Improving tax compliance and reducing the tax gap is typically one of the main objectives of a tax administration. Therefore, many jurisdictions are estimating the size of their tax gap. Although a tax gap may have a relatively simple definition, its estimation is complex and includes many nuances. This chapter provides an overview of key tax gap concepts and examples of international experiences in tax gap research.
Tax Administration 2024
Comparative Information on OECD and other Advanced and Emerging Economies
Tax Administration
Abstract
An increasing number of jurisdictions are estimating tax gaps as these estimates can provide insights on size and nature of non-compliance, emerging trends, and the general health of the tax system. To estimate the tax gap, a jurisdiction needs to take into account many aspects such as legislative frameworks the overall administrative design of their tax systems, internal operations, availability of data, and economic events. Moreover, jurisdictions need to capture unobserved events or deliberately hidden activities that add challenges to tax gap estimation.
This chapter provides an overview of:
Tax gap definition and its types;
Tax gap methodologies;
Tax gap components;
How tax gaps are used by tax administrations; and
The impact of the COVID-19 pandemic on the tax gap.
It also highlights international experience with tax gap research.1
Tax gap is generally defined as a difference between the potential tax revenue and actual tax revenue (Hutton, 2017[1]). It can be separated into two types: compliance gap and policy gap.
Compliance gap is a potential tax revenue loss due to non-compliance under current tax laws. One of the main functions of most tax administrations is reducing tax non-compliance, therefore, they mostly focus on a compliance gap rather than a policy gap (Hutton, 2017[1]). Moreover, most jurisdictions do not take into account illegal activities in their tax gap estimation due to uncertainties in taxes that could be applied.
Policy gap measures other tax revenue loss due to various tax policies. This gap can include intentional tax expenditure such as tax credits to achieve certain policy outcomes or unintentional tax expenditure such as tax avoidance due to loopholes in tax law (Hutton, 2017[1]). Since tax policies are often managed by a ministry of finance or a department of treasury, it is often out of the control of tax administrations.
Since a compliance gap is more common for tax administrations than a policy gap, the remainder of this chapter mostly focuses on methodologies and nuances for the compliance gap.
Belgium
The Belgian tax administration defines a policy gap as the difference between potential collections under the existing policy framework and potential collections under a normative or referential policy framework. Although often overlooked, policy gap estimations allow for a comparison of the sizes of the compliance gap and policy gap and their respective contribution to the total gap. Defining the gap in such a way allows decision makers to also focus on how policy choices have an effect on the total gap and if policy changes have an effect on the compliance gap.
Furthermore, the policy gap can be broken down into a “non-taxable” or a non-actionable gap, such as items that are included in the national accounts consumption statistics, but are not subject to VAT. These results can also be used in the modelling of certain tax expenditures. For example, it can give an overview of not only the size of the zero-rated goods and services, exemptions, but also of the size of public goods, non-market services.
In other words, a policy gap estimation offers an opportunity to increase budget transparency, to establish a closer link between tax expenditures and direct expenditures and potential avenues of actions for improving revenue performance by addressing either component of the gap.
European Commission
The European Commission’s annual study on the VAT gap in the EU (European Commission, 2023[2]) estimates the VAT policy gap in addition to the VAT compliance gap. The VAT policy gap is a proxy of the additional VAT revenue that could be generated if a uniform VAT rate – without any exemptions or reduced rates – were applied to all final domestic consumption of goods and services by households, government, and non-profit institutions serving households (NPISH), assuming full taxpayer compliance. In policy terms, this is a more “actionable” estimate than the estimate of the VAT compliance gap, because EU Member States have, in general, direct control over reduced rates and, to some extent, on exemptions.
The VAT policy gap consists of two main components: the VAT exemption gap, which accounts for revenues lost due to exemptions and exclusions from the tax base, and the VAT rate gap, which accounts for revenues foregone due to preferential treatment such as reduced rates and zero-rates. Changes in the size of the VAT policy gap can be attributed to changes in legal rules, such as national adjustments to reduced rates or exemptions, as well as shifts in demand composition.
Estimating the VAT policy gap provides transparency on the cost of policy choices that depart from uniform VAT rates and helps identify areas where tax policy changes could increase revenue and improve tax efficiency. The European Commission publishes annual estimates of the VAT policy gap and its components as part of its VAT gap in the EU study, which provides a standardised framework for EU Member States to assess their VAT systems. Over time, these estimates have been continuously improved in accuracy and detail, enabling policymakers to track progress and target reforms.
The latest report finds that the average VAT policy gap in the EU-27 stood at approximately 45 percent of the notional ideal revenue in 2022. The VAT rate gap has increased since 2020. In the same time frame, the VAT exemption gap, which constitutes the largest portion of the policy gap, has decreased. This decline in the VAT exemption gap is attributed to changes in demand composition, as there have been no significant changes to the VAT Directive regarding exemptions.
Note: While individual EU Member States are allowed to abolish reduced rates, most exemptions are mandated by EU law, specifically the VAT Directive (Council Directive 2006/112/EC). As a result, abolishing these exemptions would necessitate an amendment to the Directive at the European level, which demands unanimous agreement among EU Member States.
