Course grades and standardised scores most likely reflect different aspects of students’ learning and knowledge and could thus be differentially sensitive to SST changes. However, when grouping studies based on outcome measure, there was no tendency or differential results on either 1) objective test scores (two positive [54,65], three null findings [53,55,56], and four mixed results [51,57,58,63]), 2) objective grades (two positive [45,59], four null findings [46,52,56,61] and two mixed results [58,60]), 3) or self-reported grades (two positive [50,66], three null findings [48,49,62], and one each for mixed [51] and negative [64]). To judge the evidence quality of the included studies, we performed a risk of bias assessment (Fig. 3).

    View all citing articles on Scopus

    Acknowledgements: We thank the Wake County Public School System for its partnership, especially Brad McMillen, Cathy Moore, Colleen Paeplow, Sonya Stephens, and members of the district’s communications, student assignment, and transportation offices. Co-author Lenard is a former member of the district’s Data, Research and Accountability department. We thank Maria Fitzpatrick, Michelle Marcus, Sabrina Wulff Pabilonia, and seminar participants at the Southern Economic Association Annual Meeting for helpful comments and discussions. All remaining errors are our own. This paper was previously circulated under the working title: “School Start Times and Student Achievement: Evidence from High Schools.”

    View full text