This chapter looks at the resources devoted to tax administrations and provides information on their workforce. It also sets out how administrations are responding to new challenges and maintaining their capability while managing a workforce that has to adapt to a changing work environment.
Tax Administration 2025
Comparative Information on OECD and other Advanced and Emerging Economies
Abstract
Central to a tax administration meeting its role in collecting revenue and providing services to citizens and businesses, is sufficient financial resources and a skilled workforce that can deliver quality outputs efficiently and effectively. This chapter examines the financial resources available to tax administrations, and how they are spent. It also provides information on tax administrations’ workforce, and how working practices are changing.
Operating expenditures
The overall level of resources devoted to tax administration is an important and topical issue for most governments, external stakeholders, and of course tax administrations themselves. While the budgetary approaches differ, in most jurisdictions the budget allocated is tied to the delivery of performance outputs which are outlined in an annual business plan.
When looking at the budget figures, around 80% of tax administrations report an increase in their operational expenditure between the years 2022 and 2023. This is similar to the two previous periods (see Table 9.1.).
Percentage of administrations
Note: The table is based on the data from 53 jurisdictions that were able to provide the information for the years 2018 to 2023. Data for India, Israel and Türkiye have been excluded, see notes in Table A.16.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table A.16 Tax administration expenditures: Operating and salary expenditure, http://isoradata.org (accessed on 1 October 2025).
However, this data should be treated with caution. While on paper a significant number of administrations saw increases in their budget, this may be a result of added responsibilities that come with additional budget as well as additional funding for investments in technology and new services to respond to taxpayer demands. It also does not take into account increases as a result of inflation, and that a significant part of the budget is needed for salary costs, accounting for on average nearly three-quarters of operating budgets annually (see also Table D.6). Any increases in budgets can be rapidly consumed by salary increases, which may be a contractual obligation.
Components of tax administration operating expenditure
As stated above, the largest reported component of tax administration operating budgets is staff costs, with salary alone accounting for on average 73.5% of operating budgets annually, even though there are some differences among jurisdictions (see Figure 9.1.).
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.6 Resource ratios: Cost, http://isoradata.org (accessed on 1 October 2025).
Another important component is the operating cost for information and communication technology (ICT). On average ICT expenditure accounts for about 12% of operating expenditure. However, reported levels of ICT expenditure vary enormously between administrations. For those administrations able to provide ICT-related cost, 50% reported an annual operating ICT expenditure exceeding 10% of the administration’s total operating expenditure in 2023 (an increase of 10 percentage points compared to 2022) and another 23% reported figures between 5% and 10%. (See Table D.6)
Although the introduction of new technologies can be costly, it can bring significant benefits in efficiency and productivity. The examples in Box 9.1. set out some of the initiatives being adopted by tax administrations.
Finland – Real-time economy project
The Real-Time Economy project (2021–2024) was introduced by the Finnish Tax Administration to advance Finland’s digital economy. It focused on automating the transfer of financial information between companies, reducing manual input from officials and achieving significant cost savings.
The key objectives and results were as follows:
Infrastructure: Conditions were established for financial data sharing and digital wallet services. The Finnish Peppol authority was created to promote international Peppol infrastructure (a secure network for exchanging electronic invoices). Central government capabilities for processing digital procurement messages and receiving digital financial statements were developed.
Cost savings: Digital documents were projected to save nearly EUR 6 billion annually by 2030, with centralised exchanges of data producing EUR 430 million in annual benefits. The project increased the use of e-invoices by 93% and e-receipts by 20%.
Digital solutions: Structured business documents, digital wallets, and data sharing were tested in trials. Significant benefits were found for businesses and the tax administration. As a result, these were implemented and guidelines were produced, training organised, and marketing campaigns ran.
Interoperability: Solutions were designed that are compatible with international and EU standards, to enhance competitiveness and increase interoperability.
Ecosystem development: Developed a joint management model for continuous real-time economy development, defining roles for participants and cooperation principles. Legislative and information security studies were conducted.
Vision 2030 envisions a fully digitalised business economy ecosystem in Finland, enhancing predictability, risk management, and competitiveness for businesses, and improving efficiency and services for the administration. A roadmap for 2025–2030 has been set out, outlining steps for expanding the real-time economy.
Ireland – Streamlining tax and duty manuals
Revenue has introduced a new project, RevAssist, to streamline the management of its Tax and Duty Manuals (TDMs). There are approximately 1 500 Tax & Duty Manuals (TDMs) which summarise legislation and operating procedures, as well as providing guidance covering all the taxes and duties that Revenue is responsible for administering. These are publicly available to the tax practitioner community, and are complex to write in terms of the breadth of legislation covered. A high level of effort is required to ensure the accuracy of these manuals, and they need to be regularly updated to keep them up to date when legislation changes.
Therefore, actions were taken to streamline this process, including:
The use of emerging GenAI technology to improve the TDM process.
The provision of a simplified user interface to allow staff to easily search their queries across the TDMs and get a summarised response to the query, including references and page numbers of the full TDM documents.
Generate drafts of TDMs using uploaded guidance notes, new legislation, and other TDMs for reference.
RevAssist has delivered significant results. The query tool handles approximately 2 500 queries per week from the roughly 300 officials working across the Legislative and Operational Divisions, with an accuracy rate of 97%. It has enhanced the capability of Revenue’s teams by broadening their technical knowledge, and given officials quicker access to information, which in turn boosts customer service response times.