Sources: Belgium (2024) and the European Commission (2024).
Even though there is a common definition of a tax gap, it may not be exactly the same for various jurisdictions. For example, some jurisdictions include late payments as part of their tax gap while others do not. Others might not include a payment gap in their overall tax gap at all. In addition, tax gap estimates could be linked to a year of economic activity (e.g. tax year) or imply all incoming revenue during certain time frame (e.g. fiscal year). Therefore, a direct comparison of various jurisdictions should be interpreted with caution and account for differences in tax systems, definitions and methodologies.
A tax gap can be divided based on non-compliance sources such as registration, filing, reporting or payment (see Figure 11.1. and Figure 11.2.). Some jurisdictions use a term “lodgement” instead of “filing” indicating a process of filing or lodging a tax form. Jurisdictions may also use different names for their tax gaps based on the sources of non-compliance. For example, a tax gap related to payment non-compliance could be called a payment gap, underpayment gap, non-payment gap or collection gap. The most popular terms for a tax gap related to reporting non-compliance are a reporting gap, underreporting gap and assessment gap.
Most of the jurisdictions are focusing on estimating reporting (75%) and payment (64%) non-compliance as they are major sources contributing to the tax gap. (See Annex Table 11.A.2.) Moreover, registration and filing non-compliance can be more challenging to estimate as they cover a population that may be unknown to the tax administration.
Source: Annex Table 11.A.1 and Annex Table 11.A.2.
To provide a clearer view on what exactly is included in the tax gap, jurisdictions define gross tax gap versus net tax gap (see Figure 11.3.). Most jurisdictions estimate the gross tax gap (82%) and half measure the net tax gap. Close to 40% measure both gross and net tax gaps. Net tax gap usually includes some challenges as there could be a lag time in completing compliance and collections activities. (See Annex Table 11.A.2.)
Some jurisdictions (36%) use projections at least for one tax gap component to balance timeliness with other considerations such as the quality of tax gap estimates. (See Annex Table 11.A.3.)
Source: Annex Table 11.A.2.
A small number of jurisdictions publish their overall tax gap estimates in the public domain (29%) and some of them have a legal requirement to do so. Several jurisdictions do not publish their overall tax gap, but they make one tax gap component (usually, Value Added Tax) publicly available. The frequency of the publications varies from annual to every four years or it is on an irregular schedule. Almost 40% of the jurisdictions that publish their tax gap estimate, do so annually. (See Annex Table 11.A.1.)
In the United States, the Internal Revenue Service (IRS) estimates the gross tax gap that comprises of three components:
Non-filing: tax not paid on time by those who do not file on time,
Underreporting: tax understated on timely filed returns, and
Underpayment: tax that was reported on time, but not paid on time.
The majority of the tax gap comes from underreporting gap (around 80%), and non-filing and underpayment are smaller sources (on average, 9% and 11% respectively) that contribute to the gross tax gap.
Tax gap analysis has consistently shown that compliance is higher when income is subject to information reporting and even higher when also subject to withholding.
Source: The United States (2024).
Tax gap estimation has become a more popular topic for discussion among tax administrations in recent years. However, some administrations have had tax gap teams for more than 40 years (see Table 11.2.).
Note: Jurisdictions that estimate the overall tax gap have an asterisk (*) and jurisdictions that publish their overall tax gap estimates are bolded.
Source: FTA 2023 survey on tax gap estimations.
According to a presentation by the IMF during the 2024 meeting of the FTA Community of Interest (COI) on Tax Gap, the main factors to succeed in tax gap research are data, a methodology, human resources, management support, and institution orientation (see Table 11.3.).
Source: IMF presentation at the 2024 meeting of the OECD’s Forum on Tax Administration Community of Interest on Tax Gap.
Tax gap estimation is complex and requires nuanced analysis. In general, there are two main approaches to estimating the tax gap (see Figure 11.4.):
Top-down methodologies: Generally using aggregated macro-economic data (e.g. national accounts data) to estimate the size of the tax base and the theoretical tax liability. The difference between the theoretical tax liability and the actual amount of tax paid or reported is the estimated tax gap.
Bottom-up methodologies: Generally using micro-economic data (for example, audit data) to extrapolate potential non-compliance and estimate the tax gap. The most common data sources for these methodologies are either data from random audits or risk-based audits.
Most jurisdictions (89%) prefer using top-down approaches, and slightly more than half (57%) use bottom-up approaches. A top-down approach is usually a good starting point for jurisdictions new to tax gap estimation. In addition, jurisdictions tend to start with the VAT gap which has a fairly established top-down methodology. Some administrations (54%) use third-party support for the tax gap estimation such as from IMF, academics, a ministry of finance, independent consultants or other organisations. (See Annex Table 11.A.1. and Annex Table 11.A.3.)
Top-down methodology
Ideally, a jurisdiction estimates the tax gap using both approaches as they provide different insights. A top-down methodology can provide additional information about the whole population that might be unknown from the operational data and give an overview of non-compliance and risk sectors in the economy. However, it is limited and cannot provide more detailed findings. The insights from top-down methodologies can be used to inform compliance strategies at a high level.
