National Statistician’s Independent Review of the Measurement of Public Services Productivity

Published:
13 March 2025
Last updated:
14 March 2025

Chapter 8: Education

Education is the second largest individual area in Government spending terms and was therefore a priority for Atkinson (2005) and subsequent Office for National Statistics (ONS) research. Along with Healthcare, these were the only two services with quality adjustments by 2016. This, however, masked several known deficiencies in the measure, leading to its lack of use in policy discussion and subsequent suspension of the publication of a standalone education National Statistics bulletin.

Post the Bean Review (2016), improvement work was prioritised ultimately addressing one of the largest gaps in the application of Atkinson’s methods, namely the apportionment of attainment at GCSEs across the full educational experience of the student. This was considered a significant improvement until the advent of the coronavirus pandemic, when the model apportioned some of the learning loss exhibited during the pandemic to previous years. This, and discontinuities in the sitting and marking of public examinations, forced the suspension of parts of the method from 2019.

8.1 Scope and higher education

The Education productivity measure includes education offered from the ages of 0 to 18, through Early Years, Primary, Secondary, and Further Education. All Special Needs services are included within this, but importantly adult learning and Higher Education (HE), which covers university and those parts of colleges offering degree qualifications, is excluded.

The exclusion of HE is because there is a ‘market price’: there is a tuition fee charged to students to attend these courses. In the context of national accounts, alternative methods are applied to public services solely because no such market price is available (see Chapter 3). Given there is such a price, there is no need to consider alternative methods. Existing methods of ensuring prices are on a like-for-like basis should suffice.

However, whilst the Review excludes HE, it has observed there are two prices for HE: a regulated price set by government for domestic students and a competitively set price charged to international students. Given there is no assumed difference in the teaching output experienced by these two groups, how this is tackled is a question which can be considered by the ONS elsewhere, but the Review would encourage any consideration of this question to reflect the principles underpinning Atkinson and its work.

8.2 Core methodological and data challenges identified

The challenges in measuring education productivity are the change in the coverage of the service over time, the definition of the quality of the service, and the differences in the education system across the UK nations. In particular: capturing the expansion of compulsory education to age 18 in some nations, the increase in academisation of schools in England, and updating the treatment of healthcare and teacher training to reflect changes in provision.

The ONS uses several sources to create inputs, outputs and quality adjustments for each nation in the UK and coverage is good. However, some data sources became unavailable during the coronavirus pandemic, particularly in relation to outputs and quality adjustments. The Review has considered proxies and methodological adjustments to avoid breaks in the series.

The Review looked at quality adjustment, which is a composite of attainment and data on the prevalence of bullying, recognising this is a ‘service’ which delivers multiple outcomes. Currently, attainment is the main indicator of quality and assumes that, while if students achieve higher grades, this is an increase in the volume of output just as much as a school educating more students (and achieving the same grades).

The Review explored whether attainment (and the other quality adjusted measures currently used) should be complemented with a longer view of the impact of education towards lifetime outcomes, in particular human capital acquisition. This is discussed in the further work section and in greater depth in Annex D.

The Review also explored improvements to the ‘cohort split model’ used to allocate attainment at age 16 across the different years that cohort were in education. This is to address the point that (usually) an improvement in attainment cannot just be attributed to better teaching in Year 11 (i.e. the final year in which GCSEs are sat), but rather are a factor of the full educational career.

As previously noted, this model was unable to cope with the impact of the coronavirus pandemic, when the impact on attainment for the cohort sitting their exams in 2020 and subsequent years should be allocated only to that year. The quality adjustment section details how the impact of the coronavirus pandemic was prevented from impacting previous years.

8.3 Areas prioritised by the Review

Within the first year of this Review the ONS implemented some improvements to output:

  • Better reflecting the conversion of local authority (LA)-maintained schools to academies.
  • Capturing the increase in activity from the increase in funding entitlement for pre-primary.
  • Updating the treatment of healthcare and teacher training to reflect changes in provision.

These improvements were included in the first wave of improvements carried out as part of the Review and published in the Total Public Services productivity bulletin in March 2024.

In collaboration with the Department for Education (DfE), the Review identified further areas for improvement relating to quality adjustment:

  • Review of indicators of attainment during the coronavirus pandemic using the Office of Qualifications and Examinations Regulation (Ofqual) National Reference Test data.
  • Improvement of the adjustments applied to attainment to account for the cumulative nature of education following the coronavirus pandemic (Cohort-split model).
  • Inclusion of attainment as a quality measure for Further Education (FE).
  • Exploring if student wellbeing is a better indicator than the prevalence of bullying as a quality adjustment.