Japan – Centre-based system for internal administrative tasks
The NTA Japan has launched an initiative to concentrate its internal operations in specialised operation centres, to streamline operations and increase accuracy. It aims to complete this process for all 524 of its tax offices by 2026, and examples of the internal operations involved include entry processing for tax returns and refund processing.
With the extra time and resource for new tasks secured through the streamlining of operational and administrative tasks, the NTA intends to improve and enhance external operations. This includes taxpayer services, tax examinations and effective use of data.
Singapore – Leveraging geospatial technology for enhanced property valuation
IRAS has leveraged GIS technology to advance property valuation processes. In collaboration with the Singapore Land Authority’s Centre for Geospatial & Geomatics, three collaboration projects were completed:
A workflow was developed to overlay rental data and accessibility indices on building shape files, generating relevant comparables based on property and locational attributes.
An evaluation was conducted on how accessibility factors, such as driving time to transport hubs and expressways, influenced industrial property rental rates.
A tool was created to identify comparables using proximity buffers, along with property criteria filters. It also integrated map-pinning functionality and street view imagery for a comprehensive assessment.
Collectively, these efforts harnessed data-driven spatial analysis and visualisation of property geospatial attributes, incorporating network analysis where relevant, to select comparables faster and further enhance valuation robustness. IRAS enjoys efficiency gains from workflow automation of data processing and overlaying relevant information for assessment.
In tandem, IRAS is strengthening internal GIS capabilities to ensure its workforce is future-ready. IRAS is growing a community of GIS enthusiasts that drive use-case development and foster continuous innovation in property tax administration. By combining structured training with hands-on projects supported by experienced GIS users, IRAS is accelerating skills development and enhancing the application of geospatial technology in its work.
Switzerland – Digitalisation and automation of the anticipatory Tax and Stamp Duty processes
Switzerland’s Federal Tax Administration has digitalised its administrative processes for its Anticipatory Tax (withholding tax) and Stamp Duty. The digitalisation of administrative processes requires more than just the conversion of existing paper forms into digital formats. Rather, it involves a comprehensive digital transformation in which new technological possibilities are exploited, processes are optimised, and automation potential is fully utilised.
The focus of this process has been on the digitalisation and automation of mass forms – i.e. forms that are received by the tax administration in large quantities – in order to achieve efficiency gains. The automation of large volumes of incoming forms has freed up resources to focus more on the examination of special cases.
Another key element has been the optimisation of the filing process. Previously, several forms had to be filled out manually and submitted in the correct order, whereas now, an integrated service has been created which enables a single submission. In addition, a comprehensive monitoring system has been implemented that ensures seamless tracking of all business transactions – from submission to deposit, or payment – and facilitates the support process.
In the area of technical support, the system provides a direct view of the ‘taxpayer perspective’, and on the other hand, a ticketing tool was implemented to automate the processing of requests as far as possible. These measures have resulted in a sustainable increase in efficiency, which relieves the burden on both the administration and taxpayers.
Thailand – Using a large language model to classify businesses
As part of efforts to modernise its tax administration, Thailand has introduced a pilot project to employ an LLM for the classification of businesses. The transition from a cumbersome manual to a digital solution aims to improve the accuracy and efficiency of industry classification. This initiative replaces traditional methods with an intuitive system that significantly reduces potential errors, thereby enhancing the precision of industry-level risk analyses and revenue forecasts.
The adoption of LLM technology enables better identification of emerging growth industries, allowing tax administrations to effectively prioritise resources. By automating the classification process, the Thai tax administration anticipates not only reducing administrative burdens, but also achieving more dynamic and responsive tax governance.
The pilot demonstrates the potential of LLMs to transform traditional administrative tasks into agile and adaptable data-driven operations. This strategic move towards digital transformation prepares the administration to anticipate future challenges, ensuring robust industry oversight and better-informed decision-making processes in the tax administration.
w
HMRC’s Microfiche and Microfilm Digitisation team were set the task of digitising 1.5 million records and 2 billion microfiche/microfilm images, dating back to the 1960s. By doing this, HMRC significantly accelerated its capability to retrieve data to support customer information requests while ensuring compliance with data protection legislation. The majority (70%) of these requests were to provide employment evidence supporting claims for industrial injuries. Other requests were to establish the right to reside/work in the UK, pension calculations and for a range of fraud investigations.
HMRC achieved this working with three separate suppliers under one multi-functional team. Due to their age, the original format of these records was deteriorating and the technology needed to read them was obsolete and in constant need of repair. The project team worked at pace through commercial negotiations to secure a supplier to outsource this work, enabling the records to be digitalised.
One billion images were successfully digitised in just six months and by January 2024, 1.7 billion unique National Insurance contribution records dating from 1961 to 1996 had been digitally archived. This work reduced HMRC’s processing times from 239 days to just 43 days and has enabled thousands of vulnerable customers to access their records quickly and efficiently.
Sources: Finland (2025), Ireland (2025), Japan (2025), Singapore (2025), Switzerland (2025), Thailand (2025) and the United Kingdom (2025).
Comparing the averages for both salary and ICT expenditures for the 10-year period between 2014 and 2023, Table 9.2. shows small increases of 1-2 percentage points. (It should be noted that the table only takes into account information from jurisdictions for which data was available for both years 2014 and 2023, which explains the differences in 2023 averages mentioned in the previous paragraphs.)
Note: The table is based on the data from jurisdictions that were able to provide the information for the years 2014 and 2023. The number of jurisdictions for which data was available is shown in parentheses.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.6 Resource ratios: Cost, http://isoradata.org (accessed on 1 October 2025), and Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Table A.25, https://doi.org/10.1787/tax_admin-2017-en.