A top-down methodology could be used as an alternative to bottom-up approaches when random audits are unavailable and risk-based audit results cannot be extrapolated to the taxpayer population. At the same time, it could be used as a complement to view a tax gap from another prospective.
Bottom-up methodology
A bottom-up methodology usually provides more insights on non-compliance at the micro level. Therefore, it can be disaggregated and could be used to improve risk-assessment processes. However, estimates from this methodology could be limited to specific populations or taxes. Bottom-up methodologies require statistical or econometrical expertise and tax administration operational knowledge with good data understanding (OECD, 2017[3]).
One of the main data sources for bottom-up approaches is random audit, but it may not always be available in all jurisdictions. Therefore, some tax administrations use operational audits (for example, data from risk-based audits) or other available micro-level data. For example, some bottom-up methodologies make use of Census data, third-party information reporting data and administrative filing and payment data instead of audit data.
Random audits
Random audits are usually conducted based on a random sample drawn from the population of taxpayers. There are different types of random sampling such as simple random sampling, systematic sampling, stratified sampling and clustering sampling, where simple random sampling and stratified sampling are the most common among tax administrations. Some of these sampling techniques are explained in two Technical Guidance Notes published by the IMF in 2021 (Thackray, Jennings and Knudsen, 2021[4]) and 2023 (Barra, Hutton and Prokof’yeva, 2023[5]).
Random audit programmes are considered a high-quality method to estimate tax gaps in large populations of registered taxpayers (OECD, 2017[3]). The results from such audits can also help identify emerging trends in non-compliance for all taxpayers, and can help verify existing risk selection criteria, including whether they are still relevant and optimal.
The main challenges with random audits are typically related to the need for additional resources to conduct the audits and, as those audits are not risk-based but random, often not much additional tax revenue is identified. However, by using random audit results in improving risk-assessing systems, a tax administration can potentially become more efficient at operational audits, allocating resources to higher risk areas and recovering more taxes. Also, some sampling designs can help to reduce costs of random audits (for example, stratifications by risks, see Denmark’s example in Box 11.3.).
Denmark – Stratification by risks for a random audit programme for private individuals
The Danish Tax Administration stratifies or groups its private individual taxpayer population by risks and then draws a random sample of each risk group. To balance between recovering higher tax revenue and receiving insights on the overall taxpayer population, larger samples are drawn from higher risk groups than from low-risk ones. In that case, the results can still be extrapolated to the overall population, but at the same time, there is more efficient audit resource allocation than in a regular random audit program. This can also lead to higher returns on investments.
Denmark uses the results of random audits as an input to new legislation and to risk-assessment for improving tax compliance.
Netherlands – Know your taxpayer, understand their behaviour
The Dutch Tax Administration focuses on random audits as it can be an instrument for improving compliance risk management. Random audits not only give insight in compliance levels among populations, but also in what area of behaviour the compliance level could be improved. This creates the opportunity to develop specific interventions to target behaviour.
Compliance risk management in the Netherlands consists of five steps, namely identification of uncertainties (possible mistakes), analysing underlying causes, prioritisation, treatment (intervention) and evaluation. The Netherlands’ random audits programme contributes to step 1, identification of uncertainties (and therefore compliance risks) and step 2, analysis of what causes the mistakes made to enable designing interventions that change non-compliant behaviour into compliant behaviour or to strengthen existing compliant behaviour.
Improving compliance behaviour depends on behavioural change. The administration uses the COM-B system as a model of behaviour that provides a basis for designing interventions aimed at behavioural change. In this behaviour system capability, opportunity and motivation interact to generate behaviour that in turn influences these components. Applying this to intervention design, the task is to consider what the behavioural target would be, and what components of the behavioural system would need to be changed to achieve that.
Firms, for instance, that are selected into the random audits programme are checked completely, that is on all types of taxes such as Value Added Tax and Corporate Income Tax. Therefore, random audits give a complete overview of tax behaviour of firms. But they do not give insight into why firms behave in this manner. To understand the behaviour of firms, information is collected using a survey among tax officers. Questions with respect to characteristics of firms, use of tax practitioners and judgements on what underlying causes might lead to mistakes made (for example, causes based on capability, opportunity and motivation such as lack of knowledge and financial constraint) are answered after every audit. These survey results help to understand the audit data and, hence, behaviour of firms.
The combination of insight in compliance levels due to audits and in underlying causes help to shape specific interventions to change the compliance behaviour of taxpayers.
Sources: Denmark (2024) and the Netherlands (2024).
Risk-based audits
Some jurisdictions that do not have random audit programmes rely on either top-down approaches or a use of risk-based audit results. Even though random audit methods are usually considered a high-quality method for tax gap estimation (OECD, 2017[3]), risk-based audits could provide additional insights on non-compliance due to deeper audit procedures.
