homework

Sleep Health 3 (2017) 451–457

Contents lists available at ScienceDirect

Sleep Health Journal of the National Sleep Foundation

journal homepage: sleephealthjournal.org

The economic implications of later school start times in the

United States

Marco Hafner, MPhil, MSc a, Martin Stepanek, MSc a,b, Wendy M. Troxel, PhD c,⁎ a RAND Europe, Westbrook Centre/Milton Rd, Cambridge CB4 1YG, United Kingdom b Institute of Economic Studies, Faculty of Social Sciences, Charles University, Prague, Czech Republic c RAND Corporation, Health Division, Pittsburgh, PA 15213, United States

a b s t r a c ta r t i c l e i n f o

⁎ Corresponding author. E-mail address: [email protected] (W.M. Troxel).

https://doi.org/10.1016/j.sleh.2017.08.007 2352-7218/© 2017 National Sleep Foundation. Published b

Article history: Received 15 June 2017 Received in revised form 15 August 2017 Accepted 18 August 2017

Keywords: School start times Adolescent sleep Cost–benefit analysis Economics Economic modeling

Numerous studies have shown that later school start times (SST) are associated with positive student outcomes, including improvements in academic performance, mental and physical health, and public safety. While the benefits of later SST are very well documented in the literature, in practice there is opposition against delaying SST. A major argument against later SST is the claim that delaying SST will result in significant additional costs for schools due to changes in bussing strategies. However, to date, there has only been one published study that has quantified the potential economic benefits of later SST in relation to potential costs. The current study investigates the economic implications of later school start times by examining a policy experiment and its subsequent state-wide economic effects of a state- wide universal shift in school start times to 8.30 AM. Using a novel macroeconomic modeling approach, the study estimates changes in the economic performance of 47 US states following a delayed school start time, which includes the benefits of higher academic performance of students and reduced car crash rates. The benefit–cost projections of this study suggest that delaying school start times is a cost- effective, population-level strategy, which could have a significant impact on public health and the US economy. From a policy perspective, these findings are crucial as they demonstrate that significant economic gains resulting from the delay in SST accrue over a relatively short period of time following the adoption of the policy shift.

© 2017 National Sleep Foundation. Published by Elsevier Inc. All rights reserved.

Introduction

Inadequate sleep among adolescents has emerged as a public health epidemic.1 Even though teens need an average of 8 to 10 hours of sleep each night, about 60 percent of middle school students report weeknight sleep duration of less than nine hours and only about 7 per cent of high school students report 9 hours or more of sleep per night.1 The existing literature has shown that a lack of sleep among adolescents is associated with numerous adverse out- comes, including poor physical and mental health, behavioral prob- lems, suicidal ideation and attempts, attention and concentration problems, and suboptimal academic performance.2–9 In addition, in- sufficient sleep is associated with motor vehicle crashes, the leading cause of death of teenagers.10

Many factors have been found to be associated with adolescent sleep loss, including busy social lives, school work, participation in afterschool activities, and use of technology in the bedroom.11

y Elsevier Inc. All rights reserved

Furthermore, known biological changes in adolescent sleep–wake cycles contribute to delayed sleep–wake cycle.12 Rise times are pri- marily determined by a factor of public policy, and that factor is school start times (SST).13 In order to accommodate the known bio- logical shift in adolescent sleep–wake cycles leading to later bedtimes and later wake-times, major medical organizations recommend that middle and high schools start no earlier than 8:30 AM.14,15 Despite these recommendations, a Centers for Disease Control and Preven- tion (CDC) study estimated that 82% of middle and high schools start before 8:30 AM, with an average start time at 8:03 AM, showing significant variance of SST across different states.16 While the benefits of later SST are well-documented in the literature, in practice there is often opposition against delaying SST. A major argument against later SST is the claim that delaying SST will result in significant additional costs for schools due to changes in bussing strategies.

To our knowledge, however, there has only one been one pub- lished study to date that has aimed to quantify the potential benefits of later SST in relation to potential costs. Specifically, the analysis by the Brookings Institution17 examined the cost-benefits of delaying school start times and found a benefit–cost ratio of 9:1 for a 1hour

.

1 In economics this is referred to as a ‘multiplier effect’, which is when extra income leads to more spending in the economy which subsequently can create more income.

452 M. Hafner et al. / Sleep Health 3 (2017) 451–457

later start time among middle and upper grades. In other words, for every $1 spent, the return is $9. Costs were estimated to be $150 per year per student, based on data from a single district in North Carolina18 and were determined by a change in the school bus system, from a three-tiered bus system to a single-tier system. Cumulatively, the study estimated an average $17,500 gain per student in terms of lifetime earnings compared to $1950 in costs per student over his/her school career. While the Brookings Institution analysis shows a high benefit to cost-ratio, it is important to highlight that the time horizon for the potential benefits is protracted over the average working life of an individual (e.g., about 45 years). However, from a policymaker's perspective, it is important to have a more granular understanding of the timeframe for when these benefits are likely to accrue.

