Public Services Productivity Review, impact of improved methods on total public service productivity: 1997 to 2021

The Chancellor of the Exchequer commissioned the National Statistician to review how productivity measures of public services could be improved in June 2023. Since then, the Office for National Statistics (ONS) has been working on the Public Services Productivity (PSP) Review in partnership with government departments, public service experts and academia, to develop better methodology and data sources to improve public service measures of productivity.

The National Statistician published their National Statistician's Independent Review of the Measurement of Public Services Productivity on 13 March 2025. This article illustrates the impact of implemented methods and data improvements to the latest annual public services productivity measures, as recommended by the PSP Review. Changes to the ONS measures of Healthcare, Education, Public Order and Safety, and Social Security Administration productivity, and to the direct measurement of labour in other service areas are covered in later sections of this release. 

Section 4: Impact on total public service productivity from 1997 to 2021 describes the impact on the total public services productivity index. It also highlights the impact on contribution-to-growth. Not all sectors are highlighted within this article; however, the data has been provided in the accompanying dataset, Public Services Productivity Review, impact of improved methods on total public service productivity.

Estimates for public service productivity in 2022 can be found in our Public service productivity: total, UK, 2022 article. We also publish timelier quarterly public service productivity estimates. Annual and quarterly estimates of public service productivity are not directly comparable since we apply quality adjustments to output in our annual estimates in order to account for changes in the quality of the services provided.

The ONS will prioritise the PSP Review recommendations for future work and continue to develop the public services productivity measures, implementing further improvements in future statistical releases.

• The combined impact of methodology and data changes introduced in March 2025 by the Public Services Productivity (PSP) Review was to slightly increase the estimate of total public service productivity growth between 1997 and 2021.

• The total public services productivity compound annual growth rate (CAGR) between 2010 and 2019 was estimated to be 0.9%, up from 0.8%.

• Revisions to Healthcare productivity growth were the largest cause of these revisions; year-on-year Healthcare productivity growth in 2021 was revised up from 10.8% to 14.5%, a partial recovery from the strong decline seen in 2020 during the coronavirus (COVID-19) pandemic.

• The sector with the largest revision was Social Security Administration, where new experimental output measures pushed up annual productivity growth between 2010 and 2019 from a CAGR of negative 0.5% to positive 1.0%.

• Other data and methods improvements increased the CAGR of Education productivity between 2010 and 2019 from 1.9% to 2.1%, and the Public Order and Safety (POS) productivity CAGR from negative 0.3% to positive 0.8%.

• Tax administration productivity was introduced for the first time as official statistics in development.

Figure 1 shows that the combined impact of methodology and data changes introduced by the Public Services Productivity (PSP) Review was to slightly increase the estimate of total public services productivity growth between 1997 and 2021. The compound annual growth rate (CAGR) for the period 2010 and 2019 is 0.9%; it was previously estimated to be 0.8%. The CAGR for the period 1997 to 2019 remains unrevised at a lower 0.2%.

There was a reduction in the fall in productivity seen during the coronavirus (COVID-19) pandemic, and a stronger recovery in 2021. The fall reduced from 14.5% to 14.1% in 2020, and estimated growth in productivity increased from 6.5% to 7.3% in 2021.

This can also be seen in Figure 2, which shows the difference in inputs, quality-adjusted output, and productivity growth between 1998 and 2021. The largest revisions were seen in 2020 and 2021, with large increases in inputs estimates but larger increases in our estimates of quality-adjusted (QA) output. The causes of these revisions are discussed in Section 5: Impact on contribution to growth.

To create the total public service productivity index, we combine the growth rate in each service area. This is weighted by each service area's expenditure share. For example, expenditure on Healthcare in 2021 made up 42% of total expenditure on public services. As such, growth in Healthcare contributes more to the total index than any other service area because it has the largest expenditure share.

Total productivity growth was revised up by 0.8 percentage points from 6.5% to 7.3% in 2021, resulting from improvements to methods and data sources. This was primarily caused by a 3.6 percentage point increase in the estimate of Healthcare productivity growth, which had a 42% weight in the basket. This was partially offset by a 3.8 percentage point decrease in the estimate of Education productivity growth because this only has a 16% weight in the basket, alongside decreases in some other service areas.