However, risk-based audits tend to capture information on taxpayers with higher risks and, thus, these results are not representative of the overall population. The audited population is usually selected by specific criteria and not chosen randomly, leading to results with selection bias. This bias could be addressed by applying statistical or econometrical methods. The common methods include a Heckman method or Heckman Correction approach, extreme value method, cluster analysis, post stratification, and propensity score matching (Fiscalis TGPG, 2018[6]; Barra, Hutton and Prokof’yeva, 2023[5]).
On average, there are more bottom-up methodologies from random audits than from risk-based audits (see Figure 11.5.).
Source: Annex Table 11.A.4.
Non-detection multiplier
Some jurisdictions (21%) are applying a non-detection multiplier to their tax gap estimates (see Annex Table 11.A.3.). Non-detection multiplier or uplift factor is a multiplier that is applied to a tax gap estimate to account for undetected non-compliance not captured through tax gap estimates. For example, auditors may not always identify all sources of non-compliance when conducting audits due to various reasons. Therefore, any tax gap estimates from these audit data may be missing undetected non-compliance. Non-detection multipliers could be applied to some tax gap components or some methodologies and not necessarily to the overall tax gap.
There are several methods to develop a non-detection multiplier as can be seen in Table 11.4. One of the methods used is called the “Delphi technique” and tax administrations from both Sweden and the United Kingdom have published papers on the use of this method (HMRC, 2020[7]; Swedish Tax Agency, 2023[8]).
Sources: FTA 2023 survey on tax gap estimations and Thackray, M., S. Jennings and M. Knudsen (2021), The Revenue Administration Gap Analysis Program: An Analytical Framework for Personal Income Tax Gap Estimation, https://doi.org/10.5089/9781513577173.005.
United Kingdom – Non-detection multiplier using Delphi technique
Historically, the United Kingdom’s (UK’s) HM Revenue and Customs (HMRC) used multipliers derived from analysis by the IRS in the United States and adopted them to their random audit results. In recent years, HMRC have developed new non-detection multipliers using the Delphi technique to better apply to the types of risks seen in the UK tax system. These multipliers help adjust tax gaps for missing non-compliance in cases that were audited.
Non-detection arises for several different reasons, including the detection capability of the auditor, tax complexity, taxpayer co-operation, availability of data, available time to conduct the audit, and the level of concealed non-compliance.
The Delphi technique is a consultative method to gather expert opinion in a systematic way and establish consensus. The technique includes three rounds of questionnaires to acquire a consensus from a panel of experts in each tax regime. Response summaries are given at the beginning of last two rounds, where the panel could amend or agree their responses. In most recent years, a Delphi technique has been applied to estimate a non-detection multiplier for Pay-as-You-Earn (PAYE) employer compliance for small businesses and corporation tax for small businesses.
It is important to review a non-detection multiplier to adjust based on more recent information as detection may change over time, for example, due to improved compliance strategies. Non-detection multipliers can also differ for results from randomly selected audits and for risk-based audits. Further information is included in the HMRC Working Paper Non-detection multipliers for measuring tax gaps (HMRC, 2020[7]).
United States – Non-detection multiplier
The IRS tax gap estimates for the Tax Year (TY) 2001 incorporated Detection Controlled Estimation (DCE) for the first time. The first iteration of DCE using TY 2001 random audit data involved estimation for two categories of income (based on the extent of third-party information reporting) and for two categories of taxpayers (based on nature and size of income). The results were then synthesised down to four “multipliers”.
Further research determined that there was an opportunity to expand DCE to allow for greater variability in the average detection rates across line items. Beginning with TY 2006, the IRS moved away from explicit multipliers and implemented a microsimulation approach to allocating DCE estimates of undetected income. The IRS has continued research on refining the microsimulation approach, focusing on improving estimates of the distribution of the tax gap.
Sources: The United Kingdom (2024) and the United States (2024).
Challenges related to tax gap estimation
Main challenges for tax gap estimation are usually related to data, methodology, resources and legislation (see Table 11.5.). Jurisdictions try to learn from international best practices and apply various techniques to overcome these challenges, but some constraints may limit capacity for tax gap estimation.
Source: FTA 2023 survey on tax gap estimations
Tax systems vary, and therefore, tax gap components may differ between jurisdictions. In general, they can be divided into five main groups:
Personal income tax (PIT),
Corporate income tax (CIT),
Value-added tax (VAT), equivalent to Goods and Services Tax (GST) for some jurisdictions,
Excise taxes and duties, and
Other tax types.
All jurisdictions that completed the underlying survey estimate the VAT gap (except the IRS as the United States that does not have this type of tax in their federal tax system), but only 39% of them estimate the overall tax gap that includes a combination of major tax types (see Figure 11.6.). Some jurisdictions are at early stages of the development of their tax gap programme and focus more on the VAT gap with a well-established top-down methodology. In certain cases, the IMF helps estimate the VAT gap for jurisdictions new to the tax gap research.
The second and third most popular tax gap components are PIT and CIT gaps, where more than 50% of jurisdictions are estimating either one of them or both. Excise gap is the rarest estimate (estimated by 29% of jurisdictions), and it typically includes excise taxes and duties on multiple excise products. Other tax gaps are also not very common (estimated by 36% of jurisdictions) and sometimes may get deprioritised due to the relatively small scale of their contribution to the overall tax gap.
