Assessing the accuracy and completeness of taxpayer reported information is one of the key functions of tax administrations and critical for supporting voluntary compliance. This chapter takes a closer look at tax administrations’ work in this area, including how they manage compliance.
Tax Administration 2022
Comparative Information on OECD and other Advanced and Emerging Economies
Tax Administration
6. Verification and compliance management
Abstract
Introduction
The audit, verification and investigation function assesses the accuracy and completeness of taxpayer reported information. This function employs on average thirty percent of tax administration staff and verifies that tax obligations have been met. While this often happens through conducting desk or field based “tax audits”, there is an increased use of automated electronic checks, validations and matching of taxpayer information. The undertaking and visibility of these and other compliance actions is critical in supporting voluntary compliance, including through their impacts on perceptions of fairness in the tax system, as well as creating a ‘deterrent effect’. This chapter therefore looks at:
How tax administrations manage compliance risks, including the use of large and integrated data sets;
The delivery of compliance actions undertaken by tax administrations including moving field audit work into a virtual environment; and
The work on tax and crime.
Compliance risk management
The OECD report The Changing Tax Compliance Environment and the Role of Audit (OECD, 2017[1]) looked at the range of incremental changes occurring across tax administrations which, taken together, were changing the nature of the tax compliance environment, allowing for more targeted and managed compliance.
A significant part of this change is driven by the increased availability of data. As digitalisation proceeds, even more tax related data from taxpayers and third parties is becoming available (for example, data from e‑invoicing, online cash registers and financial account information), which is contributing to a clearer understanding of tax gaps. Most tax administrations now apply data sciences techniques and use analytical tools as part of compliance processes (see Table 6.1), and this is explored in more detail later in this chapter. Box 6.1 also contains some examples of the range of data exploration techniques being used by tax administrations.
Another growing trend is the combination of analytics with behavioural analysis to build a more holistic understanding of compliance risks, behavioural patterns and appropriate compliance interventions. Figure 6.1 below, shows the percent of tax administrations who are using behavioural insights in their work.
Figure 6.1. Use of techniques and methodologies to improve compliance, 2020
Source: Table A.49 Techniques and methodologies to improve compliance.
Box 6.1. Examples - Data exploration
Australia - Financial data matching
The Australian Taxation Office (ATO) has developed a tool that can match up any financial transaction data, regardless of its source or form. It can match, for example:
Receipts and payments in one group of bank statements to corresponding receipts and payments in another group of bank statements;
Bank statement receipts/payments to their corresponding entries in accounting records;
Debit entries in a set of accounting records to their corresponding credit entries within the same accounting records;
Bank statements or accounting records to pay slips, invoices or inventory records.
This template can process up to 10 000 data items in a single tranche, and operates approximately 30x faster than a human analyst and has better than human accuracy on most metrics. The robust fuzzy matching engine is undeterred by abbreviations or typographical errors in the data, and automatically detects and reconciles compound payments (that is, a single entry in one data source that matches to a number of smaller entries in another data source). The template also outputs match confidence indicators, and produce exceptions report (e.g. unmatched items) automatically.
Investigators and auditors use this template to confirm the details of suspicious transactions by corroborating them within or across different data sets and identifying relationships that are not apparent to the human eye.
Austria - Digital innovations as regards electronic compliance checks and real-time audits
In 2020, the Austrian tax administration has deployed a real-time risk assessment for PIT declarations for employed individuals. Based on this scoring, cases are selected for desk audits. An evaluation of the year 2020 revealed a doubling in the hit-rate (cases identified as having a supplemental claim) while the number of cases selected decreased by more than 40%. Due to the success of the innovation in risk scoring, the Austrian tax administration launched a major project in 2021 to expand gradually this method of real-time risk assessment to all PIT declarations as well as to VAT declarations, CIT declarations and other taxes.
The main target is to select the best cases and find those with the most probable supplemental claim. The methodological approach incorporates various machine learning techniques such as tree models, regression models, balance score cards as well as ensemble techniques. In addition, also expert rules are developed, since not all risk areas provide enough historical data to train models.
The Austrian tax administration also uses supervised learning techniques for central case selection since 2016 for business, wage and customs audits. Evaluations of the case selections done with advanced analytics in comparison to manual case selections show a lift in the hit-rate by more than double, i.e. audits of cases selected by machine learning algorithms are twice as likely to result in an audit surplus of more than EUR 10 000.
Canada - Digital Forensic Investigation Tools
The Canada Revenue Agency (CRA) engages in complex investigations where large volumes of data are gathered pursuant to search warrants or other means. The CRA utilises various digital forensic tools that assist the user in narrowing down what may be relevant to an investigation thus saving substantial time and costs on an investigation, including the following:
Advanced tools that learn as the investigators are reviewing and identifying relevant data and evidence thereby allowing investigators to save time in finding documents relevant to an investigation.
Tools that enable users to view visual linkages, clusters and patterns which help investigators focus on key areas of potential interest.
Tools that identify and extract key patterns within evidence which may be relevant to the investigation such as telephone numbers, credit card numbers, social insurance numbers, etc.
Tools that perform searches of contents of data along with the metadata associated with the file, as well as “fuzzy” searches that enable a search of text that is close to what the investigator may be looking for.
Tools that create dictionaries from seized data that are used to form passwords for brute force attacks on encrypted or password protected documents.
For business intelligence analysts, the use of electronic discovery software provides structure to a high volume of unstructured data. It also focuses efforts on the most relevant information so criminal investigations can act swiftly. The software also permits collaboration from various workplaces. In the current environment, access from different worksites is essential.
Canada - Refining services through analysis
The CRA faces unique analytics problems, and has developed and empirically tested new methods to address these gaps. Scatterplot matrices are useful for exploring low-dimensional datasets (i.e., datasets with only a few columns). However, currently there are no ideal alternatives when dealing with high-dimensional data (i.e., datasets with many columns). The work in question presents a data exploration method that is easily applicable in high-dimensional contexts, and is designed to yield similar analytical information to scatterplot matrices. This is done by combining complimentary information from (a) an overall multivariate model and (b) a series of bivariate models.
