Is your data measuring children, or birthdays?

Updated 25 September 2026Version 1.0Stands on The State of the Evidence, chapters 6, 7 and 9By Shilpa Kapur SinghCite this

The current answer

Some of what a school records as a child’s deficit is made by the school. A cut-off date decides who is youngest. A language label is read as ability. An identification route follows confident parents. Each one can be removed by the same lever that made it.

Start with the cheapest test there is: break your referrals, sets and more-able list down by month of birth. If the youngest are over-represented, your data is partly measuring birthdays.

How a birthday becomes a label

The months are arithmetic, not a survey: in England the school year starts on 1 September, so a child born in late August joins the same class as one nearly a year older.

A cut-off date sets who is youngest.Nobody chose which children would be eleven months behind their classmates. The calendar did.
The youngest are judged against the oldest.At four or five, eleven months is nearly a fifth of a lifetime. A normal gap in maturity reads as a gap in ability.
The judgement becomes a structure.A lower set, a referral, a place off the more-able list. Each is recorded against the child.
The structure outlasts the gap.The maturity gap closes with time. The set, the label and the expectations often do not.

The scale of it

The research project

countries in which the youngest in a year group scored lower in international maths and science tests, with the gap still visible years later.

Bedard & Dhuey, 2006, TIMSS

21.6%

of pupils in England are recorded as having English as an additional language. The label records a home language, not how well a child speaks English, and it is English proficiency that attainment follows.

DfE, 2025/26; Strand & Hessel, 2018

718,838

education, health and care plans in England in January 2026, up 12.5% in a year. When identification grows this fast, how a school identifies matters as much as whom it finds.

DfE, January 2026

What to do first

Run the birth-month query

Break referrals, sets and more-able identification down by month of birth. It costs one query and settles the question for your school.

Separate the language judgement from the ability judgement

Where one person judges both at the same moment, split them: different instruments, ideally different people.

Treat universal screening as an experiment

Pilot it, measure who it finds that referral missed, and decide from your own data.

Decide what you stop attributing to the child

Write down which recorded gaps you now think the system made, and change the structure, not the child’s file.

Tool · DT-002

Is your school manufacturing disadvantage?

Twelve questions across three places where schools most often record a design choice as a child’s deficit. You get a profile by section and one first move.

About five minutes. Nothing is stored or sent.

What would change our mind

Birth-month gaps in referral and grouping that disappear once a school controls for them properly, across several systems.
Universal screening that finds no one referral had missed.
Evidence that placing English learners by language status carries no later cost to their attainment.

Where this comes from

Evidence brief EB-002
White paper WP-002
Self-audit DT-002
The State of the Evidence, ch. 6, 7 and 9

What changed

25 Sept 2026
Published as a living dossier, version 1.0.
June 2026
Evidence brief EB-002 and white paper WP-002 published.
May 2026
First published in The State of the Evidence.

How to cite

Kapur Singh, S. (2026). Is your data measuring children, or birthdays? Living dossier, version 1.0, 25 September 2026. The EdJournal.