Cost of collection
It has become a fairly common practice for tax administrations to compute and publish (for example, in their annual reports) a “cost of collection” ratio as a surrogate measure of their efficiency / effectiveness. The ratio is computed by comparing the annual expenditure of a tax administration, with the net revenue collected over the course of a fiscal year. Given the many similarities in the taxes administered by tax administrations, there has been a natural tendency by observers to make comparisons of “cost of collection” ratios across jurisdictions. Such comparisons have to be treated with a high degree of caution, for reasons explained in Box 9.2.
In practice there are a number of factors that may influence the cost/revenue relationship, but which have nothing to do with relative efficiency or effectiveness. Examples of such factors and variables include macroeconomic changes as well as differences in revenue types administered. These factors are further elaborated in Box 9.2.
Despite those factors, the “cost of collection” ratio is included in this report and the ISORA database for two reasons:
1. The “cost of collection” ratio is useful for administrations to track as a domestic measure as it allows them to see the trend over time of their work to collect revenue and, as pointed out in Box 9.2., they may be able to account for the main factors that can influence the ratio; and
2. The inclusion of the “cost of collection” ratio and the accompanying comments set out in Box 9.2. can serve as a prominent reminder to stakeholders of the difficulties and challenges in using the easily calculated “cost of collection” ratio for international comparison.
Table 9.3. illustrates the change in the average cost of collection ratio between 2014 and 2023. For the 46 administrations covered by the table, the ratio steadily declined since 2014, except in 2020 which was most likely a result of declining revenue collections during the COVID-19 pandemic.
Note: The table shows the average cost of collection ratio for 46 jurisdictions that were able to provide the information for the years 2014 and 2018 to 2023. Data for India, Israel and Türkiye have been excluded, see notes in Table A.16.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.6 Resource ratios: Cost, http://isoradata.org (accessed on 1 October 2025), and OECD (2017), and Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Tables A.27 and A.49, https://doi.org/10.1787/tax_admin-2017-en.
Observed over time, a downward trend in the “cost of collection” ratio can appear to constitute evidence of a reduction in relative costs (i.e. improved efficiency) and/or improved tax compliance (i.e. improved effectiveness). However, experience has also shown that there are many factors that can influence the ratio which are not related to changes in a tax administrations’ efficiency and/or effectiveness and which render this statistic highly unreliable in the international context:
Changes in tax policy: Tax policy changes are an important factor in determining the cost/revenue relationship. In theory, a policy decision to increase the overall tax burden should, all other things being equal, improve the ratio by a corresponding amount, but this has nothing to do with improved operational efficiency or effectiveness.
Macroeconomic changes: Significant changes in rates of economic growth etc. or inflation over time are likely to impact on the overall revenue collected by the tax administration and the cost/revenue relationship.
Abnormal expenditure of the tax administration: From time to time, a tax administration may be required to undertake an abnormal level of investment (for example, the building of a new information technology infrastructure or the acquisition of more expensive new accommodation). Such investments are likely to increase overall operating costs over the medium term, and short of offsetting efficiencies which may take longer to realise, will impact on the cost/ revenue relationship.
Changes in the scope of revenues collected: From time to time, governments decide to shift responsibility for the collection of particular revenues from one agency to another which may impact the cost/revenue relationship.
From a fully domestic perspective, an administration may be able to account for those factors by making corresponding adjustments to its cost or collected revenue. This can make tracking the “cost of collection” ratio a helpful measure to see the trend over time of the administration’s work to collect revenue. If it were gathered by tax type, it may also help inform policy choices around how particular taxes may be administered and collected.
However, its usefulness with respect to international comparison is very limited. While administrations may be able to account for the above factors from a domestic perspective, it will be difficult to do this at an international level as such analysis would have to consider:
Differences in tax rates and structure: Rates of tax and the actual structure of taxes will all have a bearing on aggregate revenue and, to a lesser extent, cost considerations. For example, comparisons of the ratio involving high-tax jurisdictions and low-tax jurisdictions are hardly realistic given their widely varying tax burdens.
Differences in the range and nature of revenues administered: There are a number of differences that can arise here. In some jurisdictions, more than one major tax authority may operate at the national level, or taxes at the federal level may be predominantly of a direct tax nature, while indirect taxes may be administered largely by separate regional/state authorities. In other jurisdictions, one national authority will collect taxes for all levels of government, i.e. federal, regional and local governments. Similar issues arise in relation to the collection of social insurance contributions.
Differences in the range of functions undertaken: The range of functions undertaken by tax administrations can vary from jurisdiction to jurisdiction. For example, in some jurisdictions the tax administration is also responsible for carrying out activities not directly related to tax administration (for example, the administration of certain welfare benefits or national population registers), while in others some tax-related functions are not carried out by the tax administration (for example, the enforcement of debt collections). Further, differences in societal views may influence what an administration does, how it can operate and what services is has to offer. The latter may have a particularly significant impact on the cost/revenue relationship.
Finally, it should be pointed out that the “cost of collection” ratio ignores the revenue potential of a tax system, i.e. the difference between the amount of tax actually collected and the maximum potential revenue. This is particularly relevant in the context of international comparisons – administrations with similar cost/revenue ratios can be some distance apart in terms of their relative effectiveness.