As an example, this may be applied when mining call centre databases, wherein callers’ requests are logged and tracked as they are filled. Such a database may have thousands of variables to investigate. These models developed by the CRA, permit analysts to sift efficiently through these databases to find information that can refine business practices or inform decision-making. In order to ensure this approach provides accurate insights, we have tested it rigorously via a series of simulation studies. In these contexts, the Agency commonly uses these methods to look at which factors relate to the time it takes to complete external requests. This information is then used to spot bottlenecks, oddities, or inefficiencies.
Italy – Using data to estimate the VAT Gap
The Bottom-Up Tax Gap Estimation initiative is a new methodology for the estimation of the VAT gap based on the use of information from the tax assessment database. The estimates are obtained by combining traditional methods, modern machine learning techniques and ‘nearest neighbor’ procedures.
This “machine learning assisted” methodology produces a final outcome of a set of individual values that can be used to obtain estimates of the VAT gap at different levels of detail. This work will support tax administration by providing clear focus for compliance activity, and data for policy making. It also allows the gap to be examined by different taxpayer behaviours.
See Annex 6.A for supporting material.
Sources: Australia (2022), Austria (2022), Canada (2022) and Italy (2022).
Increasing availability of data
As more and more data is stored electronically, and the transfer, storage and integration of data has become easier through the application of new techniques and processes, there has been a huge increase in the amount of data available to tax administrations for compliance purposes. New data sources include:
Data from devices: Data can be collected from devices that register transactions such as online cash registers and trip computers for taxis and trucks, and also gate registrations from barriers and weigh bridges.
Data from banks, merchants or payment intermediaries and service providers: This allows direct verification of income or assets reported by the taxpayer. Some jurisdictions already receive transaction details or transaction totals for taxpayers on a regular basis.
Data from suppliers: Collecting data from suppliers, either directly or through the taxpayer, allows a more complete picture to be drawn about the activities and income of the taxpayer. This is seen in the increasing use of e-invoicing systems which, as noted in Chapter 4, allows some tax administrations to prefill tax returns.
Data from the customer: This is easiest in cases where the number of customers is limited and known, but increasingly mechanisms to leverage customers in compliance are being used, for example in the verification of cash receipts.
Unstructured data concerning the taxpayer: Increasingly electronic traces relevant to business activities and transactions can be found on the internet and in social media. Also the analysis of unstructured data can improve response times and accuracy as set out in the example from Australia in Box 6.1.
Data from other government agencies: Data held by other government agencies for example for licencing, regulatory or social security purposes can be relevant in verifying tax returns or in risk assessments. For example, the French tax administration uses information from local property maps, municipalities, the real estate registry and aerial photographs available on the web to identify real estate that, due to its use and value, is subject to special tax (see Box 6.2).
Data from international partners: New international exchanges of data commencing under the Common Reporting Standard and Country-by-Country Reporting is massively increasing the quantity of data available on international activity and providing useful information for audit and case selection processes and in some cases for prefilling of tax returns.
Box 6.2. Examples - Increased availability of data
Argentina - Simple and Pro-Forma VAT Tax Return
A Digital VAT Ledger is compulsory for all taxpayers liable for VAT in Argentina (around a million taxpayers), and for a defined group of VAT exempt taxpayers (even though they are not required to file VAT tax returns). During a first phase of implementation, around 20 000 taxpayers - who issued a small amount of electronic invoices and had a single activity of professional services - were selected for a new simplified procedure which comprised two parts:
A Simplified Digital VAT Ledger: This shows the taxpayers all the electronic invoices they issued and received, and taxpayers must confirm or correct the record for that particular period. When invoices are missing, they can be added to the app manually and invoices not related to the commercial activity must be deleted (e.g., personal expenses).
A Simplified VAT tax return: Based on the data of the Digital VAT Ledger, a Value Added Tax return is automatically generated and the amount due is assessed. The amounts cannot be edited. If necessary, the Digital VAT Ledger has to be modified.
As a next stage, a new digital VAT return will be gradually implemented. This return will be prefilled with all the amounts entered in the Digital VAT Ledger (sales, purchases, exports, tax debits and credits, etc.). In the future, this pro-forma return will be compulsory for all VAT liable taxpayers. Through this work there should be increased compliance, more reliable data, reduced business burdens and reduced fraud opportunities.
See Annex 6.A for supporting material.
Canada - Follow the Dollar Research Initiative
The CRA has developed its ability to integrate data from various assessment and compliance activities related to PIT, CIT, and VAT (GST/HST) tax programmes, creating a holistic understanding of effectiveness of different interventions and enabling outcomes to be quantified.
The Follow the Dollar project provides a view of the life cycle of tax returns, from initial assessment to resolution, with associated analysis. As a result of the initiative, CRA can identify trends, track the impact of interventions, policy changes, and more. This initiative allows CRA to follow the dollar in our processing from end-to-end. Beginning with the initial filing of a tax return by the taxpayer/registrant or entity, we are able to monitor the dollar through activities performed to enhance compliance, then follow it through appeals (if warranted) and finally through to payment, refund, collection or write-off, measuring outcomes along the way. In so doing, CRA is able to calculate certain metrics, such as audit yield. Additionally, this initiative is enabling further research such as exploring the write-off predictability rate for the Agency, as well as identifying clusters, trends and compliance gaps.
France - Innovative Land Tenure
To improve the process of detecting undeclared constructions or developments, the French tax administration (DGFiP) uses artificial intelligence and data enhancement based on aerial photographs taken by the Institut national de l'information géographique et forestière (IGN) as part of the "Foncier innovant" project.
The algorithms will make it possible to extract the outlines of buildings and swimming pools from the public aerial images of the IGN, which can be consulted by everyone on the website www.geoportail.gouv.fr. A computer process then verifies, from the declarations made by the owners to the tax authorities, whether the elements thus detected on the images are correctly taxed for direct local taxes (property tax in particular). An agent of the tax authorities then systematically verifies each anomaly detected before any operation of reminding the owner of the property and ultimately of taxation.
See Annex 6.A for supporting material.