8.4 Improvements to inputs estimates

The current approach of using deflated expenditure with a combination of indirect and direct labour for Education remains the preferred methodology for obtaining a comprehensive and accurate measure of education inputs. Whilst no major methodological changes have been required in this area, some minor improvements to improve data quality and increase harmonisation with other service areas are being made in the ONS Spring 2025 annual release.

Currently, the labour inputs are created based on the full-time equivalent (FTE) teacher and support staff numbers (split by school and academies for England only) and weighted together using data on salaries from the Annual Survey of Hours and Earnings (ASHE). In the past, the ONS used salaries of all workers, while the Review recommends moving to the salary of full-time workers. This change aligns with the methodology used in other services and shields the measure from any artificial growth in salary due to a change in the relative mix of full-time and part-time workers. Movement in the full-time salary reflects only actual changes in the salary bands themselves.

Further minor remedial amendments have been made within the ASHE salary indices to increase the robustness of data by applying a compound annual growth rate to impute missing values. In using the ASHE data, the Review notes the definition of some categories of staff in education has changed since 2020. This could raise some issues for comparability over time. The Review has worked to ensure the best match of these categories and the ONS will continue to review them in the following years.

Recommendation 50:

The ONS should implement improvements identified for Education inputs in Spring 2025 related to salaries.

Recommendation 51:

The ONS should continue to review if Annual Survey of Hours and Earnings remains the best data source for labour data for Education, or whether alternative sources may be preferable.

8.5 Improvements to outputs estimates

As part of this Review, developments on Education have been adopted and implemented into PSP. These developments were implemented in Public service productivity: total, UK, 2021. These developments include:

  • The refinement of categories used to assign relative weights to different school types in output to reflect changes to the landscape of Education.
  • Improvements to pre-primary inclusion, better capturing pre-primary activity.
  • Removal of teacher and healthcare training to avoid double-counting from Further Education.

Refinement of education categories for weighting

Academies, state-funded schools that are financed by central, rather than local, government and are granted a greater degree of operational independence from local authority control, were introduced in England in 2002. Initially, the academisation of schools was concentrated almost exclusively on secondary schools. However, from 2010, the process of academisation spread to primary schools and in 2011 to special schools.

In the academic year ending 2023, just over 40% of primary, 80% of secondary and just under 45% of special schools in England were academies. With this large change in the education landscape, it is important to capture the contribution of each phase of education (primary or secondary etc) to productivity as accurately as possible, such that any further changes in the landscape will give a more accurate picture of overall education productivity.

Previously, compulsory education in England was captured through three education categories: primary, secondary and special. These are now split into a further five categories, splitting out primary, secondary, and special academies from their LA-maintained counterparts, as well as splitting out alternative provision (education for students who cannot go to mainstream schools) into two separate categories (academies and LA-maintained). Updates to the expenditure weights have allowed cost weighting of each category, allowing the contribution of each to overall education productivity to be more accurately captured.

Pre-primary education

Pre-primary education was previously captured through pre-primary (all LA-maintained pre-primary schools plus pre-primary classes in primary schools) and private, voluntary and independent (PVI) pre-primary schools.

However, expenditure data for PVI pre-primaries are not available separately from LA-maintained pre-primary expenditure. Similarly, for Scotland and Wales, there are no PVI enrolment data. The scarcity of PVI data inhibited a consistent LA-maintained and PVI split in pre-primary education across the nations, so LA-maintained pre-primary schools and PVI pre-primary schools have been included as a single pre-primary figure. This allows a more accurate cost weighting for pre-primary and therefore gives a clearer contribution to productivity.

Previously, for England, those within primary schools but who are of pre-primary age were not being captured as pre-primary but rather primary pupils. All pupils under 4 years old as of 1 September on the year prior to the school census (which is taken in the January) have been considered as pre-primary, and contribute to the LA-maintained pre-primary figure. This allows a more accurate estimate of primary-age pupils.

In the academic year ending 2014, funding became available for some 2-year-olds in pre-primary education. While included in the LA-maintained pre-primary enrolment figure, these have not previously been included in the PVI pre-primary enrolment. These enrolments are now included from the academic year ending 2014 onwards.

Similarly, in 2017, the extended entitlement of 30 hours was introduced for 3 to 4 year olds in families meeting certain eligibility criteria. The ONS implemented these improvements to capture the extended childcare entitlement in output. Note that the pre-primary inclusion for Wales, Scotland and Northern Ireland will not be affected by these changes.

