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

Published:
13 March 2025
Last updated:
14 March 2025

Chapter 13: Tax administration

Tax Administration accounts for approximately 0.6% of total public expenditure on services captured within the scope of the Review (therefore excluding transfers). It reflects government activity in the administration and collection of taxes (such as Income Tax, National Insurance, Value Added Tax (VAT) etc.

By far the largest contributor to this service is HM Revenue and Customs (HMRC), with other activity provided by local authorities in terms of collecting Council Tax. However, not all HMRC activity is included, for example Tax Credits and Child Benefit relate to Social Security, rather than Tax Administration, although there are challenges in identifying these in the available spend data.

Tax Administration is currently consolidated within the “Other” grouping in the Office for National Statistics (ONS) public services productivity (PSP) statistics. This “Other” grouping contains only services calculated on the “inputs = outputs” basis and as a result, Tax Administration productivity is currently constant by assumption.

The Review has developed a provisional method focused on HMRC taxes in the first instance to directly measure tax administration output, and hence productivity for the first time. Although relatively small by expenditure weight, exploring this area alongside social security administration has provided the Review valuable insights which can be applied more widely in terms of how to best address administrative functions of this type.

As the Tax Administration method is new, the ONS determines this measure to be ‘Official statistics in development’. Once this has undergone further testing, it will become an ‘Accredited official statistic’ in line with the rest of the estimates.

The Coronavirus Job Retention Scheme was launched in April 2020 and aimed to protect jobs affected by the coronavirus (COVID-19) pandemic. This is not related to taxation as a service and expenditure is therefore recorded within “Other” groupings. Further consideration of this matter is advised in relation to whether this scheme should be considered social security administration.

13.1 Core methodological and data challenges identified

The Review has collaborated with academics and HMRC experts to develop innovative approaches to measuring Tax Administration productivity, whilst ensuring comparability with other services and adhering to UK National Accounts standards.

Inputs

Tax Administration activity, as with other services discussed in this Review, suffers from sharing a Classification of the Functions of Government (COFOG) code 1.1.2 (Financial and Fiscal Affairs) with other non-tax activity such as the management of public funds and public debt. In terms of accessing data on inputs, it is therefore not independently identifiable or measurable in the national accounts, despite being an essential service undertaken in some form by almost all governments, often by a distinct body with responsibility for tax collection alone (rather than wider public finance management).

The Review has considered viable alternative data sources to measure inputs, and the ONS has separately proposed that tax administration be considered as a potential separate category within the COFOG structure in its response to the United Nations Statistical Commission’s consultation on updating the COFOG. As with other services, implementation of any such change would need to be coordinated and funded across government.

Importantly, once a data source is identified, the scope of activity for which equivalent output metrics needs to be considered and input split accordingly.

Outputs and outcomes

Alongside gaining clarity on inputs, the Review has reflected on how to best assess measures of output, which enable productivity to be measured over time. To achieve this the Review has identified the need to balance two perspectives. Firstly, the need to identify a suitable measure for changes in the amount of service activities delivered and weights for these. Secondly, a reflection of the outcome of relevance from Tax Administration, which is to raise the right amount of revenue to fund government public services whilst spending as little to achieve this as possible.

The key challenge in this area therefore relates to traditional metrics of activity, such as the number of taxpayers’ casefiles processed poorly reflecting the outcome desired, which is the correct collection of tax. Some tax-files may result in no tax being paid, and the relationship between cost and taxes collected fundamentally relies on the design of the tax regime. Real Time Information Pay as You Earn (PAYE) is a relatively low cost form of collecting income tax and national insurance, whereas Self-Assessment is materially more expensive. Conversations with HMRC revealed their operational preference and experience in using a ‘cost of collection’ model, focused on the expenditure per pound of tax collected. The Developing outputs estimates section discusses this in greater detail.

One of the key issues arising however is the bounded nature of the desired outcome. Taking Healthcare as a counterfactual, if more operations are delivered, this would be perceived as an increase in output; each additional operation may have diminishing marginal utility, but as citizens are unlikely to volunteer for operations they do not need, one can assume there is no upper bound on demand. This is not the case for Tax Administration. The tax regime is designed to deliver a certain value of tax which is considered necessary by the fiscal authority to achieve macroeconomic aims. If HMRC were to collect too little tax compared to the theoretical liabilities (the ‘tax-gap’) or too much (effectively ‘over-taxing’ citizens), either of these outcomes would be viewed as having social detriment.