Lithuania - Increasing transparency of 2nd hand car transactions
According to the data of the Lithuanian State Tax Inspectorate (STI), used car trade, repair and car part trade sectors are classified as one of the highest risk sectors for tax evasion. Based on the results of the study concluded in 2018, the VAT gap within the sector for trade of used care was more than EUR 38 million. Therefore, STI is especially keen on performing monitoring and control activities within this sector.
In order to resolve the problem of unscrupulous salespersons operating within the used car trade sector, a new accounting framework for vehicle owners has been developed with changes in the legislation requiring all cars present within the country to bear a unique vehicle owner declaration code (SDK). The code allows a specific car to be traced back to its owner within Lithuania.
Prior to the entry of the vehicle into Lithuania, the owner of vehicle is required to have a valid SDK. The registration and sale of vehicle is not allowed without SDK and it is required to be published in all vehicle advertisements. By checking more than 6 000 vehicles on roads and parking lots and more than 22 000 vehicle advertisements on the internet, almost 21 000 total cases of missing SDK and/or failure to publicly disclose the SDK were uncovered. The offenders received warnings and their information has been logged for repeat investigations following the end of the grace period.
The newly implemented system has resulted in a positive impact being realised within a couple months after launch. It has highlighted 22 persons possibly engaged in the operation of unregistered car trade, EUR 109 000 of undeclared income have been disclosed, and six persons have registered their car trading operations. This is helping create a fair market while car salespersons causing the highest risk are being identified and purchasers can check the sale is trustworthy.
See Annex 6.A for supporting material.
Spain – Analysing feedback from verification actions
The Spanish Tax Agency’s Strategic Plan 2020-2023 includes plans for prevention and control of tax and customs fraud, and includes a complete analysis of the results of compliance actions, in order to ensure that tax regularizations become an additional tool for obtaining improvements in voluntary compliance by taxpayers. In this regard, it is important to highlight the importance of the feedback from the verification actions, consisting of gathering information on the reasons that lead to the regularisation of the taxpayers included in the control plans.
This information is incorporated into the Tax Agency's databases and allows for the qualitative evaluation of the verification actions. This helps in the planning of future actions, which will be oriented towards the programmes, profiles or risks that are of most help in reducing the most significant non-compliance. The collection of this information is obtained through predefined forms, in which the officials select the main reason or reasons for the tax regularisation and quantify it. This is done within a standardized table of reasons for regularisation, weighting the reason for regularisation against the amount of the tax assessment.
The analysis of the reasons for regularisation is carried out through computer tools that can support massive data processing. This allows structured information to be obtained on the specific risk areas of the main taxes of the Spanish tax system as well as knowledge of the reasons that lead to the regularisation.
See Annex 6.A for supporting material.
Sources: Argentina (2022), Canada (2022), France (2022), Lithuania (2022) and Spain (2022).
There are, though, some emerging risks to the availability of large data sets. In particular, it is increasingly possible for data relevant to the tax administration in one jurisdiction to be held within the territory of another jurisdiction. In these circumstances, it can be difficult to obtain the data on an automatic basis from the data holder located in another jurisdiction. This could make it more difficult to risk assess in some circumstances, as well as making it more difficult to prefill tax returns and to further develop compliance-by-design processes.
An example of this comes from the growth of the sharing and gig economy facilitated through online platforms which can operate across border. This may become an increasing risk as the online economy grows, particularly if it is accompanied by a shift from salaried employment (and the reporting of incomes by employers) to self-employment. This issue was considered in the OECD report The Sharing and Gig Economy: Effective Taxation of Platform Sellers (OECD, 2019[2]). That report looked at a number of strategies currently being adopted by tax administrations as well as their limitations and recommended the development of standardised reporting requirements to facilitate possible future automatic exchange of information between tax administrations. It also led to the development of:
A set of Model Rules that when used in legislation require digital platforms to collect information on the income realised by those offering accommodation, transport and personal services through platforms and to report the information to tax authorities (OECD, 2020[3]).
A Code of Conduct to facilitate a possible standard approach to co-operation between administrations and platforms on providing information and support to platform sellers on their tax obligations while minimising compliance burdens (OECD, 2020[4]).
Another risk that has been identified is that posed by digital financial assets (DFAs), such as cryptocurrencies. The owners of DFAs can be very difficult to trace even though they may be linked to the creation of a specific digital wallet (which is somewhat similar to a bank account). Tracking down the individuals or entities behind particular wallet addresses is at best very difficult and resource intensive.
While not a risk as such, it should also be noted that data protection requirements could limit the circumstances in which data can be kept, processed or shared. This is a key consideration for administrations in designing systems which rely on large data sets and the retention of data.
Sharpened targeting of risks
Data science
Over recent years, the application of advanced analytics to risk management and risk targeting is becoming increasingly common;
Figure 6.2 shows 80% of tax administrations reporting using big data in their work, and of those that use big data nearly all are using it to improve their compliance work.
Of the 58 tax administrations covered by this report, 52 report using data science / analytical tools with the remaining administrations in the process of preparing the use of such tools going forward (see Table A.51).
Similarly, the use of artificial intelligence, including machine learning, for risk assessments and detecting fraud is already undertaken or in the process of being implemented by the majority of administrations covered in this publication (see Table 6.1 and Figure 6.3).
This increasingly sophisticated use of analytics on expanded data sets is leading to a sharpening of risk management and the development of a range of intervention actions, including through automated processes. A selection of examples is included in Box 6.3. Additionally, the OECD report Advanced Analytics for Tax Administration: Putting data to work (OECD, 2016[5]) provides practical guidance on how tax administrations can use analytics to support compliance and service delivery.
Table 6.1. Application of data science, 2020
Percent of administrations
Source: Tables A.51 Innovative technologies: Implementation and usage (Part 1) and A.52 Innovative technologies: Implementation and usage (Part 2).
Figure 6.2. Use of big data for analytical purposes, 2022
Note: The figure is based on ITTI data from 52 jurisdictions that are covered in this report and that have completed the global survey on digitalisation.
Source: OECD et al. (2022), Inventory of Tax Technology Initiatives, https://www.oecd.org/tax/forum-on-tax-administration/tax-technology-tools-and-digital-solutions/, Table DM3 (accessed on 13 May 2022).