For enrolment, full-time equivalence (FTE) is used to calculate activity. FTE is calculated as 50% for part-time pupils and 100% for full-time pupils. This has a small impact on primary, secondary or special schools, where there are very few part-time pupils; however, it is more of a consideration for pre-primary pupils. Where part-time pupils are not identifiable (England PVI pre-primaries, Northern Ireland PVI pre-primaries, all Scotland and Wales pre-primary), pre-primary enrolment (headcount) figures were previously given a factor of 0.5 to proxy FTE (that is, pre-primary pupils were assumed to attend school 50% of full time).

Where part-time pupils are identifiable (predominantly, the England school census and some of the Northern Ireland school census), the FTE factor is higher and growing. Therefore, where part-time pupils are not identifiable, the pre-primary figure is now given a factor equal to that of the calculated FTE factor for each year, instead of the current constant factor of 0.5. As the FTE factor is not identifiable for Wales and Scotland, the weighted average of the FTE factor for England and Northern Ireland is applied.

Teacher and healthcare training

Historically teacher and healthcare training were included within Education outputs. However, by the end of the 1990s, most teacher training colleges were absorbed into universities and have therefore been increasingly reflected within higher education measures. Similarly, when degrees became compulsory for nurses joining the NHS in 2009, the number of nurses training via non-degree routes declined. Nursing degrees, delivered through universities, are also included within the UK National Accounts through existing methods. To avoid double counting, teacher and healthcare training have been removed from Education outputs.

For more details on the effects that these changes have on output, see Improved methods for total public service productivity: total, UK, 2021.

8.6 Improvements to quality adjustments

Using the National Reference Test (NRT) to inform attainment during the coronavirus pandemic

As a result of the coronavirus pandemic, exams were cancelled in schools and typical attainment data were not available. Data were either unpublished (such as the case with England Primary schools) or informed by teacher assessed grades (such as the case with England Secondary schools) which were determined not to be appropriate to inform PSP estimates due to concerns on grade inflation.

Prior to the Review, the ONS used learning loss measures published by the DfE to inform attainment from academic year ending 2020, and the first year of the Review sought to explore alternative indicators of attainment that were consistent with historical data and provided a more robust overview of academic performance during the pandemic.

The Review identified the National Reference Test (NRT) as a more appropriate indicator of attainment during the coronavirus pandemic. This was primarily because the NRT is a strong indicator of GCSE-level performance and is independent of teacher assessed grades.

Please refer to Annex D for more information.

Attainment

In the second year of the Review, the treatment of attainment was further explored. The key issues regarding improving the quality adjustment of the Education estimates relate to adapting the ‘Cohort-split’ model in the years of the coronavirus pandemic. The Review highlighted that the ONS needed to develop adjustments to the existing model, as it apportioned some of the learning loss exhibited during the pandemic to previous years. The key elements are:

  • How to address the years prior to the coronavirus pandemic for students in school during the pandemic.
  • How to align cohorts in subsequent years to actual data, given the impact of the intervention described in the previous bullet.
  • Whether in the light of the Review the weights used in the cohort-split model needed to be updated.
  • How to handle missing data in this new model.

The adjustments made to the cohort split model were reviewed extensively in collaboration with DfE, and are described in detail in Annex D. In effect the updated model confines the effects of the coronavirus pandemic on attainment to the years directly affected by the coronavirus pandemic without apportioning any learning losses to previous years. Furthermore, the weights assigned to year groups in Secondary schools were revised so that each year group has equal weighting and only considers year groups in secondary school.

Concerning the treatment of missing data, there was a distinct lack of attainment data for primary schools across the UK for some academic years. In the absence of alternative data, the NRT was used to inform data gaps for primary schools. This was because the NRT is an indicator of academic performance, which is relevant to primary schools, and there is a concern that leaving the data gaps issue untreated could produce estimates that do not account for the broader effects of the pandemic on attainment.

Recommendation 52:

The ONS should implement improvements to the quality adjustment of Education to better account for the impact of the coronavirus pandemic and to account for student well-being and Further Education attainment.

Further education attainment

The Review also explored quality adjusting Further Education (FE) according to attainment at this stage of schooling. In collaboration with DfE, the Review identified the percentage of students meeting the minimum requirements for Level 2 and Level 3 qualifications by age 19 as the most appropriate and robust indicator of attainment at FE level. It must be noted that these relate to FE institutes in England only.

Level 2 and Level 3 attainment were individually processed according to the updated cohort-split model and were weighted and aggregated into an FE attainment index according to enrolment figures for each qualification.

FE marking practices were also influenced by the coronavirus pandemic, and no alternative data are available to indicate FE performance during this period, therefore the attainment indices for L2 and L3 will need to be held constant from academic year ending 2020 to academic year ending 2024 for Level 3, and academic year ending 2026 for Level 2. The NRT was not an appropriate indicator to inform data gaps for FE due to its link to academic performance whereas FE does not have a pure academic focus and has a shift to technical and vocational qualifications.