Against this background, the current study examines the po- tential economic impact from delaying SST for middle and high schools to 8:30 AM. That is, the main research question this study aims to answer is: what are the economic implications of a state-wide universal shift in start times to at least 8:30 AM and how do they vary regionally by state and what is the expected time horizon for the benefits to occur?

Specifically, this study runs a hypothetical policy experiment in order to estimate the potential year-by-year state-wide economic changes for several US states which may occur from a state-wide uni- versal shift to 8.30 AM SST compared to the current start times in each of the states [cross-state average of 8:03 AM, as reported by the CDC16]. The analysis departs from the approach taken in the Brookings Institution study in several ways. First, instead of assuming a one hour later school start time, the current distribution of school start times across different states is taken into account and the impact of an 8:30 AM SST is modeled. Second, when calculating the benefits of SST, this study takes into account the effects on student lifetime earnings as well as the potential impact of reduced car crashes among adolescents, which can have a negative impact on future labor supply of an economy if young adults die prematurely. Third, the Brookings analysis focused only on a general potential gain per student, whereas this study looks at potential economic effects for different re- gions, taking into account the variation of school start times and eco- nomic factors across different US States. Finally, this study also takes into account potential multiplier effects of increased lifetime earnings of individuals. For instance, at any given point in time the additional money these individuals save or consume will create further opportu- nities through further income for other agents in the economy.

Methods

General modeling approach

The analysis is based on a theoretical dynamic general equilibrium model related to a system of mathematical equations to characterize the economic interaction of different agents in an economy such as households, firms, or the government. The economic model builds on the long tradition of computable general equilibrium (CGE) models, which have been extensively applied for economic policy analysis.19–21 CGE models are based on a detailed theoretical frame- work simulating the behavior of various agents and depicting rela- tionships between subjects in an economy described by a set of parameters, equations and conditions that are to be satisfied simulta- neously. The equations are then evaluated using mathematical software,22 giving a set of numerical results representing, for exam- ple, the levels of labor or capital in a simulated economy. CGE models explicitly allow for the analysis of multiple comparable scenarios which differ only in the selected set of parameters, for example by creating both a baseline (or status quo) and a ‘what if’ situation show- ing how the economy would evolve under different policy scenarios. The specific model applied in this study is a so-called ‘Overlapping

Generations (OLG) model, which simulates the behavior of different cohorts of individuals over their lifecycle (see a more detailed de- scription of the model in Appendix A).

Estimates of potential benefits and costs associated with a state-wide shift in SST

As a first step, the model simulates the economic forecast of each of the states under consideration in the baseline scenario, using the current distribution of SST across middle and high schools in 47 US states which is provided by the CDC.16 In a second step, under a dif- ferent ‘what if’ scenario (compared to current start times), the model predicts how the economic output (e.g. measured as gross do- mestic product) of each state would change if the state implemented a universal shift to 8:30 AM SST. The population affected by the policy change are students from grade 6 to grade 12. In the applied economic model, it is assumed that delaying SST leads to extended sleep duration for adolescents,14 which subsequently could impact the economy in a given state through different “channels”. Specifically, we only included effects for which there were sufficient and robust parameters from the existing literature. In particular, this study focuses on two specific ben- eficial channels that could be derived from later SST:

The first channel is mortality from motor vehicle crashes. The data for car crash mortality includes the underlying cause of death data provided by the CDC23 on weekday motor vehicle fatalities among teenagers age 16 to 18, combined with parameters from a study by the AAA Foundation for Traffic Safety, which revealed that about one fifth of fatal motor vehicle crashes involved a driver impaired by sleep- iness, drowsiness or fatigue.24 Together with the estimate by Danner & Phillips,25 which suggests that the car crash rate decreases by 16.5 per- cent due to an hour delay in SST, the potential reduction of car crash mortality rates for each state is calculated. Note that in the model, re- duced mortality levels among adolescents increase the potential future labor population and therefore has a positive effect on the economy. Thus, the labor supply effect derived from motor vehicle mortality data consists of two factors: (1) the direct impact of the individual being alive and productive; and (2) the impact on the individual's po- tential future offspring, which will subsequently be missing and hence will not contribute to the economy in the future.26

The second channel potentially contributing to the benefits of later SST is the impact on academic performance. Using data on the effect of adolescent sleep on academic performance and graduation rates from Wang et al.,9 the model predicts that longer sleep will lead to increased high-school and college graduation rates.9 Specifi- cally, Wang et al.9 estimate that one additional hour of sleep is esti- mated to increase the probability of high school graduation on an average by about 8.6% and the college attendance rate by 13.4%, both with decreasing marginal returns for each hour of additional sleep. Due to the non-linear effect of sleep duration, Wang et al.’s findings suggest that later school start times may create longer-run human capital benefits especially for those adolescents that sleep on average below seven hours a night, which has been estimated to affect more than 40 per cent of the adolescent population.1 The pos- itive effect on adolescents' academic performance and likelihood of high school graduation, in turn, impacts the jobs they are able to ob- tain in the future. This in turn, has a direct effect on how much a par- ticular person contributes towards the economy in future financial earnings. Due to the dynamic nature of the model, at any given point in time, the increased income these individuals save or con- sume will create further opportunities through additional income for other agents in the economy and hence increase overall economic output of each of the states.127 In essence, the effects of changes in

2 Both benefits and costs per student are discounted and presented in present-day values. The future benefits and costs have been discounted by a rate of 4% which is common among the macroeconomic literature.