Figure 6.3. Use of artificial intelligence, 2022
Note: The figure is based on ITTI data from 52 jurisdictions that are covered in this report and that have completed the global survey on digitalisation.
Source: OECD et al. (2022), Inventory of Tax Technology Initiatives, https://www.oecd.org/tax/forum-on-tax-administration/tax-technology-tools-and-digital-solutions/, Table DM3 (accessed on 13 May 2022).
Box 6.3. Examples - Using analytics to sharpen the targeting of risks
Australia - Electronic compliance checks and real-time audits
The ATO are trialling a new feature for taxpayers as part of the lodgement process for income tax returns to help them comply with their tax obligations. The intent of this new feature is to increase positive voluntary compliance for small businesses by addressing risks at time of lodgement using real time analytics and nudge messaging. If there is an anomaly based on industry benchmarks, the system will pop up with a message that prompts the client to double check their figures before submitting their income tax return. When a business is operating significantly outside the key benchmark range for their industry, this does not automatically mean they have done anything wrong, however it does indicate something is unusual and may prompt us to contact them for further information in the future. An example nudge message might be “Your cost of sales amount of <xx> and total expenses amount of <xx> are higher than the small business benchmarks for your circumstances. Please review these amounts and the income that has been reported.”
At present this feature is only available to individuals running a business, however, even with this limited trial the ATO have seen about 25% of clients who have received a nudge message go back and make an adjustment to their income tax return before they finalise the lodgement. This has resulted in savings to the taxpayer and the ATO through the prevention of subsequent compliance activity.
Bulgaria – Using machine learning to identify missing traders
To combat VAT fraudulent transactions the Bulgarian National Revenue Agency (BNRA) for over ten years has been using a two-step process. First, BNRA computes a rules-based system risk score for the universe of the VAT-registered taxpayers in the country. Following this, the top-ranked taxpayers are further analysed with the view of establishing if each of them is a part of a missing trader chain. This two-stage process therefore discretely combines data analytics with operational knowledge. As such it is highly labour-intensive as requires input from the most experienced BNRA tax professionals. In 2019 a co-operation was established between the BNRA and the Tax Administration Research Centre (University of Exeter Business School), and University College London. The main objective of this co-operation was to develop a scalable predictive model, which automatically identifies missing traders. A key aspect of the methodological approach developed is that the network structure of all VAT transactions (through sales/purchases invoices) provides important information which can (and should) be used to identify abnormal VAT transactions through the production chain.
In early 2021, BNRA started implementing the predictive model, partially utilising the information structure emerging from the rich VAT transactions data collected. The first results have shown that such an approach can achieve both faster identification of fraudulent VAT transactions, but also a significantly higher proportion of identified real missing traders. This means the time a missing trader begins its fraudulent activities and the time it is identified by BNRA is on average reduced by 15%. Importantly, the predictive model can now identify fraudulent transactions without relying solely on the established risk rules used in risk assessment, thereby making identification more flexible.
Chile – Detecting VAT non compliance
Chilean tax legislation allows for the VAT on specific expenses borne by taxpayers to be used as credits against VAT and other special taxes (in addition to input VAT). However, Chile does not have information that confirms these expenses, creating opportunities for fraud. As a result, a mathematical model has been generated to estimate the probability that such "special credits" are fraudulent, which together with the business analysis, allows for improved risk analysis. Some taxpayers also declare VAT credits for purchases that are not related to their business activity; usually personal expenses. Chile is therefore using models to analyse electronic invoices and estimate whether these purchases are from business activity.
In addition, six analytical monitoring dashboards have been prepared covering: VAT; Special VAT Credits; Issuance of Electronic Invoices; Income and Expenses; Electronic Fee Tickets; and Financial Ratios for Large Taxpayers. Each dashboard has a series of filters, which allow the analyst to explore different segments of taxpayers, and detect abnormal behaviour, trends, patterns, etc. This is also used to prioritise taxpayer segments for analysis, allowing a more efficient use of resources.
China (People’s Republic of) – Selecting cases for e-audit
With the rapid development of cloud computing, machine learning, data mining and other technologies, the ‘Audit big data case selection and case study and judgment system’ developed by the State Tax Administration (STA) makes full use of big data for data modelling to accurately explore various high-risk cases, and identify the key targets of tax audit.
There are two main advantages of the system:
It is supported by various national tax data sources collected by the STA as well as by visualization technology to showcase taxpayers' tax related behaviours in multiple dimensions. Users can intuitively identify the risk characteristics of taxpayers and flag their illegal behaviours;
The process of case source selection is integrated with advanced machine learning and data mining algorithms to improve the efficiency of data analysis, increasing the accuracy of risk assessment.
Since the launch of the system, this work has in particular played an important role in the two-year project to tackle false VAT invoices and export VAT refund fraud, compliance in the cultural and entertainment industry and the investigation of major cases.
Georgia - System of data processing and analysis
The creation of data processing and analysis system is part of a wider data warehouse development project that was completed in 2021. The ultimate and main goal of the program is to collect and sort data available from the Revenue Service database and other third-party data sources and use them to strengthen the analytical capabilities of the Revenue Service.
As part of the creation of a system for data processing and analysis, the data processing and analysis system infrastructure were optimized. Different types of standard reports have been created in the data processing and analysis system, which are updated on a daily basis. Tax risk modules have been implemented and automated, which allows a systematic assessment of the tax risk of taxpayers and tax returns, using specific logic and indicators.
The programme has already reduced the time required to process specific information, as well as the risk of human error related to information retrieval and processing. It has also allowed prompt monitoring of important ongoing projects (such as the automatic VAT refund process), and the continuous receipt and processing of information from third parties. Furthermore it is being used for the operational assessment, ranking and preparation of specific tax risks (e.g. undeclared inventories).
Sources: Australia (2022), Bulgaria (2022), Chile (2022), China (People’s Republic of) (2022) and Georgia (2022).