Please refer to Annex D for more information.

Student well-being

The ONS has historically used bullying rates as a component to inform quality adjustments for Education, however the Review identified that this does not capture the wider effects of school on students outside of attainment.

The Review identified that a measure of general student well-being would be more appropriate to inform quality adjustment, given the links between well-being and attainment and engagement at school. Areas such as bullying would also be accounted for under a well-being measure.

A student well-being measure was derived from the Understanding society’s harmonized UK Household Longitudinal survey (UKHLS), which covers households across the UK. A well-being index was prepared based on responses to ‘how do you feel about your school?’ and ‘how do you feel about your schoolwork?’.

In terms of the weighting of well-being so that it may be aggregated alongside other components such as attainment, planned annual expenditure share (%) allocated to funding pupil deprivation as declared in the DfE’s National Funding Formula (NFF), was used to assign weights to well-being.

Please refer to Annex D for more information.

The DfE are currently exploring other methods to assess student well-being in schools, and these will be kept under review to determine if more robust estimates of student well-being can be obtained.

Recommendation 53:

The ONS should continuously engage with the Department for Education over indicators of student well-being.

8.7 Devolved governments

The coronavirus pandemic had a strong impact on the availability and quality of data for all of the UK nations. The ONS is working with England, Scotland, Wales and Northern Ireland to overcome these limitations and improve the sources of data in the upcoming years.

Recommendation 54:

The ONS should work with the devolved governments to improve Education data sources as far as possible.

8.8 Recommendations for further work

Lifetime outcomes as alternative to attainment

The Review recognises the broader conceptual issue of whether attainment (and the other quality adjusted measures currently used) should be complemented with a longer view of the impact of education towards lifetime outcomes, lifetime earnings and human capital acquisition. However, there are related issues around attribution. This is discussed in greater depth in Annex D.

Recommendation 55:

The ONS should, as part of its research agenda, continue to explore the links between Education, the current Education quality measures, health expenditure and human capital acquisition, with specific attention to labour market returns, to better understand the output of these.

Grading policy

There is a second significant topic which the Review has not considered, and this emerges from one of the risks which were highlighted in Foxton, Grice, Heys et al (2019), namely arising from decisions made within a service which affect the measures used in these estimates. In this instance, the issue is grading policy.

Grading policy has varied under the different administrations in operation over the time period of these data (1997 to the present day). Essentially, GCSE grades can change for one of two reasons: a quantity effect or a price effect. Either students have achieved greater knowledge and better skills in their years of learning and are achieving an increase in the volume of output (more attainment), or there is a phenomenon of ‘grade inflation’ – the perception that students are achieving the same level of skill and knowledge from year to year so improvements in grades represents a drift in the grades awarded for any particular level of skill.

Broadly, if one believes improving grades reflect improved attainment, then this allows a larger share of students over time to be awarded the highest grades. Alternatively, if one believes improving grades is predominantly a result of ‘grade inflation’, a fixed proportion of the student population can be set who can receive the highest grade, and another fixed proportion receive the next highest grade and so on across the whole distribution.

The Education Act 2011 put in place a new regime, where the The Office of Qualifications and Examinations Regulation (Ofqual) was given the powers and duties to set grading policies and standards to deliver consistent and comparable standards. Ofqual’s approach to maintaining standards uses statistical evidence to allow grade boundaries to reflect changes in attainment over time (if they can be evidenced). Taking account of attainment at the primary stage, this establishes a broad expectation that grade distributions are unlikely to change significantly year-on-year.

The impact of this legislation (first introduced by the previous Labour administration) is that from 1997 to 2010 the number of students attaining the highest grades increased, whilst from 2012 to 2024, growth was much more muted. The result is the statistics published prior to the Review suggested faster productivity growth in education under the 1997 to 2010 Labour administrations than under the 2010 to 2024 Conservative administrations, although the Review considers this an unhelpful simplification which does not recognise the complexity of the policy landscape and the impact of the Education Act 2011.

It is, nevertheless, a concern to the Review that the change in regime and the creation of a more robust grading policy framework under Ofqual may be misinterpreted as a change in productivity trends. As it is a long-standing feature of the statistics and there has been no concern raised externally on this topic, the Review has not proceeded to tackle this question, but it may be an issue academic researchers may wish to interrogate.

Recommendation 56:

The ONS should engage with academic researchers and stakeholders to understand if the impact of grading policy on Education productivity is substantial enough to warrant further research.

Back to top