In 2023, the administrations included in this report employed more than 1.7 million staff (see Table A.18) highlighting the importance of effective and efficient management of the workforce to tax administration. As a result of the size of the workforce, salary costs average more than 70% of operating expenditures, meaning any significant budget change invariably impacts staff numbers. The challenge is compounded for some administrations which, due to contract restrictions or government mandates, may find it difficult to strategically down-size their operations other than through the non-replacement of staff who leave of their own accord.
Investing in Human Resources (HR) has been a key focus of the strategic planning of the RFB. This priority is reflected in the HR coordination and operational plan, which monitors and evaluates the progress of projects and initiatives monthly. This plan features:
The implementation of a user-oriented training system;
The introduction of an impact assessment for educational initiatives;
The enhancement of management and compliance reports;
The execution of an internal competition for reallocating employees among RFB’s units, facilitated by the automation of the corresponding working process;
The automation of the employee selection process for Brazilian tax and customs attachés;
The coordination of the open public competition for new employees, along with the automation of the work process;
The automation of controls on HR management, the issuance of administrative acts, and employee data querying.
These initiatives highlight the RFB's commitment to improving its HR processes through innovation and automation.
Source: Brazil (2025).
Over the10-year period 2014 to 2023, around 60% of administrations reported decreasing numbers of full-time equivalents (FTEs). However, this down-sizing trend seems to have halted in a number of jurisdictions as 60% of administrations reported increasing staff numbers between 2022 and 2023. (See Table 9.4.)
Note: Calculations based on information from 49 jurisdictions that were able to provide the data for the years 2014, 2022 and 2023.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table A.18 Total FTEs of the tax administration and FTEs of the tax administration by function: Registration and services, http://isoradata.org (accessed on 1 October 2025), and OECD (2017), and Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Tables A.46 and A.47, https://doi.org/10.1787/tax_admin-2017-en.
With most jurisdictions covered by this publication having an increasing population and labour force (see Table E.1.) and the long-term trend being declining staff numbers, FTEs have to serve more people. As can be seen in Table 9.5., on average the population and labour force per FTE increased by around 15% between 2014 and 2023. Digital transformation, technological advancements and the new applications and tools that come with this will help tax administrations not only support taxpayers in better ways but also allow them to support their staff (see also Box 9.6. and the surrounding text).
Note: The table shows the average resource ratios for those jurisdictions that were able to provide the information for the years 2014 and 2018 to 2023. The number of jurisdictions for which data was available is shown in parentheses.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.5 Resource ratios: Full-time equivalent (FTE), http://isoradata.org (accessed on 1 October 2025), and OECD (2017), and Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Table A.21, https://doi.org/10.1787/tax_admin-2017-en.
Staff usage by function
Figure 9.2. provides average allocation of staff resources (expressed in FTEs) across seven functional groupings used to categorise tax administration operations. It should be noted that since the ISORA 2023 survey, the functional groupings were revised splitting the category “All other functions” into four: (i) Dispute management; (ii) HR management; (iii) ICT support; and (iv) All other functions. This allows obtaining a better picture of staff usage by tax administrations.
Note: Excluding administrations that were unable to provide the break-down for all functions.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Tables D.7 Staff allocation by location and function: Registration, services, processing, and audit and verification, D.8 Staff allocation by function: Debt collection, dispute management, and HR management, and D.9 Staff allocation by function: ICT support and all other functions, http://isoradata.org (accessed on 1 October 2025).
While the detailed data for each administration in Tables D.7. to D.9. shows a significant spread of values and a number of outliers for each function, on average the “Audit, investigation and other verification” function and the “Registration, services, returns and payment processing” function are equally resource intensive, each employing on average close to 30% of staff.
Note: The table shows the average staff usage by function for 37 jurisdictions that were able to provide the information for the years 2014 and 2023.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Tables D.7 Staff allocation by location and function: Registration, services, processing, and audit and verification, D.8 Staff allocation by function: Debt collection, dispute management, and HR management, and D.9 Staff allocation by function: ICT support and all other functions, http://isoradata.org (accessed on 1 October 2025), and OECD (2017), Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Table A.20., https://doi.org/10.1787/tax_admin-2017-en.
As can be seen in Table 9.6., the average staff usage by function has remained relatively stable over the period 2014 to 2023. There has only been a small shift from “Registration, services, returns and payment processing” and “Audit, investigation and other verification” to “Dispute management” and “Other functions”.
Staff metrics
The ISORA survey also gathers key data concerning the age profiles, length of service, gender distribution and educational qualifications of tax administration staff: see Tables D.10. to D.16. and A.22. to A.33. While interpreting this data it should be noted that combined tax and customs administrations were allowed to use their total workforce for answering the underlying survey questions as it may be difficult for them to separate the characteristics of the tax and customs workforce.
Age profiles
While there are significant variations between the age profiles of tax administration staff (see Tables D.12. and D.13.), it is interesting to see that there are also differences when viewed across different regional groupings. This may be the result of a complex mix of cultural, economic, and sociological factors (for example, economic maturity, recruitment, remuneration, and retirement policies).
Figure 9.3. illustrates that staff are generally younger in administrations in the regional groupings of “Middle East and Africa” and “Asia-Pacific” where, on average, 25% and 30% of staff are below 35 years of age, whereas in the “Americas” and “Europe” this percentage drops to below 20%. At the same time, administrations in the “Americas” and “Europe” have a large percentage of staff older than 54 years.