Taxpayer programmes
Another approach for targeted risk management is the creation of units looking into the tax affairs of specific taxpayer segments. Two specific areas where tax administrations have found it advantageous to manage specific groups of taxpayers on a segmented basis are large business taxpayers, and high net wealth individuals (HNWIs). The rationale for focusing administration resources on managing these groups revolves around the:
Significance of tax compliance risks: due to the nature and type of transactions, offshore activities, opportunity and strategies to minimise tax liabilities; and in the case of large business, the differences between financial accounting profits and the profits computed for tax purposes.
Complexity of business and tax dealings: particularly the breadth of their business interests and in the case of HNWI, the mix of private and tax affairs.
Integrity of the tax system: the importance of being able to assure stakeholders about the work undertaken with these high profile groups of taxpayers.
Additionally, in the case of large taxpayers, a small number of taxpayers are typically responsible for a disproportionate share of tax revenue collected. Data collected as part of the 2021 International Survey on Revenue Administration (ISORA) indicates that for most jurisdictions between 30% and 60% of their total net revenue, including withholding payments on behalf of employees, was received from taxpayers covered by their large taxpayer programmes (see Figure 6.4). On average, 2.3% of corporate taxpayers covered by those programmes account for 43% of all revenue collected (see Table 6.2).
Table 6.2. Importance of large taxpayer offices / programmes (LTO/P), 2020
Note: The table shows the average percentages across the jurisdictions that were able to provide the information.
Source: Table D.9 Segmentation ratios: LTO/Ps.
While the management of these groups of taxpayers is often undertaken as a programme, in a large number of jurisdictions these programmes are also structural involving a Large Taxpayer Office or HNWI unit. The scope of the work of these units varies considerably, ranging from undertaking traditional audit activity, through to “full service” approaches (see Figure 6.5). However, on average two-thirds of tax administration staff in large taxpayer offices or programmes are working on audit, investigation and other verification related issues (see Table 6.2).
Figure 6.4. Percentage of revenue administered through large taxpayer offices/programmes, 2020
Source: Table D.9 Segmentation ratios: LTO/Ps
Figure 6.5. Large taxpayer offices / programmes: Existence and functions carried out, 2020
Source: Table A.15 Large taxpayer office / program: Existence and functions
Figure 6.6. HNWI programmes, 2020
Source: Table A.18 High net wealth individuals (HNWIs) program.
Planning for future risks
While it is key for tax administrations to understand current compliance risks and prepare appropriate response strategies, it is equally important to understand and prevent risks which may arise in the future. The increasing availability of data along with the enhanced capacity of tax administrations to handle and analyse that data allows tax administrations to more robustly assess future tax risks. Figure 6.2 highlights the large number of tax administrations who engage in forecasting, which is putting them in a position to assess where new compliance risks may arise and develop in time the necessary mitigation strategies.
This is particularly important as jurisdictions emerge from the COVID-19 pandemic. Tax administrations report that the pandemic influenced taxpayer’s compliance behaviour as government lockdowns and related measures affected the income streams of many taxpayers, resulting in reduced profits or even losses. As most administrations reduced or suspended compliance activities, this also impacted the data available to accurately assess risk. The sophisticated modelling analytics and modelling skills that tax administrations built up before the pandemic have been used to respond to these challenges, and to take account of any changes in taxpayer behaviour.
An interesting development within tax administrations is the recognition that the power of data analysis needs to be decentralised and spread more widely across the organisation. Through this tax administrations can be ready to identify emerging risks more quickly, and identify possible early interventions. As a result, tax administrations are now also exploring how artificial intelligence can be embedded into compliance processes across the organisation, and it is likely that this will be central to the digital transformation of compliance management, and risk management in the future. Examples of this can be seen in Box 6.4 below.
Box 6.4. Examples - Embedding AI across the organisation
Canada - Advanced Data Analysis Techniques For Assurance Engagements
The CRA is empowering its assurance and advisory professionals to incorporate artificial intelligence and data analytics into their work. The CRA has evolved from using traditional software, marketed specifically for internal auditors, and has invested in learning how to use tools that are more versatile to power its analytics. This has given the CRA an opportunity to use the many robust natural language processing, machine learning, and data visualization packages available in these tools to help transform how the CRA accesses, processes, and analyzes data.
For instance, using the new tools to deploy topic modelling, abstractive text summarization, and sentiment analysis has improved the CRA’s ability to quickly identify themes and meaning in unstructured data, such as interview notes and high volumes of web articles and documents, to support assurance and advisory engagements. This has enabled the CRA’s assurance and advisory professionals to perform faster and more comprehensive reviews.
Furthermore, data visualization and interfacing tools using open source programming code have opened up opportunities for the CRA to create products to help non-technical users in the organization incorporate structured data into their reports more easily. One such example is a technique that the CRA has developed to help show interconnections between corporate risks in a visually appealing manner to promote further analysis and clearer reporting.
Singapore – Developing an AI strategy
The Inland Revenue Authority of Singapore (IRAS) is scaling up the use of data and AI across all functions, including through the development of new AI solutions and Minimum Viable Products (MVP) to better manage tax non-compliance and improve service delivery. IRAS’ AI initiatives include:
Anomaly Detection – Traditionally, case selection models under Corporate Income Tax were built based on supervised learning and therefore could only detect known patterns by learning from the past. To uncover unknown risks and to complement existing models, IRAS adopted anomaly detection techniques and developed an MVP to detect anomalies in Corporate Income Tax Returns for officers’ review. The MVP has successfully identified several anomalous cases which could potentially result in notable tax recoveries.
Intelligent Network Analysis Tool – IRAS employed graph database as a solution to address challenges faced by auditors when analysing complex layers of relationships and networks. Instead of relying on commercial tools, IRAS implemented an in-house network visualizer with graph database as the underlying technology to address the auditors' needs. This new tool provides auditors with customised functionalities to analyse intricate, multi-layered relationships between entities during audits/investigations. It can also uncover relationships more than 10 connections deep in a real-time manner.
Service Quality Monitoring – IRAS has developed Natural Language Processing models to rate 100% of live chats for service quality monitoring. The new AI solution can boost objectivity and productivity of service quality monitoring, enabling IRAS to identify chats that require attention and take proactive actions to improve the service experience.
