Abstract:
Enrollment has a direct and profound impact on school funding. Beginning with the 2019-20 school year, a Midwestern urban school district, already experiencing enrollment pressures due to competition with nearby districts, charters, and private schools, pushed high school start times back an hour to better align with research suggesting a later start for adolescent learners encourages academic gains. Funding and transportation constraints required a corresponding shift to an earlier morning start for elementary schools. We study the impact of the start time change on enrollment trends using a panel of school-grade level data and a difference-in-differences methodology that employs school and year fixed effects. We find that the start time policy is associated with declines in enrollment for primary grades on the order of 4-6 students per grade — with the exception of kindergarten, where declines average 7-8 students. A series of falsification tests suggest the policy negatively impacted elementary enrollment.

Introduction or Background
The American Academy of Pediatrics (AAP) and the Centers for Disease Control (CDC) recommend that adolescents get at least 8.5-9.5 hours of sleep per night to promote well-being and academic success (Adolescent Sleep Working Group et al. 2014; Centers for Disease Control and Prevention, n.d.). In 2015, the CDC released results of the previous year’s School Health Policies and Practices Study. According to this report, 93% of high schools and 83% of middle schools throughout the country started classes before 8:30 a.m. Given circadian rhythms of middle age and adolescent children, experts suggest that such early start times might contribute to poor academic and health outcomes (Wahlstrom et al. 2014) and recommend that schools revisit these policies. In response, districts throughout the country have begun to explore and implement later start times for older children.

A now robust body of literature examines how academic performance and well-being correlate with delayed start times. Most of the research on delays in adolescent school start times suggests a positive association. In 2011, Carrell, Maghakian, and West used changes in the daily schedule at the U.S. Air Force Academy along with randomized placement of freshmen students and instructors in morning courses to examine the causal effect of school start time on achievement. They found that a 50-minute delay in class start time had a significant positive impact on student achievement roughly equivalent to one standard deviation increase in teacher quality. In 2014, Wahlstrom, et al. published the results of a three-year research study on school start times covering 9,000 students from eight public schools in three states. Their results suggest that a start time of 8:30 a.m. or later is associated with higher grades in core subjects and achievement on state and national standardized tests. The later start times were also associated with increased attendance and reduced tardiness.

Following Seattle Public Schools’ decision to delay secondary start times by nearly an hour, Dunster, et al. (2018) conducted a pre-post policy evaluation and found a statistically significant increase in daily median sleep duration as well as a 4.5% improvement in the median grades of high school students post-policy change. Heissel & Norris (2018) instrumented the hours of sunlight before school in different time zones to assess the relationship between start time and achievement and found that moving start times one hour later relative to sunrise can increase test scores in reading and math. Finally, using data from the Child Development Supplement to the Panel Study of Income Dynamics (PSID-CDS), Groen & Pabilonia (2019) examined the effect of high school start times on a nationally representative sample of students. Paired with time diaries, their analyses suggest that female students who attend schools with later start times sleep more and score higher on reading assessments. Male students do not appear to sleep longer or score higher in response to these changes.

Literature regarding the effect of start times, however, is not always consistent. For example, in 2011, Hinrichs took advantage of changes (delays) in school start times in Minneapolis and several surrounding suburban districts to examine potentially associated changes in individual-level ACT scores. The lack of change in neighboring St. Paul and additional suburban districts offered an important point of comparison. Hinrichs found no evidence of an association between a delayed start time and individual ACT scores.

Most studies exploring the relationship between start times and adolescent sleep, health, and achievement ignore tradeoffs often made between teenagers and young children to balance transportation logistics and budget constraints. A few exceptions explore the relationship between start time and outcomes for younger students. Temkin, Princiotta, Ryberg, & Lewin (2018) compared online survey data from seventh- and eighth-grade students attending eight late and three early starting schools in a diverse suburban school district. They found that students attending later start schools slept more (despite later bedtimes) and reported feeling less sleepy. In 2015, Keller, Smith, Gilbert, Bi, Haak, and Buckhalt examined the association between early elementary start times and standardized test scores in public elementary schools across Kentucky. They found a negative association between early school start time and school achievement though this was concentrated in schools serving few students that qualify for free or reduced-price lunch. In a follow-up study, Keller, Gilbert, Haak, Bi, and Smith (2017) documented a significant positive association between early school start times and negative behavioral outcomes (harassment, in-school removals, suspension, and expulsions). Some of these associations were only significant for schools in the non-Appalachian region of Kentucky.

