Still on roll, slipping from view: the attendance recovery England has not completed

01 / Still on roll

The register can see a pupil clearly while education is steadily receding from their life.

A severely absent pupil can miss at least half of their possible sessions and remain on a school roll. The record is administratively visible: the enrolment exists, the sessions are counted and the absence has a code. But visibility in data is not the same as connection to education. By the time the national measure describes severe absence, the question is no longer simply whether a pupil attended. It is whether the system around that pupil has understood the barrier, found the right response and stayed involved long enough to know whether anything changed.4

That is the tension inside England’s latest attendance figures. In the definitive 2024/25 release, overall absence and persistent absence both fell from the previous year. Severe absence rose. Around 176,000 pupil enrolments missed at least half of their possible sessions, more than twice the pre-pandemic proportion.1 The national picture therefore contains two truths at once: a broad improvement and a deepening problem at the furthest edge.

02 / The edge of recovery

The average improved. The deepest absence did not.

Overall absence fell from 7.15% in 2023/24 to 6.78% in 2024/25. Persistent absence fell from 19.95% to 18.14%. Those changes matter: fewer sessions were missed across the system and fewer enrolments crossed the persistent-absence threshold. Yet severe absence moved in the opposite direction, from 2.30% to 2.39%. It increased in every English region.1

The next evidence does not support a simple continuation of the improvement. In the accredited autumn 2025/26 comparison, overall absence rose from 6.38% to 6.47%, persistent absence from 17.79% to 18.10%, and severe absence from 2.04% to 2.12%.2 The separate daily-data series estimated full-year 2025/26 overall absence at 6.93% and persistent absence at 19.05%.3

Those series answer different questions and must not be blended. The school-census publication is the definitive accredited account; the daily feed is a faster early indicator. Its coverage, code revisions and reporting lag mean individual days can be revised materially before settling, and it does not publish a severe-absence rate. The responsible conclusion is narrower than a headline declaring success or failure: the recovery remains incomplete, and its most serious edge requires separate attention.

03 / Beneath the average

One national percentage contains several different operating problems.

Absence was not distributed evenly in 2024/25. It was higher in secondary and special schools, among pupils eligible for free school meals, and among pupils receiving SEN support or with an education, health and care plan. The differences are large enough to change the questions a leadership team should ask. They are not permission to assign a cause from a category.1

A free-school-meal indicator cannot tell a leader whether the immediate barrier is housing instability, illness, caring responsibility, transport, anxiety or a damaged relationship with school. An EHC plan does not establish that provision was unavailable, unsuitable or undelivered. A recorded illness code does not reveal the diagnosis, the continuity of education or the support a family could access. The category routes the inquiry; it does not complete it.

The same principle applies to attainment. DfE’s published analysis shows an association between absence and outcomes, but it does not prove the impact of a particular attendance intervention.10 Leaders need to resist the seductive jump from “attendance matters” to “this action caused attendance to improve.” The harder and more useful question is which mechanism changed for which pupil, in which context, and whether the change endured.

04 / The missing middle

Between an absence signal and a sustained return sits a chain of trust, capacity and coordination.

The 2026 evaluation of the national attendance mentors intervention gives this middle part of the story unusual visibility. Across five local-authority areas, the programme worked with pupils facing overlapping barriers: unmet special educational needs, anxiety and other mental-health concerns, bullying, strained peer relationships, difficulties at home, poor sleep and routine, and negative experiences of school. The evaluation found that mentors could help schools understand those causes more constructively and support gradual returns, classroom adjustments or pastoral help.11

The numbers show why a policy can be clear while delivery remains fragile. Mentors typically held caseloads of 20 to 30 young people. The average wait from referral to the start of support was 43 days. Seventy-two per cent of those offered the intervention took it up, but engagement was harder where need was complex or trust between families and schools had already deteriorated. Where school capacity or communication was weak, the intervention was less effective.11

The most important finding may be what happened after support. Families and schools reported cases where improvements weakened once mentoring ended. Only a small number of schools indicated that they might adopt the approach independently; limited resource and funding were prominent constraints. The programme could create a bridge, but it could not guarantee that the surrounding system would be ready to hold the pupil when that bridge was removed.11

05 / From signal to response

The statutory approach is support-led. That makes it demanding, not passive.

The July 2026 edition of Working together to improve school attendance describes absence as a symptom and connects attendance with curriculum, behaviour, bullying, SEND, medical conditions, safeguarding, wellbeing, pupil premium, young carers and children with a social worker.5 Its logic is not “wait and understand forever.” It is to notice early, work with families, connect the right services, formalise support when needed and use legal intervention where support has not worked, is not engaged with or is not appropriate.

1. 01

   See the movement

   Track the pupil moving towards persistent or severe absence, not only the trust’s average or the point at which a threshold has already been crossed.

2. 02

   Understand the barrier

   Check the register and listen to the pupil and family before deciding what the pattern means. An absence code records a category; it rarely supplies the full explanation.

3. 03

   Connect the response

   Join attendance with curriculum, SEND, health, pastoral, early help, transport and safeguarding evidence so the barrier reaches the system able to change it.

4. 04

   Test what changed

   Record whether support altered attendance, wellbeing and access to education, not simply whether a referral, letter or meeting occurred.

5. 05

   Escalate proportionately

   Formalise support when voluntary help has not worked or been engaged with. Use enforcement only where it is appropriate and the statutory tests are met.

Two adjacent systems also need to remain visible. Persistently and severely absent pupils in the main statistics are still on roll; children missing education are, in statutory terms, not registered at a school and not receiving suitable education otherwise.6 Children unable to attend because of health needs retain a right to suitable education, including where absence is intermittent.7 A strong attendance system reconciles those boundaries instead of assuming the register alone contains every pupil at risk of losing education.

06 / The leadership view

Replace the scorecard with a movement and response view.

A board needs the rate, but it also needs to know what is changing beneath it. A trust can improve its overall percentage while a smaller cohort moves towards severe absence. It can complete every recorded action while support arrives too late, depends on one person or disappears before a return is secure. The governing question is therefore not “what did we do?” but “what changed for the pupil, and what evidence supports that answer?”

This is where the Connected Systems become operational rather than decorative. Inclusion is the primary system, but the response may depend on workforce capacity, curriculum access, funding, transport, digital continuity or safeguarding. The National Audit Office found that national data was insufficient to judge how changes in home-to-school transport affect attendance.9 A low count under one register code should therefore not be mistaken for evidence that transport is unimportant. The same discipline applies to every connected factor: combine sources, preserve their limitations and investigate the mechanism locally.

For the opening of 2026/27, the policy shift is one of clarification rather than a new national absence threshold. The July attendance guidance sharpens cohort analysis, benchmarking, reasonable adjustments, young-carer attention and system collaboration. Keeping children safe in education 2026 takes effect on 1 September; attendance remains a potential safeguarding signal to interpret alongside other evidence, not proof of harm by itself.58

08 / What we are monitoring next

The decisive evidence will be whether improvement reaches, and holds at, the deepest edge.

School Connection will revisit this analysis when the definitive full-year 2025/26 census release is published; when DfE confirms the future publication cadence of the daily series; if severe absence moves materially; or if policy changes the support and enforcement sequence. We will also watch the linked randomised evaluation of attendance mentoring, because the current report explains implementation and contribution but does not establish causality.311

The live public-intelligence panel below is deliberately separate from this conclusion. It can add an approved, verified development to the continuing evidence record; it cannot silently rewrite the editorial judgment. Machine-detected candidates remain private until investigated and approved by an editor. A live system should accelerate scrutiny, not manufacture certainty.