Figure 3 shows the difference in the contribution of each service area towards total productivity growth, compared with our previous publication. These revisions are explored further in Section 6: Impact on Healthcare Productivity from 1997 to 2021 and Section 7: Impact on Education productivity from 1997 to 2021, respectively.

Revisions to Healthcare were also the largest contributor to overall revisions in all years since 2018. This reflects the dominance of this service within the basket of those being delivered, alongside the material movements within this service caused by the coronavirus (COVID-19) pandemic and its aftermath.  

Healthcare and Education are broadly dominant within the back series. However, other services also present more substantial revisions in earlier periods. Public Order and Safety productivity makes a large contribution to headline revisions (both positive and negative) between 2009 and 2012. Revisions to Social Security Administration (SSA) productivity make a large contribution to revisions seen between 2015 and 2020. This is because of the implementation of an experimental new output measure for SSA, which is outlined in Section 9: Impact on Social Security Administration productivity from 1997 to 2021.

We now measure the contribution of Tax Administration to total public service productivity from 2018 onwards. These measures are labelled as official statistics in development. Tax Administration made positive contributions to productivity growth in 2020.

The improvements to Healthcare outputs and quality adjustments build on improvements to output measurement, as described in our Improved methods for total public service productivity: total, UK, 2021 methodology. We introduced additional changes to the measurement of directly measured output and quality adjustments in March 2025, as a further result of the Public Services Productivity Review.

Specifically, we have equivalised the cost weights where equivalent care is provided through alternative modes of provision, for example, equivalent hospital procedures carried out on an inpatient- or day case-basis. Additionally, changes in the amount of excess bed days no longer contribute to output growth between financial years ending 2014 and 2019.

We have improved the quality adjustment for hospital services so that the adjustment for health gain resulting from surgery accounts for changes in the remaining life expectancy of patients for non-elective procedures; this is already done for elective procedures. This has resulted in a higher quality adjustment factor being applied to non-elective treatments and has increased quality adjusted output growth.

The largest cause of revisions to quality adjustment was the refinement of the "cubic splining" approach. We use cubic splining to convert estimates of UK Healthcare output that is derived using financial year data into calendar year estimates. The substantial contractions and growth over the coronavirus (COVID-19) pandemic period has produced disparities between financial year and calendar year estimates, as well as the estimates published in our Public service productivity, quarterly, UK: July to September 2024 bulletin.

To counteract this, we have introduced the "benchmarking" of calendar year estimates against the quarterly estimates from 2019 onwards. This will improve the mapping between financial and calendar years over periods that were considerably affected by the pandemic.

Figure 4 shows the effect of these improvements. More detailed discussion of these improvements can be found in Chapter 7 of the Independent Review of the Measurement of Public Services Productivity.

The main effect of these changes was to increase year-on-year Healthcare productivity growth in 2021 from 10.8% to 14.5%. Productivity growth was also revised up from negative 25.1% to negative 24.2% in 2020, but revised down from 3.6% to 1.6% in 2019.

Improvements to Education output and quality adjustment builds on improvements to output measurement described in our Improved methods for total public service productivity: total, UK, 2021 article. We introduced additional improvements to the measurement of productivity in publicly funded Education in March 2025, as part of the Public Services Productivity Review.

For inputs, the salary weights used to calculate labour inputs have been updated to reflect more accurate salary data. This increased estimates of Education input growth from an annual increase of 1.5% to 1.6% between 1997 and 2019.

For the quality-adjusted output, we have changed three main aspects. We have introduced a new quality adjustment for further education attainment in England. We have replaced the bullying quality adjustment with a broader measure of students' wellbeing in schools. We did this using data from the British Household Panel Survey (BHPS) and Understanding Society (USoc), which replaced the BHPS.

We have also amended the "cohort split" approach, which applies weighted contributions from the current year's attainment data back to previous years. This adjustment ensures that the effects of the coronavirus (COVID-19) pandemic do not affect attainment estimates for years before the pandemic. In our two previous annual publications, we suspended the use of the cohort split approach for the academic years affected by the pandemic (from 2019/2020 onwards). A more detailed explanation of these improvements can be found in Chapter 8 of the Independent Review of the Measurement of Public Services Productivity.

Improvements to the quality-adjusted Education output increased the compound annual growth rate (CAGR) rate in quality-adjusted output from 1.6% to 1.8% between 1997 and 2019. The estimates of productivity growth over this period were also revised upwards, from a CAGR of 0.1% to 0.2%. Likewise, productivity growth was revised up from a CAGR of 1.9% to 2.1% between 2010 and 2019.