Some jurisdictions may not segregate the tax gap by tax types in the same way. For example, the Netherlands Tax Administration does not use their bottom-up approach to estimate a tax gap for a specific tax revenue such as VAT but rather estimates the total gap for each entity.2
Source: Annex Tables 11.A.1. to 11.A.13.
Personal income tax gap
The PIT gap usually includes tax non-compliance from private individuals (for example, from shadow economy activities, offshore investments, capital gains), self-employed individuals, and non-residents. A tax gap for self-employed individuals or small businesses could be included in the PIT gap or CIT gap, depending on the tax system and internal processes of a tax administration.
Almost 60% of jurisdictions estimate the PIT gap. Of those:
88% estimate the reporting gap and 69% estimate the payment gap; and
81% estimate the gross tax gap and half of them estimate the net tax gap.
Top-down and bottom-up approaches are equally popular, and random audits are prevalent in bottom-up methodologies. Top-down methodologies usually include global statistics, local aggregated data and academic research. See Figure 11.7. for details.
The IMF published a Technical Guidance Note containing various methodologies for PIT gap in The Revenue Administration Gap Analysis Program: An Analytical Framework for Personal Income Tax Gap Estimation (Thackray, Jennings and Knudsen, 2021[4]).
Number of jurisdictions that conduct the relevant analysis or apply the relevant methods
Note: 16 jurisdictions conduct PIT gap estimations.
Source: Annex Table 11.A.5 and Annex Table 11.A.6.
The Danish Tax Administration estimates the PIT gap for private and self-employed individuals using random audits. In addition to the random audit programmes, the part of the PIT gap attributed to undeclared work is estimated separately using the labour input method.
The application of the labour input method relies on data from the Danish Labour Force Survey, which is compared to Danish tax data. In short, undeclared work is assumed to be represented by the number of self-reported working hours in the Labour Force Survey that is in excess of the working hours officially recorded in the tax data. By calculating the discrepancy between the two, the prevalence and value of undeclared work in Denmark, together with the associated PIT gap, can be estimated.
Source: Denmark (2024).
Corporate income tax gap
The CIT gap usually includes tax non-compliance from large corporations, small and medium enterprises, public enterprises and institutions. Half of the jurisdictions estimate the CIT gap. Of those:
All jurisdictions focus on the reporting gap and 86% on the payment gap; and
93% estimate the gross tax gap and 64% estimate the net tax gap.
While jurisdictions use either bottom-up or top-down approaches, there are two jurisdictions that estimate the CIT gap using both approaches at the same time. On average, there are more bottom-up approaches used by jurisdictions. Bottom-up approaches are usually based on random or risk-based audit data. Local aggregated data is usually the main source for top-down approaches (see Figure 11.8.).
In addition, specific methodologies for CIT gaps can be found in the IMF Technical Guidance Note Corporate Income Tax Gap Estimation by using Bottom-Up Techniques in Selected Countries: Revenue Administration Gap Analysis Program (Barra, Hutton and Prokof’yeva, 2023[5]).
Number of jurisdictions that conduct the relevant analysis or apply the relevant methods
Note: 14 jurisdictions conduct CIT gap estimations.
Source: Annex Table 11.A.7 and Annex Table 11.A.8.
Brazil – Stochastic Production Frontier for CIT
The Brazilian Tax Administration had some difficulties to estimate CIT tax gap of small companies (Simples Nacional) using traditional methods, due to the lack of reliable and detailed data originated from the simplified tax forms. However, there was some third-party information available to be used in the estimation for small companies. This scenario contributed to choosing an alternative method as a CIT tax gap estimation tool, known as Stochastic Production Frontier (SPF).
The SPF method was originally developed to estimate the production possibilities of a set of companies based on a set of inputs such as capital and labour, through production functions. Thus, a classical production frontier model establishes the theoretical limits of the production capacity of firms using such inputs. The customisation of the SPF model to a tax approach in Simples Nacional was based on the use of tax information such as: purchases obtained from electronic invoices, bank flows and remuneration paid as inputs of a production function whose product was the revenue from the company. Thus, the econometric model was adapted to estimate a frontier for revenue generation (close to which the most compliant firms would be situated) and, consequently, to estimate the degree of noncompliance for firms, as a function of the gap between their declared revenue and the corresponding boundary. The use of discriminant variables with the model also allowed to obtain a good level of details in terms of geographic and sectorial cutouts.
After the Brazilian VAT tax reform, the model is planned to be updated to allow estimation of the VAT tax gap, through the prediction of the value added by companies (revenues minus inputs).
Canada – CIT gap estimation for large corporations using risk-based audits
The Canada Revenue Agency (CRA) continually monitors large corporations and risk assesses 100% of the corporations that are determined to be at a higher risk of non-compliance. These corporations are subject to rigorous compliance audits where the CRA examines relevant books and records to ensure that all tax obligations have been met.