453M. Hafner et al. / Sleep Health 3 (2017) 451–457

lower car crash rates and higher educational attainment are translat- ed into economic benefits the moment an individual enters the labor market and continues to progress throughout his/her career until retirement.

Note that a shift to 8:30 AM SST is likely to come with some costs, and hence it is relevant to compare the economic benefits of the delayed SST to its potential costs. As mentioned, one of the most important factors driving costs is a change in the bus system from a three-tier to a one- or two-tier system, which would incur additional costs. The Brookings Institution analysis uses a cost estimate of $150 per student per year for the cost–benefit analysis, based on estimates from a school district in Wake County, North Carolina.18 Obviously, the cost will depend on the local circumstances of each state, and even at the more granular school district level it is impossible to rep- resentatively estimate them across the USA. Hence, for the purpose of illustration of the potential benefit–cost ratios four different cost esti- mates are used in the present analyses to provide a more comprehen- sive range of potential costs: $150, $350, $500 and $700. We assume that the cost per student will occur in perpetuity after the policy shift to 8:30 AM SST. This is likely overestimating the actual costs as the majority of the costs would probably accrue at the beginning of the policy shift in the form of upfront investments for new bus routes.

Calibrating the economic model with data from different sources

In the simulated model, the economic output of each state is pro- duced using different production factors including capital and so called ‘effective labor’, which combines physical labor units (e.g. number of workers) with labor efficiency units (e.g. productivity of each unit of labor), where the latter includes units of human capital (e.g. level of education). To address the first component, physical labor, a cohort-component model is applied to predict the size of the future populations in each state using current base population es- timates from the Census Bureau,28–31 as well as mortality and fertility rates data provided by the CDC.23,32 In essence, the base population evolves over time by applying assumptions on mortality and fertility so that the state population changes according to a ‘natural’ increase (births minus deaths), providing a projection of a state's population by one-year cohort groups into the future. In order to address the sec- ond component, labor efficiency, several sources of data from the US Census Bureau are utilized. First, in order to inform what share of the total population belongs to each one-year cohort group, we deter- mine, the proportions of individuals in each state according to their age, gender and ethnicity33 and complement this data with detailed information on school and college attainment. Second, in order to translate the changes in SST into potential personal financial gains of potentially higher academic performance data on the average earnings by the highest educational attainment level is used, which is collected by United States Census Bureau.34 In addition, a variety of parameters were used to calibrate the model, including, among others, the hours of work, capital-labor ratios, capital stock deprecia- tion rates and the growth rate of technological change. A full model specification and a complete description of model parameters is re- ported in Hafner et al. (2017).34

Results

This study illuminates the link between a state-wide universal delay in SST to 8:30 AM and economic gains to different US states. In what follows, the predicted cumulative gains in present-day values (2016 $ figures) aggregated across all 47 states are presented, follow- ed by a breakdown of the benefits by student and the comparison against the potential costs per student. Finally, the overall benefit– cost ratios by state are presented.

Predicted cumulative gains and benefit–cost ratios across the USA

Fig. 1 depicts the cumulative economic gains from delayed SST in present-day value across the 47 US states included in the analysis. The economic gains are displayed as higher levels of economic output that would occur if SST would be delayed compared to the status quo. Economic output is measured as gross state product (GSP), which is the equivalent of gross domestic product (GDP) at country level.

In the first year of the shift to 8:30 AM SST, the model projects no immediate economic gain, given that the first cohort of students graduating from high school is only experiencing one-year of change in the SST policy before graduation. However, as more students will benefit over time from the delayed SST as they enter the labor mar- ket, the gains are increasing over time. For instance, after year two of the policy shift, the model projects an economic gain of about $8.6 billion which represents about 0.04% of total US Gross Domestic Product (GDP). After five years, the economic gain increases to about $37 billion, and to $83 billion and about $140 billion after 10 and 15 years, respectively. On average, this corresponds to an annual gain of about $9.3 billion, which is roughly the annual revenue of Major League Baseball.35

