Is four less than five? Effects of four-day school weeks on student achievement in Oregon

1,850,238 student-year observations in maths and 1,844,010 in reading in the working-paper panel (Oregon, 2007-2015); the published version extends the panel to grades 3-8 over 2005-2019

Oregon public school students in tested grades, in districts that adopted a four-day school week largely for budgetary reasons after the Great Recession; four-day schedules are concentrated in small and remote districts

Journal of Public Economics 193, 104308. Difference-in-differences on staggered district adoption of the four-day week, with school and year fixed effects on student-level administrative test records. Full text read from IZA Discussion Paper 12204 (March 2019); the published 2021 version extends the panel and adds an instrumental-variables specification using four-day adoption as an instrument for weekly hours, whose coefficients are taken from the published abstract rather than from the text read. Two honest weaknesses. First, adoption is not random — districts choose the four-day week under financial stress, which is why the design leans on school fixed effects and event study leads; the leads are small and insignificant (t-1 = -0.005, t-2 = -0.011 for maths). Second, the maths estimate is not robust to district-specific linear time trends, falling from -0.044 (SE 0.017) to -0.028 (SE 0.018) and losing significance, while reading survives at -0.019 (SE 0.011). The schedule change also bundles three things — fewer weekly hours, longer days, earlier start times — and the paper measures all three rather than assuming one.

Switching to a four-day week lowers achievement by roughly 0.03-0.06 SD: maths -0.044 (SE 0.017) and reading -0.033 (SE 0.011) in the working-paper baseline, and 0.037-0.059 SD maths and 0.033-0.042 SD reading across specifications in the published version. Schools that ran the four-day week throughout show larger declines (maths -0.076, reading -0.055). The mechanism is measured directly and is the reason this source matters for instructional time: adopting the four-day week lengthens the school day by 37.7 minutes but shortens the week by 215.8 minutes — nearly three and a half hours of lost weekly school time — and moves the start of the day 6.0 minutes earlier. Absences fall by 1.36 days and discipline days by 0.18, so the schedule buys back some attendance and still loses on net. Converting the reduced form into a price for time, the published IV gives a one-hour increase in weekly time in school as worth 0.018 SD in maths and 0.006 SD in reading. Effects are worse for boys in both subjects and for low-income students in reading. Thompson's framing is the policy-relevant one: the four-day week is a cost-cutting measure, and its achievement-per-dollar trade-off turns out comparable to the traditional cost-cutting alternatives such as raising class size.

Low. The treatment is a district-level calendar decision driven by budget pressure, imposed on every child in the district, and the design compares the same schools before and after adoption. Family and child traits do not select into the four-day week at the individual level. The residual concern is district-level selection — districts adopting under fiscal stress may be on declining trajectories for other reasons — which the event-study leads and the district-trend specification probe, and which is exactly where the maths estimate weakens.

The direction replicates across states and authors: Morton's Oklahoma analysis and Thompson and co-authors' subsequent work on high-school achievement and engagement find comparable small negative effects, and the implied return per weekly hour sits in the same range as Lavy's cross-national within-pupil estimate and the COVID closure literature. What is not settled is the magnitude, since the maths coefficient is sensitive to district time trends and the effect is larger in the subsample of long-running four-day schools, where selection is strongest.