While risk-based audits allow the CRA to focus its efforts on higher-risk taxpayers, non-compliance identified through these audits cannot be directly extrapolated to the population given that audits are selected based on the risk of non-compliance. Therefore, the CRA uses two statistical methods to minimise this selection bias and estimate the federal CIT reporting gap for large corporations:
Extreme value methodology, a statistical approach that assumes the majority of tax non-compliance in the large corporate population is concentrated in a relatively small number of corporations. It also assumes that the magnitude of non-compliance will tend to drop off exponentially when ranking corporations according to their level of non-compliance. A regression analysis is then used to extrapolate tax non-compliance to the rest of the large corporate population in order to obtain an estimate of the tax gap. One key limitation of this method is that it can underestimate the tax gap. Therefore, the CIT reporting gap from the extreme value methodology is used as a lower-bound estimate.
Cluster analysis, an unsupervised machine learning technique in the field of artificial intelligence that helps identity subgroups or "clusters" in a population, where objects in the same cluster are more similar to each other than to those in other clusters. In the context of tax gap analysis, clustering techniques were used to determine whether large corporations could be organized into relatively distinct groups on the basis of certain key variables to estimate the potential level of non-compliance within each cluster. In contrast to the extreme value method, cluster analysis can overestimate the tax gap. Therefore, the CIT reporting gap from cluster analysis is used as an upper-bound estimate.
Sources: Brazil (2024) and Canada (2024).
Value added tax gap
The VAT gap is typically related to tax non-compliance in VAT that can include various fraud activities (for example, carousel schema) and overclaiming VAT refunds in various sectors. This tax gap component is generally the first estimate that a jurisdiction examines due to its well-established top-down methodology.
Almost all jurisdictions with a tax gap programme estimate the VAT gap. Of those:
Around two thirds estimate the reporting gap and 59% estimate the payment gap. Also, the registration gap (estimated by 19%) and filing gap (estimated by 30%) are more common for this tax type than for others.
74% estimate the gross tax gap and 44% estimate the net tax gap.
89% use top-down approaches and 33% use bottom-up approaches. Some jurisdictions developed bottom-up methodologies in addition to their top-down approaches. Only a few jurisdictions use only bottom-up approaches to estimate their VAT gaps. The main data sources for bottom-up approaches are random and risk-based audit data (see Figure 11.9.).
For more information, the European Commission (European Commission, 2023[2]) and the IMF (Hutton, 2017[1]) published details on main methodologies used to estimate VAT gaps.
Number of jurisdictions that conduct the relevant analysis or apply the relevant methods
Note: 27 jurisdictions conduct VAT gap estimations.
Source: Annex Table 11.A.9 and Annex Table 11.A.10.
The Italian Revenue Agency has implemented two different bottom-up approaches for the estimation of the VAT gap overall and of the portion due to Missing Trader Intra Community (MTIC) fraud, by using data from risk-based audits.
VAT gap: the methodology combines traditional parametric inference methods, modern machine learning techniques and nearest neighbour imputation procedures. To address the selection bias due to the non-random selection of audited taxpayers, while preserving the distribution of data, the model relies on the conditional independence assumption building up a three steps procedure. Firstly, the Italian Revenue Agency estimates the selection probabilities on a target population through a logistic model and the units are then grouped into classes of “approximately constant selection probability”. The second step includes prediction of individual VAT gap values by bagging of regression trees, within each stratum. The third step applies the nearest neighbour imputation method based on predictive means to match non-audited taxpayers with audited taxpayers.
VAT gap due to MTIC fraud: The main challenges faced in the implementation of the approach is the possible double counting related to the estimation of the gap for all actors involved in the fraud mechanism. To address this issue, we focus on Missing Trader (MT) as a main actor of the fraud. For the identification of the MT we adopt the risk criteria suggested by an internal survey. The model is based on a two-step procedure. The first step involves the estimation, through a logistic model, of the probability of being a MT. The second step computes the MTIC fraud gap multiplying the (estimated) probability of being a MT and the evaded tax.
Source: Italy (2024).
Excise gap
Excise gap typically includes tax and duty non-compliance related to excise products such as cigarettes/tobacco, alcohol, fuel, spirits, betting and gambling. This component is rare and is estimated by 29% of jurisdictions. Reporting and payment gaps are common, and are each estimated by 63% of jurisdictions that measure the excise gap. Excise gap is typically reported as a gross tax gap (by 63%), and 38% of jurisdictions estimate the net tax gap. Bottom-up and top-down approaches are equally common for excise gap estimation, depending on what data is available in the jurisdiction (see Figure 11.10.).
Number of jurisdictions that conduct the relevant analysis or apply the relevant methods
Note: 8 jurisdictions conduct excise gap estimations.
Source: Annex Table 11.A.11 and Annex Table 11.A.12.