Furthermore, Fig. 2 reports the average benefit–cost ratio per stu- dent across the 47 US states using four different estimates for costs in order to capture a range of relevant scenarios for different school dis- tricts: $150, $350, $500, and $700.2 Under the assumption that the costs per student are $150, as in the Brookings Institution analysis, the benefits are predicted to outweigh the costs per student (e.g. benefit–cost ratio is larger than 1) after about 2 years of delaying SST to 8:30 AM. After 13 years, the benefit–cost ratio would reach 3:1, meaning that every $1 invested would yield a return of $3. The ratio increases annually, reaching 4:1 after 20 years. Assuming a higher cost of $350 per student per year, a universal state-wide delay in SST to 8:30 AM on average is predicted to outweigh the cost after 7 years. Remarkably, assuming a high cost of $500, the “break-even” point would be 16 years. Even when assuming an ex- tremely high cost of $700 per student per year, which is more than four times larger than the suggested cost per student used in the Brookings study, the economic benefits of delaying SST would out- weigh the cost per student after 25 years.

Variation across states: Gains per student and benefit–cost ratios

The findings presented thus far represent average figures across the 47 US states, but a state-by-state analysis reveals significant re- gional variation in the effects. For instance, Table 1 reports the gains per student across 47 states after the policy change to delay SST to 8:30 AM.

For instance, in Alabama, the economic gain per student after 2 years is estimated to be about $31 per student. This is significantly lower than the average of $346 per student across the 47 states. Other states with relatively low gains per student are Arkansas, Idaho and Mississippi (between $177 and $190). On the other hand, states such as Delaware and Massachusetts would proportionally gain more than $700 per student after 2 years. Other states with rel- atively large gains per student are Connecticut, New Jersey, Ohio, Rode Island and Virginia. Note that the difference is mainly driven by variation in the state-wide initial average SST and underlying eco- nomic factors which also vary significantly by state (e.g. the industrial composition or average productivity levels).

Table 2 reports the benefit–cost ratios by state assuming the cost per student to implement the delayed SST to 8:30 AM would be $150

Fig. 1. Predicted cumulative economic gains from delayed SST to 8:30 AM across 47 US States. Note: The figure plots the predicted discounted cumulative gains of delayed SST to 8:30 AM aggregated across 47 US states since the change of policy (from one to fifteen years).

454 M. Hafner et al. / Sleep Health 3 (2017) 451–457

per student per year. The findings suggest that with the exception of Alabama, in every other state, the economic benefits for delaying SST would outweigh the costs within 5 years after the change. The pre- dicted benefit–cost ratio after five years varies from 0.4 (Alabama) to 4.4 (Delaware). Even after two years, a majority of states are pre- dicted to reach a benefit–cost ratio of at least 1:1, meaning that for every $1 spent, there is a $1 return on investment.

Discussion

The current study is the first to measure the economic gains asso- ciated with delaying school start times in states across the US. Using a novel macroeconomic modeling approach, the findings suggest that delaying SST to 8:30 AM could lead to significant economic gains over a relatively short period of time in the form of increased overall

Fig. 2. Predicted average Benefit–Cost Ratios of Delayed SST to 8.30 AM across 47 US States. N 47 U.S states after the change of policy).

economic performance. Departing from the previous cost–benefit analysis provided by the Brookings Institution, this study reports the estimated year-by-year and state-by-state changes in benefits from delaying SST which average across the states from 1.2 (after 2 years) to 3.2 (after 15 years). From a policy perspective, these find- ings are crucial as they demonstrate that significant economic gains resulting from the delay in SST accrue over a relatively short period of time following the adoption of the policy shift. In comparison, the Brookings Institution estimated a benefit–cost ratio of 9:1 per student, but calculated the benefits and costs over the working life of an individual, which is about 45 years on average, and hence the benefit–cost ratio cannot directly be compared to the ratios predicted in this study, which are year-on-year. However, if we apply the annu- al cost in perpetuity assumption of $150 per student per year to the Brookings Institution analysis, which found that the overall lifetime

ote: The figure plots the predicted benefit–cost (B-C) ratios of delayed SST to 8:30 AM across

Table 1 Predicted cumulative economic gain

Years after policy change (gain $ per student)

