Decomposing the Pandemic Effect on School Non-Attendance: A Three-Regime Analysis Showing Persistent Elevation Beyond Policy Resolution in Japan, 2012-2024 — Data and Code

This Zenodo deposit accompanies the manuscript by the same title. It includes the manuscript, analytic code, consolidated dataset, and all figures. This study analyzes 13 years (FY2012-FY2024) of grade-level annual school non-attendance counts in Japan, using piecewise linear regression to decompose the time series into three regimes (Pre-Act, Post-Act+COVID-onset, COVID-period). FY2024 is treated as an independent observation period for testing whether the COVID-period effect persists beyond the policy end of the pandemic emergency (May 2023 reclassification of COVID-19 from class 2 to class 5 in Japan). Key findings:- Within elementary grades, the COVID effect ratio shows a perfect monotonic age gradient (Spearman rho = -1.000, p < 0.001, n = 6)- All 9 grades show FY2024 observed values exceeding the upper 95% prediction interval of beta-extrapolated counterfactuals (aggregate excess: +33%, +87,313 students)- Cumulative COVID excess: approximately 132,856 student-years over 4 COVID years All data are publicly available from Japan's e-Stat statistics portal (MEXT School Non-attendance and Behavior Problems Survey, Series 4-4-1, FY2012-2024). All annual totals match the official MEXT-published headline figures (verified: 13/13 years).