Percentage of staff by age bands for selected regional groupings
Note: The following administrations are included in the regional groupings: Americas (9) – Argentina, Brazil, Canada, Chile, Colombia, Costa Rica, Mexico, Peru and the United States; Asia-Pacific (11) – Australia, China (People’s Republic of), Hong Kong (China), India, Indonesia, Japan, Korea, Malaysia, New Zealand, Singapore and Thailand; Europe (32) – Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Georgia, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Norway, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland and the United Kingdom; Middle East and Africa (5): Israel, Morocco, Saudi Arabia, South Africa and Türkiye.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Tables D.12 Staff age distribution: Staff below 45 years and D.13 Staff age distribution: Staff 45 years and above, http://isoradata.org (accessed on 1 October 2025).
Looking at the jurisdiction specific data, the percentage of staff older than 54 years grew in close to two-thirds of administrations over the period from 2018 to 2023 (see Figure 9.4).
Note: The table shows the movement of the percentage of staff older than 54 years for those jurisdictions that were able to provide the information for the years 2018 and 2023. Data for Iceland, Malta and Saudi Arabia has been excluded due to the mergers of the tax administration with the customs administration.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.13 Staff age distribution: Staff 45 years and above, http://isoradata.org (accessed on 1 October 2025).
Comparing the age profiles over the 10-year period between 2014 and 2023, Table 9.7. also confirms that tax administration staff is getting older. The percentage of staff older than 54 years increased by 4.1 percentage points on average.
Percentage of staff by age bands
Note: The table shows the average staff age profiles for 47 jurisdictions that were able to provide the information for the years 2014 and 2023. Data for Iceland and Malta has been excluded due to the mergers of the tax administration with the customs administration.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Tables D.12 Staff age distribution: Staff below 45 years and D.13 Staff age distribution: Staff 45 years and above, http://isoradata.org (accessed on 1 October 2025), and OECD (2017), Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Table A.22., https://doi.org/10.1787/tax_admin-2017-en.
Length of service
Figure 9.5. indicates that a significant number of administrations will not only face a large number of staff retiring over the next years, but that many of these staff will be very experienced, thus raising issues about retention of key knowledge and experience. The trend of experienced staff leaving administrations is already visible when looking at Table 9.8., which shows that between 2014 and 2023 the percentage of staff with less than 5 years of service has increased by 7.4 percentage points on average.
Source: OECD Secretariat calculations based on ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Tables D.12 Staff age distribution: Staff below 45 years, D.13 Staff age distribution: Staff 45 years and above, D.14 Length of service: Less than 10 years, and D.15 Length of service: 10 years or more, http://isoradata.org (accessed on 1 October 2025).
Percentage of staff by length of service bands
Note: The table shows the average length of staff service for 43 jurisdictions that were able to provide the information for the years 2014 and 2023. Data for Iceland and Malta has been excluded due to the mergers of the tax administration with the customs administration.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Tables D.14 Length of service: Less than 10 years, and D.15 Length of service: 10 years or more, http://isoradata.org (accessed on 1 October 2025), and OECD (2017), Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Table A.23., https://doi.org/10.1787/tax_admin-2017-en.
Gender distribution
Administrations were invited to report total staff and executive staff respectively by gender. As can be seen in Figure 9.6., while many administrations are close to the proportional line, typically female staff remains proportionally underrepresented in executive positions and significantly underrepresented in a number of administrations, something that has remained unchanged since the 2017 edition of this report (OECD, 2017[1]).
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.16 Gender distribution, http://isoradata.org (accessed on 1 October 2025).
Looking at the change over the 10-year period from 2014 to 2023, it can be noted that on average the share of female employees of total staff has increased by 2.7 percentage points, almost double the increase of the share of female executives (see Table 9.9.). The jurisdiction-level data shows that in about 60% of administrations the percentage of female executives has increased since 2014.
Note: The table shows the share of female employees of total staff and executive staff for 42 jurisdictions that were able to provide the information for the years 2014 and 2023.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.16 Gender distribution, http://isoradata.org (accessed on 1 October 2025), and OECD (2017), Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, Table A.24., https://doi.org/10.1787/tax_admin-2017-en.
To understand how the gender distribution might develop in the future, the ISORA survey asks tax administrations to report the percentage of recruits who are female. As can be seen in Figure 9.7., in three-quarters of administrations covered in this publication, females make up more than half of all staff that joined the administration during the fiscal year 2023.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.16 Gender distribution, http://isoradata.org (accessed on 1 October 2025).
Staff attrition
Staff attrition, also called staff turnover, refers to the rate at which employees leave an organisation during a defined period (normally a year). High attrition rates may result from a variety of factors, such as downsizing policies, demographics or changing staff preferences. The attrition rate should be considered together with other measures, such as the hire rate, which looks at the number of staff recruited during a defined period, when evaluating the human resource trends of an administration.
While a high attrition rate combined with a low hire rate is usually associated with a general downsizing policy, administrations should perhaps be concerned where both rates are high. Recruitment is costly, not only the recruitment process itself but also the cost and time for training and supporting new staff members.
Having attrition rates that are too low may also not be ideal. While an organisation is growing, a low attrition rate may be desirable. However, in situations where both the attrition rate and the hire rate are low, an organisation may not have the ability to recruit new skills as all positions are filled. This could be an issue particularly for administrations that are undergoing transformation and are therefore in need of staff with skills that are different from what is currently available within the administration.
While what is considered a “healthy” attrition rate differs between industry sectors or jurisdictions, the average attrition and hire rates for administrations participating in this publication of around 7-8% in 2023 would seem to point at a reasonable range for tax administrations of between 5% and 10%. This is also in-line with the average attrition and hire rates over the 10-year period from 2014 to 2023 as can be seen in Table 9.10.