Sources: Canada (2022) and Singapore (2022).
Delivery of compliance actions
The type of “compliance actions” undertaken by tax administrations to determine whether taxpayers have properly reported their tax liability is changing. As set out earlier, the increasing availability of data and the introduction of sophisticated analytical models are allowing administrations to better identify returns, claims or transactions which might require further review or be fraudulent. Furthermore, these models, many of which can operate in real-time, are allowing administrations to conduct automated electronic checks on all returns or on transactions of a particular type.
Electronic compliance checks
While traditional audits (including comprehensive, issue or desk audits) are often the primary verification activities, the use of automated electronic checks or using rules-based approaches to treat some defined risks (e.g. automatically denying a claim, issuing a letter or matching a transaction) is providing administrations with more effective and efficient ways to undertake some of this work. Box 6.5 sets out examples of this.
These approaches do, however, raise the question of how to reflect those automated electronic checks in the performance information that administrations report in ISORA data. To include all checking may distort coverage, adjustment and yield rates. However where it replaces previously undertaken manual actions it would seem appropriate to reflect what administrations are now doing in this area.
In this respect, the 2021 version of the ISORA survey invited participants to break down the total value of additional assessments raised from audit and verification actions into (i) audits and (ii) electronic compliance checks (defined as electronic checks, validation and matching of taxpayer information).
Only a few administrations were able to provide information on electronic compliance checks. However, for some of those administrations electronic compliance checks make-up an important part of the additional assessments raised through all audits and verification actions. (See Table A.34)
Box 6.5. Brazil - Electronic compliance checks
Brazil - High Performance Inspection work processes
The Internal Revenue Service of Brazil (RFB) uses an internal development platform, which brings together a set of tools for data crossing, data mining, graph analytics and application of some AI techniques. This platform also uses a big data environment to perform queries on large tables, with volumes in the order of petabytes. For example, the electronic invoice table totals trillions of records and hundreds of pieces of information about each one. In addition to ready-to-use tools, the development platform allows tax administration members to build their own tools or improve existing ones. It can be accomplished by writing new scripts, shared in a collaborative space, and cataloged to be used as automation assets. For those who are not proficient in programming languages, they can use a 'no-code' programming style created on the platform under the name of 'Visual Script'. This platform facilitated a strategy called 'High Performance Inspection' (FAPE) where tax intelligence was combined with the Big Data environment. This meant multiple regional teams could collaborate and perform cross-reference on different databases, including digital tax bookkeeping, digital accounting, electronic invoices, financial movement data, registration data, among others. FAPE also included the automatic generation of letters to taxpayers in self-assessment where there are divergences on the declared values from the data held. The high level of automation optimizes the use of the workforce, making it possible to dramatically increase the fiscal presence of the small regional teams.
In recent years, an increasing numbers of taxpayers have been summoned by means of these automatic notifications. In 2021, there were more than 40 000 taxpayers summoned, totaling BRL 7.4 billion in amounts subject to self-assessment, which would not have been possible without the automation provided by the FAPE work.
Sources: Brazil (2022).
Audits
While previous ISORA surveys distinguished between audit adjustment rates by audit type, this has changed with the introduction of the 2020 ISORA survey. Since then administrations are invited to provide information for all audits combined. A data comparison is therefore only possible for the years 2018 to 2020.
Table 6.3. Audit adjustment rates and additional assessments raised
Note: The table shows the average audit adjustment rates and additional assessments raised through audits (excluding electronic compliance checks) for those jurisdictions that were able to provide the information for the years 2018, 2019 and 2020. Data for India has been excluded from the calculations for additional assessments raised as it would distort the average ratios. The number of jurisdictions for which data was available is shown in parenthesis.
Source: Table D.22 Audits: Hit rate and additional assessments raised.
Looking at the data, there are some general observations that can be made:
On average, audit adjustment rates have remained stable over the period 2018 to 2020 (see Table 6.3). However, the rates vary significantly across the administrations covered by this report ranging from as low as 10% in Slovenia and Norway to as high as 95% and more in Brazil, Bulgaria and Morocco (see Figure 6.7.). (High adjustment rates can of course result from highly targeted audits.)
The importance of audits can also be seen when looking at the additional assessments raised. On average, the additional assessments raised from audits correspond to around 4.5% of total revenue collections. This has been relatively flat over the years 2018, 2019 and 2020 (see Table 6.3). Looking at the jurisdiction level data, it can be seen that there are significant differences across the 51 administrations that were able to provide data for 2020 (see Figure 6.8).
Breaking this down by tax type, it shows that the ratio of additional assessments raised to tax collected is the greatest for corporate income tax (CIT). On average, CIT additional assessment raised as a percentage of CIT collected is 11.7%, more than double the percentage for value added tax (4.6%) and more than four times the percentage for personal income tax (2.5%) (see Figure 6.9).
In many jurisdictions, the additional assessments raised through large taxpayer offices or programmes (LTO/P) make-up a significant share of the total additional assessments raised from audits (see Figure 6.10). On average, LTO/Ps contribute around 30% of the total additional assessments raised from audits (see Table 6.2).
Figure 6.7. Audit adjustment rates, 2020
Source: Table D.22 Audits: Hit rate and additional assessments raised.
Figure 6.8. Additional assessments raised through audit as percentage of tax collections, 2020
Source: Table D.22 Audits: Hit rate and additional assessments raised.
Figure 6.9. Additional assessments raised through audit as percentage of tax collected by tax type, 2020
Source: D.23 Audits: Additional assessments raised by tax type.
Figure 6.10. Additional assessments raised from audits undertaken by LTO/P as a percentage of additional assessments raised from all audits, 2020
Source: Table D.9 Segmentation ratios: LTO/Ps.
Moving audit work to a virtual environment
Traditionally, administrations apply a variety of different audit types including comprehensive audits, issue-oriented audits, inspections of books and records, and in-depth investigations of suspected tax fraud. Often those audits require the administration to visit the taxpayer’s premises (so called field audits).
Advances in technology have led administrations to consider new ways of engaging with taxpayers during the audit process including the electronic submission of audit related documentation. This trend was accelerated as a result of the COVID-19 crisis as the closure of tax offices and the move to remote working for large numbers of tax officials changed how they approached audits.