State 2 5 10 15 20

Alabama 31 246 953 2220 3712 Arizona 374 1498 3325 5440 7619 Arkansas 190 904 1908 3349 4867 California 335 1357 3097 5216 7523 Colorado 294 1199 2876 5158 7515 Connecticut 517 2209 5193 8863 12,639 Delaware 733 3061 7242 12,278 17,275 Florida 456 1783 3943 6525 9145 Georgia 269 1098 2485 4109 5835 Hawaii 476 2295 4688 7553 10,653 Idaho 180 735 1676 2802 4044 Illinois 259 1082 2539 4499 6700 Indiana 273 1273 2824 5097 7459 Iowa 395 1625 3681 5756 7983 Kansas 289 1478 3117 5215 7500 Kentucky 263 1442 3043 4916 6850 Louisiana 381 1587 3723 6422 9232 Maine 275 1148 2748 4702 6740 Massachusetts 704 2693 5673 9050 12,535 Michigan 331 1367 3248 5381 7551 Minnesota 360 1443 3395 5671 8046 Mississippi 177 856 1844 3112 4452 Missouri 341 1396 3302 5877 8468 Montana 338 1366 2942 4656 6473 Nebraska 258 1096 2527 4358 6433 Nevada 191 795 1897 3330 4896 New Hampshire 286 1343 3117 5416 7801 New Jersey 551 2207 5112 8482 11,886 New Mexico 373 1469 3268 5204 7226 New York 295 1244 2970 5298 7862 North Carolina 342 1411 3215 5413 7717 Ohio 410 1646 3510 5665 7879 Oklahoma 291 1153 2568 4226 5989 Oregon 294 1198 2758 4810 7088 Pennsylvania 276 1212 2987 5381 7900 Rhode Island 537 2406 5358 8703 12,081 South Carolina 317 1380 3050 4747 6513 South Dakota 296 1300 2744 4438 6287 Tennessee 225 967 2286 3998 5858 Texas 333 1335 3007 5022 7162 Utah 228 931 2327 3956 5654 Vermont 365 1523 3464 5876 8365 Virginia 595 2129 4774 7706 10,765 Washington 503 2308 4786 7664 10,662 West Virginia 235 972 2263 3813 5443 Wisconsin 360 1498 3572 6034 8610 Wyoming 461 1982 4497 7553 10,702 Average 346 1461 3309 5552 7906

Notes: The table reports the predicted discounted cumulative economic gains per student of delayed SST to 8:30 AM across all 47 US states, compared to the status quo with current distribution of SST.

455M. Hafner et al. / Sleep Health 3 (2017) 451–457

gain of a student is $17,500 for a one-hour shift in SST, and further as- sume a 45 year time horizon, then the predicted adjusted benefit– cost ratio of the Brookings Institution analysis is approximately 6:1, instead of 9:1. By taking a more comprehensive and more detailed national approach, the figures presented in Table 2 suggest that after only 15 years (about a third of the working life of an individual), the benefit–cost ratio across the 47 states is about half of the benefit– cost ratio of the Brookings analysis. If the estimates reported in Table 2 would be extended to 45 years, the ratio is predicted to in- crease to about 7.5:1, which is about 1.2 times larger than the esti- mated adjusted benefit–cost ratio by Brookings Institution (of 6:1), even though the current analysis implies generally a net increase in SST of less than an hour (approximately 30 minutes).

Overall, this study only applies parameters in the calibration pro- cess of the model for which robust empirical evidence is available concerning the impact of sleep loss on affects adolescents' health

and academic performance. Specifically, we utilized available data on car crash mortality and impaired academic performance. There- fore, our models are conservative in nature, as we did not take into account other potential impacts of insufficient sleep, such as the ef- fects on mental health, including depression and suicide, or other po- tential negative effects related to obesity or other morbidities that are also associated with insufficient sleep. Hence, the reported benefits in this study are likely an underestimation of the full benefits related to delaying SST.

On the cost side, this study uses a previous estimate of $150 per student per year but also uses higher cost estimates to assess the ef- fectiveness of the policy change at a broader range of costs. Since costs will vary by school district, the costs applied in the current model serve for illustration purposes, but represent ostensible ranges. Furthermore, beyond increased transportation costs, it is pos- sible that there could be other costs which are not included in our model calculations, such as the costs of having to reschedule after- school activities. In addition, due to later school start time costs, par- ents could incur costs associated with having to go to work later or before-or after school childcare and there could be a potential loss of income associated with a reduction in after-school employment for adolescents. However, in our analysis, on average, the delay SST to 8:30 AM only reflected an average delay of 30 minutes. In reality, given that many schools start before 8 AM, it is also possible that a greater “dose” of the intervention (i.e., more than a 30 minute change) could result in even greater benefits to outweigh the costs. Nevertheless, even if higher cost estimates (e.g. $500 per student per year) are applied, which likely would cover some of these difficult to quantify additional potential costs to parents and the wider socie- ty, the benefits from delaying SST would still outweigh the costs after fifteen years. Moreover, in conjunction with the highly consistent and robust data showing the widespread consequences of adolescent sleep loss on health, safety, and academic performance,36–42 these benefit–cost projections suggest that delaying school start times is a cost-effective, population-level strategy which could have a signifi- cant impact on public health and the US economy.