Note: The table shows the average attrition and hire rates for 48 jurisdictions that were able to provide the information for the years 2014 and 2023.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.10 Staff dynamics, http://isoradata.org (accessed on 1 October 2025).
However, when looking at specific administration data, it becomes apparent that “attrition and hire” rates cover a very broad range. Figure 9.8. shows the relationship between tax administration attrition and hire rates. It illustrates that there are a number of administrations with attrition and hire rates well above 10% (upper-right box), while others show very low attrition and hire rates (lower-left box).
Note: Attrition rate = number of staff departures/average staffing level. Hire rate = number of staff recruitments/ average staffing level. The average staffing level equals opening staff numbers + end-of-year staff numbers/2.
Source: ADB, CIAT, IOTA, IMF, OECD, International Survey on Revenue Administration, Table D.10 Staff dynamics, http://isoradata.org (accessed on 1 October 2025).
While recruitment rates may vary year on year, the challenge of training and knowledge transfer are consistent. Tax administrations are increasingly facing challenges to both recruit staff with the right skills and retain existing staff. Box 9.4. outlines innovative approaches being taken by France and Hungary to attract new candidates and motivate existing staff.
France – Career development
DGFiP launched a significant project with the objective of constructing a more supportive environment for helping officials to develop their careers.
The project was carried out with the cooperation of tax administration managers, either through a survey (conducted among 30 000 officials) or through a large number of workshops (over 130) to identify areas for improvement.
From the responses, the following was implemented:
A web application that asks users about their career paths and aspirations, and in response offers illustrations of possible career paths that match their profiles. It also provides access to a comprehensive library of 60 inspiring career paths within the administration, and offers a quiz to test knowledge of the HR rules to career pathways.
An upgrade of the intranet dedicated to HR with access to career routes/mapping for officials.
A newsletter sent to all managers, gathering job offers published every week and corresponding to their positions.
Enhanced support to prepare officials for the recruitment process when applying for roles, including the appointment of mentors, access to specific intranet spaces with all the necessary documentation, and the recording of video conferences.
Days dedicated to immersion in different areas of work to help discover other occupations.
Hungary – Internship programme for data scientists
The NTCA has launched a new internship programme for data scientists completing their studies, aimed at recruiting young talent for data-intensive roles. The programme also served as a pilot project to explore innovative recruitment approaches.
In addition to social media and short video messages, the NTCA disseminated information about the programme to targeted candidate groups through a network of university contacts, specifically tutors in relevant fields.
As past efforts indicated that prolonged recruitment processes significantly increase the likelihood of candidates opting for other opportunities, either midway or even after a successful application, the application process was simplified and sped up.
Undergraduate students are eligible to apply from their third semester onwards, regardless of their field of study. Candidate selection was based on an analytical exercise that tested practical skills. While offering positions across numerous NTCA field offices, the NTCA centralised candidate communication and streamlined evaluations. Consequently, applicants received feedback on their submissions within just a few days.
Initial findings suggest that “direct marketing” methods are effective in reaching the target group. Coupled with mentoring, interns can produce value very quickly. Although they start on temporary contracts, there is the possibility to apply for a permanent position. The advantage is that by this time they will have good understanding of organisational procedures, datasets, and have built strong working relationships with colleagues.
For further information, please see here: https://www.youtube.com/watch?v=KZZNvfmWE8k (accessed on 1 October 2025).
Sources: France (2025) and Hungary (2025).
Virtual training
During the COVID‑19 pandemic, many tax administrations reported transitioning their training programmes from face-to-face contact to a virtual environment. Tax administrations report that these practices brought significant benefits to the administration and participants, and as a result they are becoming permanent features of training programmes. Previous editions of this report showcased examples of this, such as the use of live online training sessions or pre-recorded videos/webinars (OECD, 2021[2]) or interactive applications that can also be used as part of recruitment efforts (OECD, 2023[3]).
While moving to a virtual training environment may have some up-front costs, it may save costs in the longer term as once produced, pre-recorded training material can be viewed at any time, from anywhere. Remote training can reduce travel expenses and can allow staff to learn at their own pace and convenience as well as increasing the number of staff members that can follow a course. New technologies are also helping facilitate the collaborative learning aspects, increasing the quality of the training experience.
In addition to the virtual training programmes produced by tax administrations themselves, international organisations are also producing e-learning courses specifically designed for tax administrations. One of these programmes is the Virtual Training to Advance Revenue Administration (VITARA) initiative, a joint project of four organisations, which is described in more detail in Box 9.5.
VITARA is an online course specifically designed for tax administrations. The course consists of several short, structured online modules. It is a joint project of four organisations: CIAT, the IMF, IOTA and the OECD.
The VITARA course content is tailored to senior managers and executives of tax administrations of developing countries. Yet other tax officials, including from advanced economies, could also benefit from following the different course modules. All modules are free of charge and there are no prerequisites or qualification requirements.
The VITARA curriculum represents a comprehensive training package for tax administration management. It consists of two parts and includes the following topics:
Part A. Institutional governance, management, and support
Institutional governance
Compliance risk management
Organization
Strategic management
Information technology and data management
Reform management: (i) Fundamentals and (ii) Specific Topics
Human resource management
Performance management
Enterprise risk management
Part B. Core functions of tax administration
Introduction to tax administration
Taxpayer registration
Taxpayer services
Filing of declarations
Payment and debt collection
Audit program
Dispute resolution
Revenue management
By October 2025, twelve VITARA modules have been developed in English, and some are also available in Spanish, French and Arabic. New English modules as well as further translations will follow. The available modules (underlined above) can be accessed here: https://www.imf.org/en/Capacity-Development/Training/ICDTC/Search?sortby=Relevancy&sortdir=Descending&keywords=VITARA (accessed on 1 October 2025).