We note several important gaps in existing literature. First, the evidence that informs ideas about the relationship between start time and elementary student sleep, health, and academics is thin. Second, most scholars fail to explore whether and how the relationship between sleep and achievement differs across important subpopulations. In 2017, the National Association for the Advancement of Colored People (NAACP) Boston took a strong stand against a policy design that required elementary students to start earlier to ensure adolescent sleep. Leaders expressed concerns that “Parents of color are disproportionately in lower-wage jobs, and are less likely to have the flexibility needed to build their schedules around a school day that ends at 1:15 or 1:55, let alone pay for any resulting need in after school care” (NAACP Boston, 2017). In addition, few studies acknowledge the relationship between policy changes and other important bottom-line outcomes such as enrollment. Like many urban Midwestern districts, our study district has registered steady enrollment declines for the past several years. In the context of open enrollment policies between this urban district and well-resourced suburban districts, and a robust charter sector, the district’s decisions on timing may exacerbate such declines if parents find the change in start time difficult to manage with work schedules.

In this study we investigate whether and how changes in school start times impact enrollment trends across grade spans. In particular, we explore enrollment in transition years for elementary, middle, and high school, years where parents may be more likely to enroll their child in a different school due to inconvenient school start and stop times. Furthermore, to recognize NAACP Boston’s concern for parents of color, we also examine different racial demographics for every enrollment pattern by asking: Did the impact on enrollment vary by race or other socioeconomic characteristics? Disaggregating enrollment patterns for transition school years and racial subgroups help to focus our research on the populations most likely to be affected in an attempt to fill noted gaps in existing literature.

Methodology
We employ a difference-in-differences (DD) model, which compares the change in enrollment in schools in two treatment arms (start time changed later and earlier) to the change in enrollment in schools that did not experience a change in start time — these schools act as the counterfactual. Using the DD model allows us to isolate a causal effect of the treatment. The key assumption is that the decision of which schools are treated was unrelated to their enrollment trends leading into the policy change.

We implemented the DD by creating an interaction between the 2019-2020 school year — our only year after the change in start time, which we label “post”— and a binary indicator for whether the school changes start time — i.e., the treatment status. This allows us to identify the effect of the policy on schools in the treatment group, in comparison to the trends in enrollment that prevailed in schools in the control group. In order for the interaction of post and treatment status to isolate the change in enrollment attributable to the policy, we need to control for school and year fixed effects (this is a more robust version of controlling for the pre-period and treatment status).

The year fixed effects span 2014 through 2019 and control for the existing trends in enrollment across all schools in the district. The school fixed effects control for the variation between schools — for instance, some schools are bigger than others — and allow us to focus on changes in enrollment within schools over time. To highlight enrollment trends between grades, we subsetted by elementary school grades and transition grades. In order to look at variations in enrollment trends within racial groups, we replaced our total enrollment independent variable with a variable for enrollment by race.

Results:
We find that, on average, Kindergarten through 4th grade saw steeper enrollment declines in schools that changed start time relative to schools with no change in start time. Results indicate that, on average, schools that changed to an earlier start time lost on the order of four or five more students per grade than schools that did not change start times in first through fourth grades. The results in 2nd and 4th grade are significant at the p<0.1 level, and the results for 3rd grade are significant at the p<0.05 level. The results for first grade are of a similar magnitude to the other grades but are not statistically significant at conventional levels. Kindergarten experienced the largest impact: schools that changed earlier lost around eight more kindergarteners than schools that did not change start times. This result is significant at the p<0.05 level. We see no significant differences in enrollment for schools that changed to later start times.

Next, we investigate if enrollment trends were more or less pronounced within particular racial groups. Across racial groups, elementary schools that changed earlier compared to ones that did not lost an average of 4.5 students per grade p<0.01. We find that for elementary schools that changed to an earlier start time, Black/African American p<0.05 as well as Asian p<0.01 enrollments declined by an average of two students more per grade than schools that did not change. There is some evidence of an impact on Asian, Latinx, and White enrollment in schools that changed to later start times. However, we note that only two elementary schools are included in the change-later arm of the treatment.

Lastly, we perform a series of falsification and robustness checks. The falsification tests use 2015 and 2018 as placebo policy changes (i.e., we act as if the start time changed in these years), and we find no statistically significant impact. This is reassuring and suggests that the schools that changed start time in 2019 were not different from schools that did not change start time in the pre-period. We also interact treatment status with each year leading into the policy change (this is commonly called an “event study” analysis) to further assess pre-trends. The results are not robust to this specification. However, the statistical power needed is hampered by the fact that we have only one year post-treatment. Future research will use additional post years, but this will be limited by the fact that COVID-19 has drastically altered enrollment patterns.

Conclusions
We find evidence of a decline in enrollment in elementary schools that changed to an earlier start time, particularly in Kindergarten and particularly amongst Black/African American and Asian student populations. The district changed start times to accommodate concerns about adolescent health, achievement, and, presumably, enrollment. However, our results make clear that the change also influenced families’ choices at elementary locations, some of which moved to start times as early as 7:30 a.m. Districts face real budget constraints — running three bus routes per day rather than one or two bus routes would require an increase in revenue — and yet, changes in start times designed to address mental health and achievement concerns may inadvertently exacerbate enrollment declines which then further exacerbates gaps in funding.

