School-Level Variation — How Wide Is the Within-Board Gap?
Within any given board, how much do schools vary? Is most of the variation between boards or between schools within the same board? How many schools are persistently low-performing?
Variance Decomposition — Between-Board vs Within-Board
The intraclass correlation (ICC) measures what share of total variance in school L3/4% lies between boards. A high ICC means boards differ systematically; a low ICC means most variation is between schools within the same board — and board-level policy misses the point.
ICC = between-board variance / total variance. Higher ICC → more variation is systemic (board-level). Lower ICC → variation is school-level.
Within-Board School Distribution — 2024–25
Each dot is one school. Boards with wider spreads have more internal inequality. Select a board to highlight it, or view all boards ranked by median.
× marks indicate suppressed data. Schools with fewer than 15 assessed students are suppressed in the source data — their metric values are withheld. These schools are shown as × at x = 0 rather than silently omitted.
Boards ordered by median L3/4%. Red dashed line = province average. Box = IQR. × = suppressed (n < 15). Hover for school details.
Within-Board Spread Rankings
Which boards have the most internal inequality? IQR and range of school L3/4% within each board, 2024–25.
Sorted by IQR (interquartile range). Red = IQR > 20pp, Orange = 15–20pp, Teal = < 15pp.
Within-Board Spread Over Time
Is within-board inequality growing or shrinking? Average IQR and SD across all boards, by year.
Persistently Low- and High-Performing Schools
Schools that scored below the provincial average in all 4 years are "persistently low." Schools above in all 4 years are "persistently high." The middle group moved across the threshold at least once.
Based on schools with data in at least 3 of 4 years. "Persistently below" = below provincial average in every year with data.
Bottom 20 Persistently Low Schools
Turnaround Schools — Biggest Improvers
Schools with the largest positive change from their earliest to latest year. These are potential case studies for what drives improvement.
Small-School Volatility — Does Size Predict Noise?
Schools with fewer students tend to show larger year-over-year swings — a well-known statistical artifact (small sample sizes = noisier estimates). This scatter shows school size vs magnitude of year-over-year change.
Each dot is one school-year transition. Red line = linear regression. Smaller schools (left) tend to show larger swings. Log scale on x-axis.