Figure 5 shows that productivity growth in 2021 was revised downwards by 3.8 percentage points, from 13.1% to 9.2%. This was because of specific improvements to the treatment of primary school attainment data over the pandemic.

There have been several improvements to the methodology and data used to estimate output in the Public Order and Safety (POS) service area. This service area includes the activity of fire and rescue services, prisons, probation services, and law courts. Law courts have five subcomponents: magistrates' courts, county courts, Crown Court, Crown Prosecution Service, and legal aid.

Improvements to POS output estimates include updates to the data and methodology used for the output of the Crown Court and legal aid services. We also used more granular data in our estimates of quantity output for prisons and probation services and in the timeliness quality adjustment used in the estimate of Crown Court output. We also removed the reoffending quality adjustment from the civil courts component of legal aid, which reflects the fact that civil courts do not hear criminal cases. A more detailed explanation of these improvements can be found in Chapter 12 of the Independent Review of the Measurement of Public Services Productivity.

Improvements to POS inputs include changes to the deflators used to deflate goods and services expenditure in prisons and court services. It also includes a change to how the labour inputs of fire and rescue services are measured, which is described in Section 10: Impact of direct labour measurement on Police, Defence and Public Order and Safety inputs.

Figure 6 shows that the effect of the changes to inputs was quite minor, while there were larger revisions to output. Revisions to output were negative during the 2000s. This is because the contribution of improvements in reoffending rates were reduced. Revisions to output were positive from 2007 onwards, reflecting the increased number of legal aid activities captured by the new methodology.

This had a large effect on productivity growth from 2010 to 2019. It increased the compound annual growth rate (CAGR) observed over this period from negative 0.3% to positive 0.8%. There were also positive revisions in 2021, as the decline in productivity was revised up from negative 3.3% to negative 2.3%.

In our previous statistical releases, we assumed Social Security Administration (SSA) output was equivalent to inputs for SSA from 2018 onwards. This is because output estimates became less robust after the introduction of Universal Credit (UC), resulting in no measurable SSA productivity growth. We introduced a new SSA output index as part of the Public Services Productivity Review (labelled as official statistics in development), as explained in Chapter 14 of the Independent Review of the Measurement of Public Services Productivity. This provided a data update for existing benefits, which are measured on a cost-weighted activity basis. This index also reflects the transition to UC from its legacy benefits, using a new benefit-weighted activity method to capture the allocative efficiencies of this transition.

We also introduced an experimental quality adjustment based on "fraud and error" rates from 2008. This adjusts output growth to account for changes in the number of overpayments or underpayments in benefits administered by the Department for Work and Pensions. Underpayments and overpayments were used to produce an overall error rate. This was then used to generate a "correctness" index (where the correctness rate is the inverse of the error rate). We then used this to adjust output to account for benefits that were correctly administered.   

The sum of improvements resulted in an increase in the compound annual growth rate (CAGR) of SSA productivity from a negative 1.1% to negative 0.5% between 1997 and 2019. The CAGR between 2010 and 2019 was revised up from negative 0.5% to positive 1.0%.

Figure 7 shows that these revisions to output methods were the largest cause of the positive revisions to SSA productivity. However, negative revisions to inputs growth also pushed productivity higher. These revisions were caused by improved deflators used to derive constant price estimates of intermediate consumption and labour inputs.

We introduced the direct measurement of labour inputs used in public services in March 2025, based on a recommendation from the Public Services Productivity Review. "Indirect labour measurement" measures the labour input as current price expenditure on labour deflated by a suitable price deflator. "Direct labour measurement" estimates the labour growth contribution of staff in employment at each rank/grade, using granular data on salary and the number of full-time equivalent staff. This allows the measure to account for changes in the quality of labour. More detail can be found in Annex F of the Independent Review of the Measurement of Public Services Productivity.

The Public Services Productivity Review has introduced new direct labour measurement for Defence and fire and rescue services within Public Order and Safety (POS). For Police services, more granular salary data for England have been introduced into the direct labour measurement and we now have greater labour data coverage for the devolved governments.