To support learners, and to provide easy and free access to the content of the VITARA online modules (particularly to those that have no stable internet access), the VITARA partners have started reproducing the module content in so called “Reference Guides”. By October 2025, seven Reference Guides have been published:
Human Resource Management (CIAT, IMF, IOTA, OECD, 2024[4])
Institutional Governance (CIAT, IMF, IOTA, OECD, 2024[5])
Organisation (CIAT, IMF, IOTA, OECD, 2024[6])
Strategic management (CIAT, IMF, IOTA, OECD, 2023[7])
Reform Management Fundamentals: Setting Up a Reform Program (CIAT, IMF, IOTA, OECD, 2024[8])
Reform Management Specific Topics: Managing a Reform Program (CIAT, IMF, IOTA, OECD, 2024[9])
The Audit Program (CIAT, IMF, IOTA, OECD, 2025[10])
Supporting staff
The changes tax administrations are managing, whether technology, policy or budget driven, are constant. In addition, the wider digital transformation of the economy is changing the service expectations of taxpayers. To maintain staff motivation and performance, tax administrations are considering the best way to support staff through these changes, as well as ensuring they have the right tools for the tasks.
Technology is playing a key role in this. This includes the use of artificial intelligence, machine learning and robotic process automation (RPA) to automate some of the core tasks within a tax administration. Table 6.4. in Chapter 6 highlights the rapid growth in the use of such services with, for example, 60% of administrations reporting that they now using RPA, an increase of close to 40 percentage points since 2018. This is helping tax administrations respond to budgetary and workforce pressures as it is freeing up resource for staff to be focussed on more complex tasks.
Box 9.6. illustrates the different ways tax administrations are assisting their staff, from automation to boosting productivity and upskilling technology skills. Some administrations, such as France, have even taken a proactive approach to prevent unwelcome behaviours.
Austria – Voice bot to automate telephone inquiries
The Austrian Ministry of Finance has developed a voice bot to automate telephone inquiries regarding the use of ID Austria authentication on the FinanzOnline platform. The high volume of repetitive inquiries placed a significant burden on the customer service team. The voice bot helps relieve staff by independently handling simple and standardized requests. Using modern AI technologies and speech processing, the bot provides precise answers and efficiently supports typical processes.
In addition to reducing wait times, the voice bot improves service quality and offers 24/7 support. Continuous learning processes allow the system to adapt to new requirements and remain up to date through regular updates. Experience from other organizations shows that such automation solutions can significantly reduce call volumes and shorten processing times.
A key advantage of the project is its scalability. The voice bot can easily be extended to other administrative areas, such as tax-related inquiries or document applications. Automating such processes contributes to a more efficient and user-friendly public administration. With its learning architecture and continuous optimization, the voice bot provides a sustainable solution that supports the long-term digitalisation and modernisation of the public sector.
Canada – Robotic process automation
To streamline repetitive and manual tasks, the CRA has implemented RPA. RPA is a software solution designed to automate routine business processes by mimicking employee actions, without utilising learning, judgment, or assumptions.
RPA is capable of handling complex tasks across multiple systems and applications, with great speed and precision. It can perform functions such as data entry and calculations, collecting, organising, and validating information, and documenting and filing results. Benefits of RPA include:
Improved employee experience and satisfaction
Reduced carbon footprint
Optimised workloads
Refocused employee efforts on higher-value work
Since its adoption, the CRA has automated over forty processes, saving the equivalent of more than 187 000 hours of labour and millions of dollars. One significant application of RPA is the automation of the triage process for one-way electronic Requirements for Information (eRFI) responses from financial institutions. This automation has reduced the manual effort required to handle the large volume of eRFI responses, enabling collectors to receive information faster, which helps accelerate payments and close accounts.
This RPA is expected to handle over 30 000 transactions, significantly accelerating file triage for collectors. The CRA continues to explore additional areas where RPA can be applied to further enhance programmes and deliver faster, more efficient services to its taxpayers.
Finland – Productivity management
The Finnish Tax Administration has taken steps to improve its productivity management through using operational management systems, which is related to how the administration ensures it meets its objectives. Several systems (both externally acquired and internally developed) are being used to facilitate planning, the scheduling of shift work, and monitoring. The goal is to automate as much as possible, and centralise control/ management. The main focus areas are:
Resource planning: This includes allocating resources against goals and skills based on given priorities, managing employees' time, and calculating how much resource is needed for each task to allocate it efficiently.
Task management: Using operational systems to prioritise tasks.
Operational personnel planning: Ensuring staff are assigned to the correct task and staff skills are developed to meet the requirements of their role.
This brings numerous benefits, to both employees and managers. Employees can see their tasks on a weekly basis, which improves the accuracy of the work plan and increases work predictability. It also enables staff to see that they are doing tasks which suit their skillset.
For managers, it reduces the time they spend on planning and manually updates, freeing up time for other work such as coaching employees. It also increases the transparency of the group's work and makes it easier to agree on the division of tasks.
France – Addressing inappropriate or aggressive behaviours
DGFiP has adopted a policy of zero tolerance towards any form of inappropriate behaviour or aggression.
