How Test Growth Models Can Hide Achievement Gaps

A test score can rise while a student remains far behind classmates who started with stronger skills. This is the central weakness of many state “growth models”: they track movement from an earlier result, but movement alone does not show whether students have reached a sound level of achievement.

For families in Australia, the issue has a familiar ring. NAPLAN reports, My School comparisons and school improvement targets often emphasise progress over time. That information can be useful, yet a rising growth score may sit beside persistent gaps linked to poverty, disability, language background, location or access to experienced teachers.

What growth scores actually measure

A growth model usually compares a student’s current test performance with an earlier result, sometimes against students with similar previous scores. A child who moves from the 20th to the 35th percentile may be described as making strong progress, even though the child is still performing below most peers.

This approach answers a limited question: how far did the result move? It does not fully answer whether the student can read complex texts, solve age-appropriate problems or meet the expected curriculum standard. A school can therefore report positive growth while a sizeable share of students remains below benchmark.

The distinction matters in New York, where debates over Common Core testing and state accountability have made score interpretation politically significant. Families seeking a plain-language explanation can review Common Core myths alongside official score reports.

Why starting points shape the story

Students begin assessments with very different levels of preparation. A child who starts well below proficiency may make several years of learning progress and still need intensive support. Another student may begin above the standard, make modest measured gains and remain comfortably ahead.

A single growth percentage can flatten those differences. It may reward a school for lifting middle-performing students while masking the number of children who have not crossed the basic proficiency line. Reporting the average can make the pattern appear healthier than the distribution actually is.

The same problem appears in Australian classrooms. A school in western Sydney, regional Queensland or the Northern Territory may serve students with very different levels of prior opportunity. Comparing their average growth without examining the starting profile risks confusing improvement with equal achievement.

How statistical models can soften inequality

Many growth systems use scaled scores, predicted performance bands or “value added” calculations. These methods can be technically sophisticated, but their outputs depend on test design, attendance, cohort size and the variables included in the model.

When results are adjusted for prior performance, a school may look highly effective because students exceeded a statistical prediction. That can be useful evidence, but it is not a direct measure of teaching quality or educational fairness. Small cohorts can also produce unstable results, especially in rural and remote schools.

Media coverage can amplify the problem when a positive percentage becomes the headline. Families may find broader education coverage useful for seeing how testing claims are framed, but the original data and definitions still deserve close inspection.

Read achievement levels beside progress

A clearer picture comes from placing growth beside proficiency, distribution and student group results. Parents should look for the share of students below, meeting and exceeding the expected standard rather than relying on one average progress figure.

Useful checks include:

• The percentage meeting the relevant achievement benchmark
• Results broken down by year level and student group
• The gap between the lowest and highest performance bands
• Whether gains continue across several testing cycles

A school with high growth and low proficiency may be doing valuable recovery work. It may also need additional funding, specialist staff and time. The growth figure should begin a conversation about support, not end it with a congratulatory label.

What Australian families should watch

Australian parents are used to seeing NAPLAN results presented through My School, school newsletters and local media. A “good result” can mean a high average, strong improvement or performance above similar schools, and those descriptions are not interchangeable.

Look especially closely at schools serving remote communities, newly arrived families or large numbers of students with disability. In Melbourne’s outer suburbs, for example, a school may show strong annual growth because targeted programs are helping students catch up, while the overall achievement gap remains wide.

When reading a report, check:

• Whether the comparison is with the state, statistically similar schools or the previous cohort
• How many students were tested and how many were absent
• Whether results are reported separately for disadvantaged students
• Whether improvement reflects one year or a sustained pattern

Plain speaking helps here: a school can be making solid progress and still have a fair bit of ground to cover. Both statements can be true.

A fairer way to judge educational progress

Growth data is most informative when combined with achievement thresholds, classroom evidence and student wellbeing. Attendance, access to qualified teachers, reading intervention, curriculum breadth and the experiences of students who opt out or miss tests all affect what a score can tell us.

New Yorkers United for Kids frames this wider concern around testing pressure, student data and local control. Australian communities face their own policy settings, but the principle travels well: public reporting should help families understand whether children are learning securely, not merely whether a statistical line has moved upward.

For a practical starting point, take one school report and record its proficiency rate, growth rate, subgroup gaps and cohort size in four separate columns before drawing any judgement.