Sweden currently imposes around 15 different excise taxes, with the most significant in terms of revenue being energy taxes, carbon tax and taxes on alcohol and tobacco. The Swedish Tax Agency (STA) has conducted tax gap assessments specifically for alcohol and tobacco, congestion charges, and some energy taxes related to a tax rebate scheme. Due to the varied designs and target populations of these excise taxes, there is no general approach for assessing the overall tax gap in this area, especially since certain errors in excise taxes that give rise to tax gaps are near impossible to capture. Therefore, each excise tax must be evaluated individually.
The tax gap for alcohol and tobacco mainly arises from illegal production and imports, which are difficult to detect using internal data. The STA therefore relies on an external survey for its tax gap assessment, conducted by another government agency that specialises in alcohol and tobacco use in the Swedish population. Congestion charges are another particular area. It is assessed by identifying registration plates that have been intentionally or unintentionally concealed from traffic cameras. Energy taxes, on the other hand, are assessed using traditional random audits.
Source: Sweden (2024).
Other tax gap components
Other tax gaps generally contain non-compliance related to other types of taxes that exist in the jurisdiction’s tax system that may differ from others. Some examples include payroll, regional taxes, social security contributions, and inheritance taxes. Slightly more than one third of jurisdictions (36%) estimate other tax gap components. Jurisdictions usually estimate both payment gap (70%) and reporting gap (90%). Around 70% of jurisdictions estimate the gross tax gap and half estimate the net tax gap. Bottom-up approaches are more common (used by 70%) than top-down methodologies (used by 40%) mostly due to availability of random audits (see Figure 11.11.).
Number of jurisdictions that conduct the relevant analysis or apply the relevant methods
Note: 10 jurisdictions conduct other tax gap estimations.
Source: Annex Table 11.A.13 and Annex Table 11.A.14.
The tax gap for undeclared wages is estimated by a bottom-up approach using mainly tax return data. In Latvia, all employers are requested to submit monthly tax returns on wages that includes following data: detailed data on employees, hours worked during taxation period (month), renumeration and taxes withheld (SSC and PIT). A profession of an employee is also reported to the tax authority when employment relations are established.
Given that the wage depends on the following main parameters: 1) the employment industry; 2) territorial location of an employer; 3) profession of an employee and length of a working day, the Latvian Tax Administration’s approach is based on comparison of an actual wages with an average wage for each profession type in particular industry and region (further – “average wage”). For profession types recognised to have extra high risks of undeclared wages the “average wage” is set based on research of publicly available job advertisements.
As employee’s decision on choice of a particular job can be affected by non-fiscal factors (for example, location or a social “package” provided by employer), for the purpose of tax gap estimates, an assumption is made: an employee has received undeclared wages if their actual wage is less than 70% of an “average wage”. The difference between an “average wage” and actual wage is used to estimate the amount of an undeclared wage for each employee and employer. After an undeclared wage is estimated the effective SSC rate is applied for calculation of the SSC gap (same procedure is consequently applied for PIT gap estimation).
The main advantage of this method is possibility to personalise the tax gap for an employee and an employer and to analyse the phenomenon from different perspectives, for example, to identify risky industries, to define a demographic profile of an employee receiving undeclared wages.
Source: Latvia (2024).
Jurisdictions have various reasons to estimate tax gaps including supporting data-driven decision making. The main applications mentioned by jurisdictions are as follows:
Satisfying legislative requirements
Providing transparency to the public and parliament
Monitoring emerging trends and checking the health of the tax system
Identifying areas in the tax system and administration that may need improvements
Informing compliance areas on risks of non-compliance and the underlying behavioural drivers
Driving additional compliance research
Providing information to policy makers and enabling data-driven decisions
Facilitating organisational investments and planning
Measuring long-term performance of a tax administration or alongside other indicators
Although tax gap estimates can provide a lot of useful insights for tax administrations, they might not be a good basis for explicit performance targets due to a number of limitations (OECD, 2017[3]). Nevertheless, some jurisdictions do use tax gap estimates to understand overall performance of the tax system.
Tax gaps can be used not only by tax administrations but by other parties as well such as the public, politicians, other government departments, academics and other organisations. Their tax gap use could vary from the use of the tax gap estimates by the tax administration. For example, in Sweden, beyond the tax administration use, tax gap estimates are used to raise the quality of national accounts statistics.
The Australian Taxation Office (ATO) recognises that estimates of tax gap alone are not always a good measure of specific agency performance given that the tax gap measures whole of system performance, which is also impacted by factors outside of the agency’s control.
The ATO recognises these limitations in using tax gap as a performance measure, and particularly that it is not the best indicator of short-term performance. It is also hard to set targets without considering the historical context of tax gap. For this reason, Australia uses longer-term tax gap trend analysis to assess agency performance and finds that it is a useful indicator of the medium-term performance, particularly in the context of ATO’s stewardship role of the tax and super systems.
Sometimes tax gaps will go up or down through no fault of, or unrelated to the actions of the agency. For this reason, the ATO’s performance assessment includes meaningful qualitative context so that the audience can clearly understand the extent to which the ATO has contributed to the improvement or sustainment of the tax gap over the medium term.
Changes of tax gap performance are indicators of the overall health of the system including the agency and so, in the ATO, tax gap estimates form part of the strategy development and resource planning decisions.