These findings must be interpreted within the constraints of the study and the specific modeling approach. First, our model is a simu- lated or hypothetical “natural experiment” that presupposes a state- wide universal shift in school start times to 8:30 AM or later. This pre- supposition may seem unjustified given that start times are generally determined at the local district level. However, there are several examples of proposed policy initiatives in states across the country, including a bill in California that mandates that California middle and high school start no earlier than 8:30 AM.43 Thus, the hypothetical policy shift modeled in the current analysis is potential- ly a conceivable strategy. Second, we focused on the cost-benefits of later SST for the 47 states for which there was available data from the CDC on SST and the average start time was before 8:30 AM, and therefore do not have estimates for Maryland, District of Columbia, North Dakota, and Alaska. Third, the specific modeling approach taken in this study is in part based on assumptions that may influence the modeling outcome. It is important to emphasize that whenever an assumption had to be made, we aimed to make sure that the spe- cific assumption would be conservative, hence leading to a potential underestimation of the potential true effect. Finally, as mentioned, our model focuses on two specific factors that drive costs: the impact of sleep insufficiency on motor vehicle crashes/mortality and academic achievement/high school graduation rates. These factors were chosen because we were able to collect and derive robust estimates from the literature. However, as mentioned, there are nu- merous other costs associated with mental and physical morbidity that were not included in our model. For instance, the combined pub- lic health costs of the obesity epidemic in children and adolescents and its associated cardiovascular morbidities are estimated at $45

Table 2 Predicted benefit–cost ratio by state: cost $150 per student per year

Years after policy shift

State 2 years 5 years 10 years 15 years 20 years

Alabama 0.1 0.4 0.8 1.3 1.8 Arizona 1.3 2.2 2.6 3.1 3.6 Arkansas 0.7 1.3 1.5 1.9 2.3 California 1.1 2.0 2.5 3.0 3.6 Colorado 1.0 1.7 2.3 3.0 3.5 Connecticut 1.8 3.2 4.1 5.1 6.0 Delaware 2.5 4.4 5.7 7.1 8.2 Florida 1.6 2.6 3.1 3.8 4.3 Georgia 0.9 1.6 2.0 2.4 2.8 Hawaii 1.6 3.3 3.7 4.4 5.0 Idaho 0.6 1.1 1.3 1.6 1.9 Illinois 0.9 1.6 2.0 2.6 3.2 Indiana 0.9 1.8 2.2 2.9 3.5 Iowa 1.3 2.3 2.9 3.3 3.8 Kansas 1.0 2.1 2.5 3.0 3.5 Kentucky 0.9 2.1 2.4 2.8 3.2 Louisiana 1.3 2.3 2.9 3.7 4.4 Maine 0.9 1.7 2.2 2.7 3.2 Massachusetts 2.4 3.9 4.5 5.2 5.9 Michigan 1.1 2.0 2.6 3.1 3.6 Minnesota 1.2 2.1 2.7 3.3 3.8 Mississippi 0.6 1.2 1.5 1.8 2.1 Missouri 1.2 2.0 2.6 3.4 4.0 Montana 1.2 2.0 2.3 2.7 3.1 Nebraska 0.9 1.6 2.0 2.5 3.0 Nevada 0.7 1.1 1.5 1.9 2.3 New Hampshire 1.0 1.9 2.5 3.1 3.7 New Jersey 1.9 3.2 4.0 4.9 5.6 New Mexico 1.3 2.1 2.6 3.0 3.4 New York 1.0 1.8 2.4 3.1 3.7 North Carolina 1.2 2.0 2.5 3.1 3.6 Ohio 1.4 2.4 2.8 3.3 3.7 Oklahoma 1.0 1.7 2.0 2.4 2.8 Oregon 1.0 1.7 2.2 2.8 3.3 Pennsylvania 0.9 1.8 2.4 3.1 3.7 Rhode Island 1.8 3.5 4.2 5.0 5.7 South Carolina 1.1 2.0 2.4 2.7 3.1 South Dakota 1.0 1.9 2.2 2.6 3.0 Tennessee 0.8 1.4 1.8 2.3 2.8 Texas 1.1 1.9 2.4 2.9 3.4 Utah 0.8 1.3 1.8 2.3 2.7 Vermont 1.2 2.2 2.7 3.4 4.0 Virginia 2.0 3.1 3.8 4.4 5.1 Washington 1.7 3.3 3.8 4.4 5.0 West Virginia 0.8 1.4 1.8 2.2 2.6 Wisconsin 1.2 2.2 2.8 3.5 4.1 Wyoming 1.6 2.9 3.6 4.4 5.1 Average 1.2 2.1 2.6 3.2 3.7

456 M. Hafner et al. / Sleep Health 3 (2017) 451–457

billion a year, and sleep loss is longitudinally associated with in- creased risk of obesity in children and adolescents.38 Further, insuffi- cient sleep among teens is associated with an increased risk of engaging in property and violent crime.40 The direct and indirect costs of crime, including direct economic losses, increased insurance rates, loss of productivity, and various aspects of the criminal justice system, from police, to courts, to juvenile facilities and prisons are estimated in the billions of dollars.44 In addition, the robust associa- tion between insufficient sleep and poor sleep quality and adolescent risk for mental health problems and other risk-taking behaviors, including substance use, could also contribute to substantial societal costs. Taken together, our estimates suggest substantial benefits rela- tive to costs on a state-wide basis related to a universal change in SST and, if anything, these estimates are likely conservative estimates of the true cost savings.

