Schools That Beat the Odds

Which schools significantly outperform what their students' family backgrounds would predict — and do the schools that beat G6 achievement expectations also add more G3→G6 growth than expected?

Raw achievement scores largely reflect the communities schools serve. A school in a high-income neighbourhood scoring 85% isn't necessarily "better" than one in a newcomer community scoring 55% — it may simply have more advantaged students. This page strips away the demographic signal and asks: once we account for socioeconomic context, which schools over- or underperform, and along which dimensions?

Two SES-adjusted dimensions. An OLS regression predicts each school's outcome from five demographic predictors drawn from the Ministry's School Information Finder: % low-income families, % parents without a post-secondary degree, % students whose first language isn't the language of instruction, % recent newcomers, and a language-of-instruction dummy (English vs French). Residuals — the gap between actual and predicted — measure contextually-adjusted performance. Both dimensions are standardised to SD units so they can be compared on the same scale.
Ecological inference. The demographics come from Ministry enrolment files matched across years; missing or mismatched years reduce coverage for some schools. The G3→G6 value-added is a school-level estimate — it captures "how well do students in this school's typical cohort progress?" not a direct student tracking. Treat individual school positions as indicative, not definitive.

Find a school

Two dimensions: G6 achievement vs. value-added

Quadrant breakdown

G3 vs. G6 achievement (both SES-adjusted)

Schools that outperform demographic expectations at G6 tend to do so at G3 as well — performance relative to context is largely persistent across grades. The high correlation here confirms a stable school-level effect rather than a grade-specific artefact.

Performance advantages persist across grades. r ≈ 0.59 between G3 and G6 SES-adjusted scores means that schools which outperform their demographic expectations at Grade 3 tend to do so at Grade 6 as well, though there is meaningful variation — roughly a third of the variance is grade-specific. The dashed 45° line shows where G3 and G6 advantages would be equal; the regression line sits close to it, indicating no systematic grade-level drift in the advantage.

Top outperformers: both dimensions above expectation

Schools in the "Beats both" quadrant, ranked by the sum of their G6 achievement and value-added z-scores, filtered to those with ≥ 15 average G6 students per year.

How to read this page

Concept What it means
SES-adjusted The demographic model predicts a school's expected score from five indicators. "SES-adjusted" = actual minus predicted.
SD units Residuals expressed in standard deviations. +1 SD ≈ top 16%; −1 SD ≈ bottom 16%.
G6 achievement (SES-adjusted) Beats the prediction for weighted-average G6 L3/4% across Reading, Writing, and Math (2023–25).
Value-added (SES-adjusted) Beats the prediction for Bayesian G3→G6 VA — a school whose students make more Grade 3→Grade 6 progress than other schools with similar demographics.
Beats both Top-right quadrant: above expectation on both attainment level and growth rate. These schools are not just serving advantaged students — they are adding value on top.
r ≈ 0.72 Strong-moderate correlation between the two measures. A school that beats G6 expectations tends to also add more VA than expected, but not always — the two capture different things.
Small school caution. Schools with fewer than 15 students per year are filtered out by default. Their residuals are noisy: a single cohort's fluctuation can look like a large over- or underperformance. Bayesian shrinkage already pulls VA estimates toward zero for small schools, but the WLS residuals are not shrunk. Use the filter slider to explore sensitivity.