For all three service areas, revisions to the inputs index also reflect the introduction of more suitable deflators for intermediate consumption. This better reflects the mix of goods and services purchased by different service areas. We also introduced a new capital deflator for Defence.

Figure 8 shows the effect of these improvements on the inputs indices for Police, Defence, and POS. Revisions to the Defence inputs index has reduced the compound annual growth rate (CAGR) from 0.8% to 0.2% between 1997 and 2019. Over the same period, the CAGR of both the POS and Police inputs declined from 0.7% to 0.6%.

The Public Services Productivity (PSP) Review has developed estimates for Tax Administration, which are published for the first time in our Public service productivity: total, UK, 2022 article as official statistics in development. The measure covers the period 2018 to 2022.

Tax Administration was captured within the "other" service area grouping before the PSP Review. This is measured on an "inputs equals outputs" basis. Because of this, Tax Administration productivity was assumed to be constant. Output for taxes accounting for 89% of Tax Administration expenditure is now measured directly, with the remaining output calculated on the "inputs equals outputs" basis.

Figure 10 shows that Tax Administration productivity increased between 2018 and 2020, but has since declined. This variation in Tax Administration productivity growth is caused by the variance in our measure of inputs, which fell steeply in 2020. This is because HM Revenue and Customs (HMRC) resources were redirected to coronavirus (COVID-19) pandemic support in 2020. Inputs then increased in 2021 and 2022, reflecting increased resourcing of transformation and compliance functions.

Tax Administration is removed from the "other" service area grouping for the period 2018 to 2022 on a like-for-like basis. Tax Administration is still captured in the "other" grouping for the period 1997 to 2017.

Comparison of Office for National Statistics Tax Administration productivity and HM Revenue and Customs "cost of collection" efficiency 

The official statistics in development method of measuring Tax Administration productivity developed by the PSP Review maintains coherence and comparability with other PSP service areas. However, it differs substantially from HMRC's preferred measure of performance - the "cost of collection", which compares tax revenue with the cost of collecting it. 

HMRC have been using the cost of collection to measure value for money for over 10 years. This long time series is valuable in demonstrating efficiency savings over time. These estimates are published in HMRC's Annual Report and Accounts.

The cost of collection is generally presented as the cost (in pence) per pound of revenue collected. The cost of collection was 0.50 pence per pound collected in financial year ending 2022. In a scenario where tax rates go up, tax revenue increases, and the costs of collecting the tax stays the same, the cost of collection measure would fall and reflect efficiency gains. This would not necessarily be the case in the Office for National Statistics's (ONS's) productivity estimates, unless the increase in tax revenue was caused by increased numbers of taxpayers, rather than just the tax rate.  

HMRC's cost of collection approach and ONS's productivity estimates can be framed through output and inputs, but reflect different concepts in their measurement. 

Cost of collection efficiency inputs:  

• HMRC expenditure on tax collection, in current prices 

Cost of collection efficiency outputs:  

• total revenue collected, in current prices 

ONS Tax Administration productivity inputs:  

• HMRC expenditure on tax collection, converted to volume terms (that is, deflated)  

ONS Tax Administration productivity outputs:  

• revenue adjusted activity (that is, the number of taxpayers, registered traders, operators for each tax)

Figure 9 shows inputs and output for both measures, reflecting the differences outlined above.

In the ONS measure, there is greater variation over the period between the inputs and output indices. This results in a more volatile productivity index, compared with the cost of collection efficiency index where inputs and output indices follow very similar trends. The ONS and HMRC measures are shown in Figure 10.

Public services

These are services delivered by or paid for by government (central or local). This includes publicly funded services delivered by non-government providers, for example, the provision of nursery places by the private sector where these places were funded by the government.

Direct output measurement

Uses a cost-weighted activity index to estimate the non-quality-adjusted output of a service provided, such as the number of students in state schools, adjusted for attendance and weighted by school phase. This differs from indirect output measurement, where output is assumed to be equal to inputs.

Quality adjustment

A statistical estimate of the change in the quality of a public service using an appropriate metric, such as safety in prisons as part of the Public Order and Safety adjustment.

Service area

The way we refer to the breakdown of public services into nine areas, closely following the Classification of the Functions of Government (COFOG).

Classification of the Functions of Government

The Classification of the functions of government (COFOG), as explained on the Eurostat website, is the structure used to classify government activities. It is defined by the United Nations Statistical Division.