In summary, it is important to put this economic data in context. The findings of this study, as well as the Brookings Institution find- ings, suggest that the benefits of later start times far out-weigh the immediate costs. Moreover, when paired with the substantial

literature demonstrating the dire public health consequences of in- sufficient sleep among adolescents, the multitude of health and aca- demic benefits associated with later start times, and the lack of any scientific evidence to suggest that there are benefits to having adoles- cents start school earlier, these data provide a strong case to counter the argument that changing school start times is too costly to endeav- or a change. Policymakers, educators, and community members should shift from the narrow and often short-sighted focus on the costs of shifting to healthy start times to a focus on the significant benefits associated with later SST, including demonstrable long- term public health and economic benefits.

Supplementary data to this article can be found online at https:// doi.org/10.1016/j.sleh.2017.08.007.

References

1. Basch CE, Basch CH, Ruggles KV, Rajan S. Prevalence of sleep duration on an aver- age school night among 4 nationally representative successive samples of American high school students, 2007–2013. Prev Chronic Dis. 2014;11:E216.

2. Pasch KE, Laska MN, Lytle LA, Moe SG. Adolescent sleep, risk behaviors, and de- pressive symptoms: are they linked? Am J Health Behav. 2010;34(2):237–248.

3. Short MA, Gradisar M, Lack LC, Wright HR, Dohnt H. The sleep patterns and well- being of Australian adolescents. J Adolesc. 2013;36(1):103–110.

4. Pallesen S, Saxvig IW, Molde H, Sorensen E, Wilhelmsen-Langeland A, Bjorvatn B. Brief report: behaviorally induced insufficient sleep syndrome in older adoles- cents: prevalence and correlates. J Adolesc. 2011;34(2):391–395.

5. Lowry R, Eaton DK, Foti K, McKnight-Eily L, Perry G, Galuska DA. Association of sleep duration with obesity among US high school students. J Obes. 2012;2012.

6. Lytle LA, Pasch KE, Farbakhsh K. The relationship between sleep and weight in a sample of adolescents. Obesity. 2011;19(2):324–331.

7. Hart CN, Carskadon MA, Considine RV, et al. Changes in children's sleep duration on food intake, weight, and leptin. Pediatrics. 2013;132(6):e1473-480.

8. Miller NL, Shattuck LG, Matsangas P, Dyche J. Sleep and academic performance in US military training and education programs. Mind Brain Educ. 2008;2(1):29–33.

9. Wang K, Sabia JJ, Cesur R. Sleepwalking through School: New Evidence on Sleep and Academic Performance. IZA Discussion Paper No. 9829; 2016.

10. Richard Fischer P. Mission Not Accomplished: Teen Safe Driving, the Next Chapter. Washington, DC: GHSA and Ford Motor Company Fund; 2016.

11. Carskadon MA. Factors Influencing Sleep Patterns of Adolescents. In: Carskadon MA, editor. Adolescent Sleep Patterns: Biological, Social, and Psychological Influ- ences. New York: Cambridge University Press; 2002. p. 4–26.

12. Crowley SJ, Acebo C, Carskadon MA. Sleep, circadian rhythms, and delayed phase in adolescence. Sleep Med. 2007;8(6):602–612.

13. Knutson KL, Lauderdale DS. Sociodemographic and behavioral predictors of bed time and wake time among US adolescents aged 15 to 17 years. J Pediatr. 2009; 154(3):426–430.

14. Adolescent Sleep Working Group. School start times for adolescents. Pediatrics. 2014;134(3):642–649.

15. Owens JA, Belon K, Moss P. Impact of delaying school start time on adolescent sleep, mood, and behavior. Arch Pediatr Adolesc Med. 2010;164(7):608–614.

16. Wheaton AG, Ferro GA, Croft JB. School start times for middle school and high school students - United States, 2011-12 school year. MMWR Morb Mortal Wkly Rep. 2015;64(30):809–813.

17. Jacob BA, Rockoff JE. Organizing Schools to Improve Student Achievement: Start Times, Grade Configurations, and Teacher Assignments. The Hamilton Project, Brookings Institution; 2011.

18. Edwards F. Early to rise? The effect of daily start times on academic performance. Econ Educ Rev. 2012;31(6):970–983.

19. Allan G, Lecca P, McGregor P, Swales K. The economic and environmental impact of a carbon tax for Scotland: a computable general equilibrium analysis. Ecol Econ. 2014;100:40–50.

20. Lofgren H, Cicowiez M, Diaz-Bonilla C. MAMS–A Computable General Equilibrium Model for Developing Country Strategy Analysis. In: Dixon PB, Jorgenson DW, ed- itors. Handbook of Computable General Equilibrium Modeling, vol. 1A. Waltham, MA: Elsevier; 2013. p. 159–276.