Therefore it cannot be the case that if HMRC was to continuously increase the tax collected as a fraction of Gross Domestic Product that this would be seen as a positive outcome, if this was not the mandate set for it by the fiscal authority. In other words, unlike private activity where, as long as there is a positive price extra output is always ‘a good thing’, HMRC could over-supply tax revenue and the fiscal authority would, if it was in the position of paying for this service, not be willing to pay for this – the market equivalent price for this activity would be zero and hence the additional output would have no value. As such it cannot be the case that the raw tax take can be used to quality adjust direct measures of output.

This Review therefore proposes an additional model to ‘revenue-adjust’ direct measures of output in lieu of a more complete quality adjustment which the Review has begun to develop with HMRC, but did not reach a state of suitable maturity for implementation at this time.

Recommendation 93:

The ONS should ‘revenue-adjust’ direct measures of Tax Administration output whilst developing, with HMRC, appropriate quality adjustments, including using factors such as the ‘tax gap’ to reflect the difference between the desired and achieved levels of tax collected, customer satisfaction and call waiting times.

13.2 Developing inputs estimates

The following data sources were considered but the Review deemed them not suitable for PSP purposes due to a lack of specificity in their coverage.

  • The ONS’ UK National Accounts’ data: It is not possible to measure tax administration inputs via this traditional method because it does not have a unique COFOG code. Therefore, alternative sources were explored in developing a suitable approach to measuring Tax Administration inputs for PSP.
  • OSCAR (Online System for Central Accounting and Reporting): this cross-government management information system provides HM Treasury with key spending data from each department. For the Review’s purposes, it was deemed unsuitable because of its inability to differentiate HMRC expenditure on Tax Administration activity from other financial and fiscal affairs within COFOG 1.1.2.
  • HMRC Annual Report and Accounts: these accounts did not allow the Review to specifically identify HMRC’s Tax Administration activity, and detailed input breakdowns were also not fully or appropriately defined.

HMRC Cost of Collection as a data source

Expenditure data used in HMRC’s ‘Cost of Collection’ statistics was the only identified source that provided suitable coverage on Tax Administration activity. ‘Cost of Collection’ statistics are published annually in the HMRC annual reports and accounts and are presented as the ratio of Tax Administration expenditure per pound collected in tax revenue.

The expenditure component is calculated by allocating expenditure to Tax Administration functions based on reported staff time spent on those activities. Only the ratio is published (available via the ‘Historical data series’ file), with the expenditure component being provided directly to the ONS.

Cost of collection expenditure figures do not provide a breakdown by input components, so OSCAR data are used to estimate the proportion allocated to labour, intermediate consumption and consumption of fixed capital within this expenditure. Input breakdowns in OSCAR are preferred over national accounts data because it can be assumed that Tax Administration is more homogenous to wider HMRC activity than it is to wider activity under COFOG 1.1.2.

Due to the availability of these data the new Tax Administration productivity measure begins in 2018, and the ONS plans to publish these from Spring 2025 as part of the Public service productivity: total, UK. An equivalent amount of expenditure will be removed from the “Other” grouping series from 2018 to avoid double counting.

Recommendation 94:

The ONS should publish provisional estimates of Tax Administration productivity in Spring 2025 and adjust the “Other” grouping accordingly.

13.3 Developing outputs estimates

HMRC has worked closely with the Review, and shared information on the range of performance indicators published in their Annual Report and Accounts, including cost of collection which is their preferred measure of the productivity of taxation. The inverse of the cost of collection, defined as the ratio of revenue collected over Tax Administration expenditure is a measure of the productivity of Tax Administration.

The Review decided not to use this measure because it relates the outcome directly to the inputs (the spend) without utilising as with other services a direct measure of output. Whilst there is a precedent for this in adult social care, this was not viewed as an optimal approach, in part because tax revenue is viewed by the national accounts as a ‘transfer’ rather than value added and using the value of a transfer as the measure of output of public (and private) industries is not permitted.