Our results are robust to various specifications; using different years as placebo treatments (both 2015 and 2018) to see if there were any significant changes in enrollment that did not relate to the policy, we found no significant differences before Fall 2019. We only have one year of enrollment data after the start time policy was implemented, however, limiting our ability to conduct additional tests and event study analysis. Unfortunately, future enrollment data in the district will be heavily affected by the COVID-19 pandemic and the many challenges it presents to mental health, achievement, attendance, and enrollment. When students in this district (and large urban districts around the country) return for in-person instruction, start times are again likely to be staggered in waves. Future research should assess several remaining facets of school start time change. For example, when students leave a district in response to a time change, where do they go? Does variation in start times in surrounding districts and charter schools correlate with movement? How does socioeconomic status correlate with start time changes between districts? The NAACP Boston’s concerns about working parents — concerns spotlighted in the face of the COVID pandemic by those attempting to balance parenting and work — should be taken seriously by school administrators dedicated to addressing gaps in opportunity and achievement. In addition, the long-term effects of shifts in school start time on attendance, enrollment and achievement should be examined.

Acknowledgements:
We would like to thank Adri Arquin, Emma Curchin and Olivia Matzke for valuable research assistance.

Authors: Dr. Lesley Lavery, associate professor of political science, Macalester College, St. Paul, MN, llavery@macalester.edu; Dr. Kristine West, associate professor of economics, St. Catherine University, St. Paul, MN,klwest@stkate.edu; Jessica Brown, student, Macalester College, jessicabrown8341@gmail.com; Elizabeth Burton, student, Macalester College, lizzyburton09@gmail.com; Emma Kettle, student, St. Catherine University,


Work Cited
Carrell, S. E., Maghakian, T., & West, J. E. (2011). A’s from Zzzz’s? The causal effect of school start time on the academic achievement of adolescents. American Economic Journal: Economic Policy, 3(3), 62-81.

Dunster, G. P., de la Iglesia, L., Ben-Hamo, M., Nave, C., Fleischer, J. G., Panda, S., & Horacio, O. (2018). Sleepmore in Seattle: Later school start times are associated with more sleep and better performance in high school students. Science advances, 4(12), eaau6200.

Glatter, H. (2017, December 15). NAACP Opposes BPS Scheduling Changes. Retrieved August 21, 2020, from https://www.bostonmagazine.com/education/2017/12/15/naacp-school-time-boston/

Groen, J. A., & Pabilonia, S. W. (2019). Snooze or lose: High school start times and academic achievement. Economics of Education Review, 72, 204-218.

Heissel, J. A., & Norris, S. (2018). Rise and shine the effect of school start times on academic performance from childhood through puberty. Journal of Human Resources, 53(4), 957-992.

Hinrichs, P. (2011). When the bell tolls: The effects of school starting times on academic achievement. Education Finance and Policy, 6(4), 486-507.

Keller, P. S., Gilbert, L. R., Haak, E. A., Bi, S., & Smith, O. A. (2017). Earlier school start times are associated with higher rates of behavioral problems in elementary schools. Sleep health, 3(2), 113-118.

Keller, P. S., Smith, O. A., Gilbert, L. R., Bi, S., Haak, E. A., & Buckhalt, J. A. (2015). Earlier school start times as a risk factor for poor school performance: An examination of public elementary schools in the commonwealth of Kentucky. Journal of Educational Psychology, 107(1), 236.

Quality Education for Every Student [QUEST]. (2017). [Statement on Boston Public Schools’ Proposed Bell Times: Changes Will Disproportionately Harm Families of Color] [Statement]. Retrieved August 21, 2020, from https://www.facebook.com/questbps/photos/pb.103276616497476.-2207520000.1513343712./916925265132603/?type=3&theater

Perkinson-Gloor, N., Lemola, S., & Grob, A. (2013). Sleep duration, positive attitude toward life, and academic achievement: The role of daytime tiredness, behavioral persistence, and school start times. Journal of adolescence, 36(2), 311-318.

Temkin, D. A., Princiotta, D., Ryberg, R., & Lewin, D. S. (2018). Later start, longer sleep: implications of middle school start times. Journal of School Health, 88(5), 370-378.

Wahistrom, K. (2002). Changing times: Findings from the first longitudinal study of later high school start times. Nassp Bulletin, 86(633), 3-21.

Wahlstrom, K., Dretzke, B., Gordon, M., Peterson, K., Edwards, K., & Gdula, J. (2014). Examining the impact of later high school start times on the health and academic performance of high school students: a multi-site study.