Achievement Adjusted for Family Background
How much of the variation in school achievement is explained by family/community background? Which schools outperform expectations given their student demographics?
Raw achievement scores largely reflect the communities schools serve, not school effectiveness. A school in an affluent suburb scoring 85% L3+4 isn't necessarily better than one in a newcomer neighbourhood scoring 55%. This page uses student questionnaire responses as proxies for socioeconomic context, then identifies schools that beat — or miss — their predicted scores.
What Can We Proxy?
EQAO doesn't collect household income or parental education. But the student questionnaire includes signals correlated with socioeconomic context:
| Proxy | What it captures | Direction |
|---|---|---|
| Other first language | % whose first language ≠ instruction language | Higher → more newcomer/multilingual |
| No computer access | % who never/rarely use a computer at home | Higher → fewer household resources |
| No internet access | % who never/rarely have strong internet | Higher → connectivity gap |
| Multilingual home | % speaking another language ≥ as often as English | Higher → more linguistically diverse |
These are imperfect proxies based on self-reports by 8–11-year-olds. They correlate with socioeconomic context but are not direct measures of family income or education. Treat as rough indicators.
How Each Proxy Relates to Achievement
Each dot is one school. The regression line shows the average relationship — steeper slopes mean the proxy explains more variation.
Correlation Summary
Negative correlations are expected: more disadvantage → lower achievement. r² shows the share of variance explained by each proxy alone.
Composite Disadvantage Index
We combine the available proxies into a single "disadvantage index" by z-scoring each proxy and averaging. Schools with higher index values serve communities with more socioeconomic challenges.
Each dot is one school, colored by its residual (blue = outperforming expectations, red = underperforming). The regression line shows the average relationship between community disadvantage and achievement. Schools far above the line are adding more value; schools far below are underperforming their demographic context.
Value-Added Rankings — Who Beats the Odds?
The residual from the regression above is a rough "contextual value-added" estimate: how much better (or worse) a school performs compared to what we'd predict from its demographic profile. This is not true value-added (which requires tracking individual students over time), but it adjusts for measurable intake differences.
Blue bars (right) = schools scoring higher than predicted by their demographic profile. Red bars (left) = schools scoring lower. Hover for details.
Value-Added Table
How Much Does Adjusting Change the Rankings?
This scatter compares each school's raw rank (by L3/4%) to its adjusted rank (by residual). Schools near the diagonal kept similar positions; schools far from it were substantially re-ranked.
Each dot is one school. Diagonal = no change in rank. Red dots moved more than 20% of total positions. The more scatter off the diagonal, the more adjusting for demographics changes our picture of school effectiveness.
Stability Across Years — Is Value-Added Noise or Signal?
A key question: if a school looks like an outperformer in one year, does it still look like one the next? If residuals are stable, the signal is real. If they bounce around, it's mostly noise.
Each dot is one school appearing in two consecutive years. If value-added is mostly signal, dots cluster along the diagonal (positive correlation). If it's noise, they scatter randomly. The pooled correlation across all year-pairs indicates how much a school's "outperformance" persists.
Board-Level View — Aggregate Disadvantage vs Achievement
Each dot is one school board. Boards in the upper-left (low disadvantage, high achievement) are in favourable contexts; boards in the lower-right face greater challenges.
Boards above the regression line are outperforming their demographic context at the aggregate level. This is suggestive but less precise than school-level analysis (boards contain diverse schools).
Methodological Notes
What this is: Contextual value-added — we adjust for measurable student demographics at the school level. This reveals which schools perform better or worse than we'd expect given who they serve.
What this is not: True value-added requires tracking the same students over time (e.g., comparing a student's G3 score to their G6 score). EQAO cross-sectional data can't do that. A school might appear to "add value" because it attracts higher-performing transfer students, not because of its teaching.
Proxy limitations:
- Self-reported by 8–11-year-olds — "Do you have a computer at home?" is noisier than census data on household income
- Technology access may partly reflect parental screen-time rules, not just household resources
- Home language captures linguistic diversity but not directly socioeconomic status — multilingual professionals and recent refugees both speak other languages at home
- The composite index explains only a fraction of the total variance (see r² above)
What would make this better: Linking school-level data to census tract demographics (median income, % university-educated, % recent immigrants) would provide much stronger SES proxies. The Ontario Learning Opportunities Index (LOI) does this for TDSB schools.