In consultation with experts in HMRC, the measure developed by the Review uses an estimate of the number of taxpayers for each regime as a direct measure of output, a similar approach to that already used for social security administration, where the number of benefit claims and caseload processed is used as the activity measure. However, it should be noted this approach prices in ‘false positives’ into the output measure. That is, there are a number of casefiles processed where no tax is collected due to the circumstances of the tax-payer. For example, PAYE will automatically report on all staff in a payroll to HMRC, not just those eligible to pay tax in the relevant time period. Equally whilst all self-employed workers will complete a Self-Assessment, whether they have any additional tax to pay under that regime will depend on their activity in the year and their eligibility for tax reliefs etc.

This ‘over-head’ of ‘false positives’ is obviously dependent on tax policy (e.g entitlements and thresholds) and clearly in this context changes in such policies would result in productivity falling or rising dependent on whether the number of ‘non-tax-generating’ casefiles increases or falls. Whilst the Review accepts this ‘overhead’ is a necessary cost of ensuring tax is properly collected, its impact on productivity estimates is an issue which would merit further research once statistical estimates are being created.

Importantly, therefore, this direct measure of output requires a formal treatment to relate it to the desired outcome, which is the collection of tax.

The Review has been able to source direct output coverage of eleven taxes accounting for a broadly consistent 88% to 89% of total HMRC Tax Administration expenditure for the period 2018 to 2023 using the HMRC’s ‘Cost of Collection’ measure in the calculation of inputs. This covers Income Tax, VAT, Corporation Tax and National Insurance Contributions and several other taxes.

Table 7: Taxes included and basis of measurement

TaxActivity measure
Income TaxNumber of taxpayers
National Insurance Class 1 (employees)Number of taxpayers
National Insurance Class 4 (self-employed)Number of taxpayers
Corporation TaxNumber of taxpayers
Capital Gains TaxNumber of taxpayers
Value Added TaxNumber of registered traders
Insurance Premium TaxNumber of registered traders
Air Passenger DutyNumber of operators
Landfill TaxNumber of operators
Climate Change LevyNumber of operators
Aggregates LevyNumber of operators
Inheritance TaxLiabilities upon death (Estates eligible for Inheritance Tax) plus Lifetime Transfers

Source: HMRC publication “Number of taxpayers and registered traders”, with the National Insurance component being provided directly to ONS.

 

Unit costs can be calculated by dividing the expenditure on administering each tax by the number of taxpayers of that tax. Detailed cost of collection data has been supplied by HMRC to the ONS, which will enable the calculation of unit costs for each of the eleven taxes in Table 7 (treating Class 1 and Class 4 National Insurance as two instances of a single tax). This may be a topic worthy of further investigation as Income Tax is collected via PAYE and Self-Assessment and National Insurance is collected via the same two channels, which may be better understood in a more disaggregated fashion.

Recommendation 95:

The ONS should work with HM Revenue and Customs to continue to ensure data on different taxes is utilised in as granular a fashion as benefits productivity estimates.

There are a number of taxes without activity data, which account for around 11% of HMRC administrative expenditure in recent years. These are treated on an “inputs = outputs” basis at this time due to data availability. The inclusion of this “inputs = outputs” component enables the output measure to account for changes in the administration of taxes that lack activity data. However, as the productivity of the administration of these taxes cannot be measured, the inclusion of the “inputs = outputs” component has a slight ‘dampening’ effect on productivity, slightly reducing the size of productivity changes.

While this measure covers nearly 90% of the output of tax collection, it does not cover all tax-related administrative activity undertaken by HMRC. In particular, it does not cover customs, borders and trade activity and tackling non-compliance (i.e. tax evasion and avoidance). As explained in Chapter 14 the primary non-tax elements of the HMRC portfolio, such as Tax Credits and Child Benefit are accounted for in social security administration. The developed metric currently also omits local authority functions in relation to council tax and business rates. Initial investigations identified data on customs duties and import taxes and the ONS intends to further investigate the possibility of expanding to full coverage in these services.

Recommendation 96:

The ONS should explore how to complement existing HM Revenue and Customs data to complete coverage of Tax Administration activities.