This was observed in the 2021 OECD report Tax Administration: Digital Resilience in the COVID-19 Environment (OECD, 2021[6]) which noted of the 32 administrations covered by that report, close to ninety percent shifted parts of their field audit work to a virtual / digital environment. Moreover, 76% of those administrations plan to continue moving field audit work to a virtual/digital environment going forward. This is supported by an increased use of technology in audits which is helping drive efficiency. Box 6.6 highlights some leading practices, and the example from Estonia is of note as it is helping drive upstream compliance.
Box 6.6. Examples - Technology in audits
Chile – Tax radar
In Chile, the Servicio de Impuestos Internos (SII) is using data-driven decision making by providing, in the most intuitive way possible, large amounts of data for staff to use in their work. The Tax Radar dashboard aims to generate a view that allows staff to identify groups of relevant taxpayers on which the SII might not be carrying out control actions. The user can see the full-range (universe) of taxpayers and manipulate it through different institutional categories, such as global risk classification, monetary-impact for the tax system, type and size of business, current status and jurisdiction. It also incorporates different types of non-compliance, allowing the users to drill down on the information and establish different types of treatment action for the group or situation. The dashboard can also be used as a starting point to a more detailed analysis or as a knowledge management tool. This dashboard, and the analytics behind it, helps the SII save around 1 000 hours in analytical workload a month.
See Annex 6.A for supporting material.
Estonia - The e-service of tax behaviour rating
In Estonia, the Estonian Tax and Customs Board (ETCB) has developed a new e-service – Tax Behaviour Ratings which is available since summer 2020 for businesses. In this service, a variety of data is used as an input to create a risk assessment of the business. This includes data on turnover, number of employees, average salary, tax debts, tax compliance, and the company directors. A tax compliance behaviour rating is then calculated, which includes the risk of tax audit for each legal person registered in Estonia.
Taxpayers can then see themselves and their business partners through the eyes of the ETCB as well as comparing themselves with their peers. The taxpayer can also share their ratings with other users of the service, or publically, so that they can demonstrate they are a compliant business. This helps build trusted relationships and incentivise compliant behaviour.
(See Annex 6.A for links to supporting material.)
Georgia – Audit case management system
In order to automate and effectively implement tax audit processes, in 2018 the Revenue Service of the Ministry of Finance of Georgia launched the development of a new electronic audit system. The purpose of the system is:
Effective Process Management and Monitoring - The system gathers all the audits in the Audit Department (both completed and ongoing) and the manager has the opportunity to receive complete information about the audit, according to the audit processes, including activities carried out within the framework of the tax inspection as well as the obtained and created documentation.
Analysis and Feedback of Audit Results - A comprehensive analysis of completed tax audits is possible in the system, including an analysis of identified tax liabilities by areas of activity. It is also possible to analyse information in relation to the identified risks, which allows for further planning.
Allocation of resources - Managers can see how many issues an auditor is inspecting, and the complexity of those issues. This can help in case allocation, as the system reflects the time spent by the auditor on the tasks performed.
Sources: Chile (2022), Estonia (2022) and Georgia (2022).
Tax crime investigations
Tax crime refers to a conduct that violates a tax law and can be investigated, prosecuted and sentenced under criminal procedures within the criminal justice system. There is a range of organisational approaches for conducting tax crime investigations and the ISORA 2021 survey looked at the responsibility for directing and conducting those investigations.
The information gathered through the ISORA 2021 survey shows that 55% of the tax administrations covered in this publication are involved in conducting tax crime investigations (Table A.36). The majority of those administrations have responsibility for both conducting and directing tax crime investigations, while the others have responsibility for solely conducting investigations, under the direction or authority of another agency, such as the police or public prosecutor (see Figure 6.11).
Box 6.7. Recently published OECD reports on Tax Crime
Fighting Tax Crime – The Ten Global Principles, Second Edition
First published in 2017, Fighting Tax Crime - The Ten Global Principles (OECD, 2021[7]) is the first comprehensive guide to fighting tax crimes. It sets out ten essential principles covering the legal, institutional, administrative, and operational aspects necessary for developing an efficient and effective system for identifying, investigating and prosecuting tax crimes, while respecting the rights of accused taxpayers. The second edition addresses new challenges, such as tackling professionals who enable tax and white-collar crimes, and fostering international co-operation in the recovery of assets. Drawing on the experiences of jurisdictions in all continents, the report also highlights successful cases relating to the misuse of virtual assets, complex investigations involving joint task forces, and the use of new technology tools to fight tax crimes and other financial crimes. The second edition is joined by a series of “country chapters”, detailing jurisdictions’ domestic tax crime enforcement frameworks as well as the progress made in implementing the Ten Global Principles.
Ending the Shell Game: Cracking down on the Professionals who enable Tax and White Collar Crimes
White collar crimes like tax evasion, bribery, and corruption are often concealed through complex legal structures and financial transactions facilitated by lawyers, accountants, financial institutions and other "professional enablers" of such crimes. These crimes have significant impacts on government revenue, public confidence and economic growth, including the recovery from COVID-19. The report Ending the Shell Game: Cracking down on the Professionals who enable Tax and White Collar Crimes (OECD, 2021[8]) highlights the damaging role played by professional enablers, and the importance of concerted domestic and international action in clamping down on them. Drawing from the experience from several jurisdictions, it also includes recommended counter-strategies for deterring, disrupting, investigating and prosecuting the professionals who enable tax and white collar crimes.
In the cases of administrations that do not have any responsibility for conducting tax crime investigations, this work is done by another agency, such as the police or public prosecutor. This could also be a specialist tax agency, established outside the tax administration.
Figure 6.11. Role of administrations in tax crime investigations, 2020
Note: In some jurisdictions, the organisational approach for tax crime investigations may depend on the tax offence or tax-related criminal proceedings. In those cases, an administration may have selected multiple answer options. This is why the percentages add up to more than 100%.
Source: Table A.36 Tax crime investigations: Role of the administration and number of cases.