To achieve these objectives, a national action plan has been adopted, which includes measures to refresh and modernise the system for reporting, monitoring and dealing with incidents. This has led to the development of the Sign@lFiP application.
Launched in 2024, Sign@lFiP makes it easier to detect, report and follow up on any incident suffered by an official. The tool was developed based on feedback from a random sample of officials. The Sign@lFiP application is accessible to all staff via the staff portal. It enables them to report either an ‘external’ incident, i.e. one that has occurred in an interaction with an external user/person (for example, verbal altercations, insults, etc.), or an ‘internal’ incident, i.e. one that has occurred in their relationship with another staff member. This function also makes it possible to report any behaviour of a sexist or sexual harassment nature, combatting any form of discrimination or harassment by supporting the reporter and punishing the perpetrator.
Lastly, the tool includes a module designed to keep all those involved in employee protection (such as the head of department, the person in charge of legal protection, the line management chain and senior management) informed on the volume of alerts and, with the official’s consent, also their individual content.
France – Generative artificial intelligence assistant for document searches
DGFiP has combined GenAI with powerful document search capabilities in its GenAI assistant called Caradoc, which enables officials to access critical information faster and focus on higher-value tasks, reducing administrative burdens.
Caradoc features a robust GenAI-supported document search engine that operates in two distinct modes:
Single file upload: Officials can upload a document and submit specific questions or queries directly related to the document's content. Caradoc not only provides a response, but also highlights the exact excerpts from the document used to generate the answer.
Collection mode: The collection mode expands Caradoc’s capabilities by allowing users to create a personalised collection of documents (for example, policy guidelines, legal texts, internal procedures) and perform comprehensive, cross-document searches. Users can pose global or complex questions, which Caradoc answers by simultaneously extracting relevant information from multiple documents to formulate a comprehensive answer. The excerpts used to formulate the response are also highlighted, ensuring transparency and a deeper understanding of the results.
In both modes, the assistant retrieves the most relevant information while highlighting the specific documents excerpts used to generate responses, ensuring transparency and traceability.
This solution clearly underscores DGFIP’s commitment to leveraging AI-driven innovation to modernise the tax administration and support its workforce.
Singapore – Technology-essentials training roadmap
IRAS has introduced the Technology-Essentials (Tech-E) training roadmap to advance the technology skills of officers who do not work in the information and communications technology or data areas. The aim is to improve technology proficiency and accelerate officers’ data and digital skills, allowing them to effectively operate in the vast digital ecosystem to maintain long-term competitiveness and agility.
In developing the Tech-E training roadmap, IRAS identified the essential data and digital tools (tech tools) that officers need to perform their tasks effectively. These tech tools cover automation, analytics, design and user experience, and generative AI. The design of the roadmap also recognises that mature or mid-career officers may require more support in meeting proficiency levels in data and digital skills.
The tech tools are categorised into different levels:
Level 1: Officers become proficient in technology relevant to their work, enabling them to perform functional tasks independently.
Level 2: Officers identify value-adding opportunities using technology and facilitate innovation.
Level 3: Officers use technology to solve complex problems and devise innovative solutions.
The training roadmap offers a blend of e-learning and classroom training, providing officers the flexibility to choose courses that best suit their individual needs and learning preferences. IRAS adopts a phased approach in implementing the training roadmap, which enables officers to progressively acquire relevant technological skills. This strategy allows officers to advance from baseline proficiency to in-depth technological skills over about four years, while balancing their day-to-day work responsibilities.
Sources: Austria (2025), Canada (2025), Finland (2025), France (2025) and Singapore (2025).
References
[10] CIAT, IMF, IOTA, OECD (2025), VITARA Reference Guide: The Audit Program, OECD Publishing, Paris, https://doi.org/10.1787/3c674401-en.
[8] CIAT, IMF, IOTA, OECD (2024), Reform Management Fundamentals: Setting Up a Reform Program, OECD Publishing, Paris, https://doi.org/10.1787/4d33b619-en.
[9] CIAT, IMF, IOTA, OECD (2024), Reform Management Specific Topics: Managing a Reform Program, OECD Publishing, Paris, https://doi.org/10.1787/1fc51bcf-en.
[4] CIAT, IMF, IOTA, OECD (2024), VITARA Reference Guide: Human Resource Management, OECD Publishing, Paris, https://doi.org/10.1787/6ef70142-en.
[5] CIAT, IMF, IOTA, OECD (2024), VITARA Reference Guide: Institutional Governance, OECD Publishing, Paris, https://doi.org/10.1787/1ade5674-en.
[6] CIAT, IMF, IOTA, OECD (2024), VITARA Reference Guide: Organization, OECD Publishing, Paris, https://doi.org/10.1787/ab075e83-en.
[7] CIAT, IMF, IOTA, OECD (2023), VITARA Reference Guide: Strategic Management, International Monetary Fund, Washington, DC, https://doi.org/10.5089/9798400223488.069.
[3] OECD (2023), Tax Administration 2023: Comparative Information on OECD and other Advanced and Emerging Economies, OECD Publishing, Paris, https://doi.org/10.1787/900b6382-en.
[2] OECD (2021), Tax Administration 2021: Comparative Information on OECD and other Advanced and Emerging Economies, OECD Publishing, Paris, https://doi.org/10.1787/cef472b9-en.
[1] OECD (2017), Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, OECD Publishing, Paris, https://doi.org/10.1787/tax_admin-2017-en.