These metrics are combined using a cost weighted activity index. Importantly, as with other areas, and as discussed in Chapter 3, cost weighting is widely used and is generally the only feasible method, but that does not make it conceptually strong.

The Review has devoted significant resource to understanding how best to address a problem also observed in other services and discussed in Chapter 3 of this report. Namely that the relative costs of undertaking various public services are only a proxy for the value of that activity. In short, this approach captures well the changes to the rate of technical efficiency (productivity gains from improving existing processes) but struggles more when trying to capture allocative efficiencies (gains from changing to an entirely different process). As such this chapter discusses how to take account of tax revenue.

13.4 Quality adjustment

The outcome which matters in relation to Tax Administration is the collection of the right amount of tax, as defined by the decisions taken by the fiscal authority. However, tax revenue is viewed by the UK National Accounts as a ‘transfer’ rather than value added and transfers are excluded from the measurement of output of public (and private) industries. Therefore, the Review has had to consider the most appropriate way to balance the inclusion of tax revenue whilst not breaching national accounting rules if the ONS wishes these metrics to flow into the national accounts data.

This issue has traditionally been avoided by focussing on the administration of taxation. This administration function has been included as part of general government final consumption expenditure and contributes to overall economic output in the UK National Accounts and therefore the PSP measures, and the purpose of this Tax Administration is to collect revenue for government. However, this approach fails to account for quality change in the service.

The Review has reflected on the gap between a measure of administrative activity, in terms of the number of taxpayers and the concept of revenue collected against the tax plans of the fiscal authority.

Pending this process, the Review has developed a temporary adjustment to direct output. This applies a ‘revenue adjustment’ to the output index that weights tax administration activity by the revenue raised from different taxes, divided by the spend to collect this revenue. Whilst this uses a simple measure of tax revenue it enables a better reflection of the relationship between taxpayers, spend per taxpayer and revenue per taxpayer. This enables the measure to account for allocative efficiency, whereby a shift towards lower cost-per-£-raised taxes will increase productivity.

Data for revenue adjustments can be sourced from the publication HMRC tax receipts and National Insurance contributions for the UK which provides the receipts by financial year for the taxes covered in this analysis.

Although applying a revenue adjustment is in keeping with the spirit of HMRC’s preferred measure of value for money, which is the ratio of tax revenue relative to the cost of collecting it (referred to as the “cost of collection”), other aspects of quality in tax collection are not currently addressed by the measure. The ONS intends to investigate further developments that could be made to account for changes in quality. In particular, the ONS is interested in exploring the possibility of using the “tax gap” as a quality adjustment. This is the difference between the theoretical amount of revenue that should be collected by HMRC and the amount that is actually collected.

Recommendation 97:

The ONS should investigate further developments that could be made to account for changes in quality such as fraud and error, customer satisfaction and waiting times, but recommends that treatment of the ‘tax gap’ should be the priority area for focus.

13.5 Devolved and Local Government

The developments in this Review reflect tax collection by HMRC only, excluding Customs. Locally administered taxes and taxes collected by devolved governments have not been prioritised.

13.6 Recommendations for further work

While substantial progress has been made in the Review, the measures developed only cover taxes administered by HMRC, excluding Customs Duties. Locally administered taxes, such as council taxes and business rates, are not currently included. Further work to expand the scope to all taxation activities and to consider quality adjustment of these areas will depend on the exact form of the direct measures of output which are chosen for these.

Recommendation 98:

The ONS should review data on taxes administered locally and in devolved governments in relation to inputs and outputs, to explore the feasibility of further improving coverage of Taxation Administration.

In addition to further work to develop coverage and quality adjustment, the time period covered by Tax Administration productivity measure should be considered. The ONS’ new initial Tax Administration productivity measure will begin at 2018. This is due to limited input data being available prior to this, however it may be possible to extend this further to 2010 if Customs were to be included and aspects such as weights estimated using more recent data. The ONS should continue to work with HMRC to investigate data sources and potential solutions to this.

Recommendation 99:

The ONS should investigate the feasibility of extending the time period for Tax Administration productivity measure to pre-2018.

Back to top