Table 6.4 shows the total number of cases referred for prosecution during the fiscal year for the 32 administrations that have responsibility for conducting tax crime investigations. While the number of cases referred for prosecution was similar in 2018 and 2019, there was a significant reduction in the number of cases referred for prosecution during 2020.
This is also reflected in the jurisdiction level data, which shows that close to three quarters of administrations that have responsibility for conducting tax crime investigations referred fewer cases for prosecution in 2020 (see Table A.36).
Table 6.4. Tax crime investigation cases referred for prosecution
Note: Only includes administrations that have responsibility.
Source: Table A.36 Tax crime investigations: Role of the administration and number of cases.
There could be many reasons for this reduction. This could include a genuine decline in cases, administrations reducing staff in this area as part of a wider reallocation of resources due to the pandemic, or the pandemic may imposed constraints on the ability to refer cases for prosecution. Future editions of this series will be able to identify if the reduction this year was a ‘blip’ caused by the pandemic or the start of a long-term trend.
Finding better ways to fight tax crime is a high priority as money laundering, corruption, terrorist financing, and other financial crimes can threaten the strategic, political and economic interests of jurisdictions. Tax administrations, as gatekeepers to a sound financial system, play a critical role in countering these activities and are in possession of information that could be crucial for a successful criminal tax investigation. Fighting tax crime requires improved transparency and greater efforts to harness the capacity of different government agencies, including across borders, to collectively deter, detect and prosecute these crimes through a whole of government approach and international co-operation, as Box 6.8 illustrates.
Box 6.8. Norway – Emerging risks
A threat assessment the Norwegian Tax Administration (NTA) made available in late 2021 showed increased risks of tax and work-related crime, post COVID-19. The digital shift has also contributed to the creation of new digital platforms used by criminals and professional enablers. Five main threats stand out: criminal networks; misuse of financial support schemes; work-related crime, professional enablers, and the opportunities of illegal activities through digitalisation. NTAs approach involves being prepared, cross-disciplinary collaboration, and as far as possible achieving real-time developments and changes.
Regarding criminal networks, NTA anticipate that these networks will use digital platforms to a greater extent in future as a means to carry out and coordinate their criminal activities. Serious crime is often systematic by nature. This may include cross-border activities in combination with other types of crime, and the use of other peoples “digital verified identities”.
NTA have also seen that public registers are vulnerable and can be manipulated by criminal actors, for example via incorrect registrations and reporting. Some actors have and are expected to deliberately exploit the arrangements granting them payment deferrals before claiming insolvency. Different professional enablers such as lawyers, accountants and other roles within finance are considered to contribute to hiding assets or concealing the real ownership in connection with bankruptcy.
Source: Norway (2022).
References
[8] OECD (2021), Ending the Shell Game: Cracking down on the Professionals who enable Tax and White Collar Crimes, OECD, Paris, https://www.oecd.org/tax/crime/ending-the-shell-game-cracking-down-on-the-professionals-who-enable-tax-and-white-collar-crimes.htm (accessed on 13 May 2022).
[7] OECD (2021), Fighting Tax Crime – The Ten Global Principles, Second Edition, OECD Publishing, Paris, https://doi.org/10.1787/006a6512-en.
[6] OECD (2021), “Tax Administration: Digital Resilience in the COVID-19 Environment”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/2f3cf2fb-en.
[4] OECD (2020), Code of Conduct: Co-operation between tax administrations and sharing and gig economy platforms, OECD, Paris, https://www.oecd.org/tax/forum-on-tax-administration/publications-and-products/code-of-conduct-co-operation-between-tax-administrations-and-sharing-and-gig-economy-platforms.pdf (accessed on 13 May 2022).
[3] OECD (2020), Model Rules for Reporting by Platform Operators with respect to Sellers in the Sharing and Gig Economy, OECD, Paris, https://www.oecd.org/tax/exchange-of-tax-information/model-rules-for-reporting-by-platform-operators-with-respect-to-sellers-in-the-sharing-and-gig-economy.htm (accessed on 13 May 2022).
[2] OECD (2019), The Sharing and Gig Economy: Effective Taxation of Platform Sellers : Forum on Tax Administration, OECD Publishing, Paris, https://doi.org/10.1787/574b61f8-en.
[1] OECD (2017), The Changing Tax Compliance Environment and the Role of Audit, OECD Publishing, Paris, https://doi.org/10.1787/9789264282186-en.
[5] OECD (2016), Advanced Analytics for Better Tax Administration: Putting Data to Work, OECD Publishing, Paris, https://doi.org/10.1787/9789264256453-en.
Annex 6.A. Links to supporting material (accessed on 13 May 2022)
Box 6.1. – Italy: Link to a document with more details on the new methodology for the estimation of the VAT gap: https://www.oecd.org/tax/forum-on-tax-administration/database/b.6.1-italy-vat-gap-estimation.pdf
Box 6.2. – Argentina: Link to a presentation providing additional information on the Simple and Pro-Forma VAT Tax Return: https://www.oecd.org/tax/forum-on-tax-administration/database/b.6.2-argentina-simple-and-pro-forma-vat-tax-return.pdf
Box 6.2. – France: Link to a presentation that explains more about the process of detecting undeclared constructions: https://www.oecd.org/tax/forum-on-tax-administration/database/b.6.2-france-foncier-innovant.pdf
Box 6.2. – Lithuania: Link to a document that contains more details regarding the project to increase transparency of second hand car transactions: https://www.oecd.org/tax/forum-on-tax-administration/database/b.6.2-lithuania-car-dealerships.pdf
Box 6.2. – Spain: Link to a presentation with more detail regarding the project to analyse feedback from verification actions: https://www.oecd.org/tax/forum-on-tax-administration/database/b.6.2-spain-tax-audit-feedback.pdf
Box 6.6. – Chile: Link to a video providing more details on the Tax Radar dashboard: https://www.youtube.com/watch?v=Aic6RTav5AM
Box 6.6. – Estonia: Link to the tax administration’s website on the e-service of tax behaviour ratings, including explanatory videos: https://www.emta.ee/en/business-client/e-services-training-courses/advice/what-tax-behaviour-ratings-tell-you