21. Zodrow GR, Diamond JW. Dynamic Overlapping Generations Computable General Equilibrium Models and the Analysis of Tax Policy: The Diamond–Zodrow Model. In: Dixon PB, Jorgenson DW, editors. Handbook of Computable General Equilibri- um Modeling, vol. 1. Elsevier; 2013. p. 743–813.

22. MATLAB [computer program]. Natick, MA: MathWorks. 2017. 23. Centers for Disease Control and Prevention (CDC). CDC WONDER: compressed

mortality file, mortality for 1999–2015 with ICD 10 codes. http://wonder.cdc. gov/cmf-icd10.html, Accessed date: 5 November 2017.

24. Tefft BC. Prevalence of Motor Vehicle Crashes Involving Drowsy Drivers, United States, 2009–2013. Washington, DC: AAA Foundation for Traffic Safety; 2014.

25. Danner F, Phillips B. Adolescent sleep, school start times, and teen motor vehicle crashes. J Clin Sleep Med. 2008;4(6):533–535.

26. Taylor J, Hafner M, Yerushalmi E, et al. Estimating the Economic Costs of Antimicro- bial Resistance: Model and Results. Santa Monica, CA: RAND Corporation; 2014.

27. Hegeland H. The Multiplier Theory. , vol. 9Lund, Sweden: CWK Gleerup; 1954 xii.

457M. Hafner et al. / Sleep Health 3 (2017) 451–457

28. U.S. Census Bureau. Annual estimates of the civilian population by single year of age and sex for the United States and states: April 1, 2010 to July 1, 2015. https://www2.census.gov/programs-surveys/popest/datasets/2010-2015/state/ asrh/.

29. Census Bureau US. Annual state resident population estimates for 5 race groups (5 race alone or in combination groups) by age, sex, and Hispanic origin: April 1, 2010 to July 1, 2015. https://www2.census.gov/programs-surveys/popest/ datasets/2010-2015/state/asrh/.

30. Census Bureau US. Annual state resident population estimates for 6 race groups (5 race alone groups and two or more races) by age, sex, and Hispanic origin: April 1, 2010 to July 1, 2015. https://www2.census.gov/programs-surveys/popest/ datasets/2010-2015/state/asrh/.

31. Census Bureau US. Estimates of the total resident population and resident popula- tion age 18 years and older for the United States, states, and PuertoRico: July 1, 2015. https://www2.census.gov/programs-surveys/popest/datasets/2010-2015/ state/asrh/.

32. Centers for Disease Control and Prevention (CDC). CDC wonder. https://wonder. cdc.gov.

33. U.S. Census Bureau. Data: 2015 population data. https://www.census.gov/data. html.

34. Hafner Marco, Stepanek Martin, Troxel Wendy M. Later school start times in the U.S.: An economic analysis. Santa Monica, CA: RAND Corporation; 2017. https://www. rand.org/pubs/research_reports/RR2109.html.

35. Brown M. MLB Sees Record Revenues Approaching $10 billion for 2016. Forbes; 2016.

36. Owens J. Insufficient sleep in adolescents and young adults: an update on causes and consequences. Pediatrics. 2014;34(3):e921–e932.

37. Mitchell JA, Rodriguez D, Schmitz KH, Audrain-McGovern J. Sleep duration and ad- olescent obesity. Pediatrics. 2013;131(5):e1428-434.

38. Magee L, Hale L. Longitudinal associations between sleep duration and subsequent weight gain: a systematic review. Sleep Med Rev. 2012;16(3):231–241.

39. Troxel WM, Ewing BA, D'Amico EJ. Examining racial/ethnic disparities in the asso- ciation between adolescent sleep and alcohol or marijuana use. Sleep Health. 2015; 1(2):104–108.

40. Umlauf MG, Bolland JM, Lian BE. Sleep disturbance and risk behaviors among inner-city African-American adolescents. J Urban Health. 2011;88(6):1130–1142.

41. Vorona RD, Szklo-Coxe M, Lamichhane R, Ware JC, McNallen A, Leszczyszyn D. Ad- olescent crash rates and school start times in two central Virginia counties, 2009- 2011: a follow-up study to a southeastern virginia study, 2007-2008. J Clin Sleep Med. 2014;10(11):1169–1177.

42. Wheaton AG, Chapman DP, Croft JB. School start times, sleep, behavioral, health, and academic outcomes: a review of the literature. J Sch Health. 2016;86(5):363–381.

43. Myers J. School Day wouldn't Begin before 8:30 AM in California under Bill that Clears the State Senate. Los Angeles Times; 2017. http://www.latimes.com/ politics/essential/la-pol-ca-essential-politics-updates-school-day-wouldn-t- begin-before-8-30-1496189944-htmlstory.html, Accessed date: 1 June 2017.

44. Jurisdictional and Program Self-Assessment. Consequences of Juvenile Delinquen- cy and Violence. Jurisdictional Technical Assistance Package for Juvenile Correc- tions. Rockville, MD: Office of Juvenile Justice and Delinquency Prevention (OJJDP); 2000.