Source: Australia (2024).
The COVID-19 pandemic was an unprecedented event that affected the whole world. Many jurisdictions implemented different measures to contain the pandemic and to address the economic impact as a result of, for example, lockdowns and other precautionary measures taken.
The pandemic also had an impact on tax administrations as many governments took action to support individuals and businesses by extending tax payment terms or suspending the collection of outstanding tax debt. (CIAT/IOTA/OECD, 2020[9])
Many jurisdictions have already noticed an impact of the pandemic on their tax gap estimates and some are already adjusting their methodologies accordingly. However, the full impact on the tax gap will need to be examined in the future.
The economic impacts of the interactions between the COVID-19 pandemic and economic stimulus measures evolved over time and lasted over a number of years. These economic disruptions would have had a direct impact on taxpayers’ financial situations and compliance behaviours, and hence tax collections.
For tax gap measurement, the COVID-19 pandemic presented both challenges and opportunities. The biggest challenges to measuring tax gap related to understanding the COVID-19 impacts on data. The ATO noticed significant changes in national account statistics and other external data with large changes that were difficult to validate as no other information existed on similar-sized economic shocks. The COVID-19 pandemic saw consumption patterns change leading to a significant reduction in the GST / VAT tax gap.
The ATO also had to contend with challenges with internally-generated data. Gap estimates that rely on operational audit data were impacted as the number of audits across some market segments declined as the ATO shifted its focus to supporting taxpayers. While the ATO was still able to generate estimates, the reliability rating of some of these “provisional” estimates was reduced to reflect a smaller sample size used to generate the estimate. Australia also changed the method for the medium business tax gap estimate to a combination of a logistic and Poisson Pseudo Maximum Likelihood regression model. This change was necessary as the assumptions underpinning the existing Extreme Value Method were no longer valid due to a reduction in compliance cases and coverage.
In terms of opportunities, the COVID-19 pandemic provides the ATO with an opportunity to apply tax gap thinking to economic stimulus measures for generating “payment” gap estimates for payment programs designed to support businesses and individuals.
Source: Australia (2024).
Tax gap estimation is not an easy exercise and it usually requires a lot of experience and diverse expertise. For further information on tax administration’s work on tax gap estimation see Table 11.6. which contains a selection of links to a number of tax gap reports.
References
[5] Barra, P., E. Hutton and P. Prokof’yeva (2023), Corporate Income Tax Gap Estimation by using Bottom-Up Techniques in Selected Countries: Revenue Administration Gap Analysis Program, International Monetary Fund, Washington, DC, https://doi.org/10.5089/9798400246265.005.
[9] CIAT/IOTA/OECD (2020), “Tax administration responses to COVID-19: Measures taken to support taxpayers”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/adc84188-en.
[10] Council of the European Union (2006), “Council Directive 2006/112/EC of 28 November 2006 on the common system of value added tax”, Official Journal of the European Union OJ L 347, https://data.europa.eu/eli/dir/2006/112/2024-01-01.
[2] European Commission (2023), VAT gap in the EU – 2023 report, Publications Office of the European Union, https://doi.org/10.2778/911698.
[6] Fiscalis TGPG (2018), The Concept of Tax Gaps Report II: Corporate Income Tax Gap Estimation Methodologies by FISCALIS Tax Gap Project Group (FPG/041), European Commission, Brussels, https://taxation-customs.ec.europa.eu/system/files/2018-07/tgpg-report-on-cit-gap-methodology_en.pdf (accessed on 10 September 2024).
[7] HMRC (2020), Non-detection multipliers for measuring tax gaps, HMRC, http://www.gov.uk/government/publications/non-detection-multipliers-for-measuring-tax-gaps/non-detection-multipliers-for-measuring-tax-gaps (accessed on 10 September 2024).
[1] Hutton, E. (2017), The Revenue Administration-Gap Analysis Program: Model and Methodology for Value-Added Tax Gap Estimation, International Monetary Fund, Washington, DC, https://doi.org/10.5089/9781475583618.005.
[3] OECD (2017), “The measurement of tax gaps”, in Tax Administration 2017: Comparative Information on OECD and Other Advanced and Emerging Economies, OECD Publishing, Paris, https://doi.org/10.1787/tax_admin-2017-19-en.
[8] Swedish Tax Agency (2023), Using the Delphi method to determine a non-detection multiplier for the tax gap assessment, Swedish Tax Agency, Malmö, https://skatteverket.se/download/18.7da1d2e118be03f8e4f45d8/1703064480910/Using%20the%20Delphi%20method%20to%20determine%20a%20non-detection%20multiplier%20for%20the%20tax%20gap%20assessment.pdf (accessed on 10 September 2024).
[4] Thackray, M., S. Jennings and M. Knudsen (2021), The Revenue Administration Gap Analysis Program: An Analytical Framework for Personal Income Tax Gap Estimation, International Monetary Fund, Washington, DC, https://doi.org/10.5089/9781513577173.005.
Annex 11.A. contains a set of fourteen tables that hold the data provided by members of the OECD Forum on Tax Administration’s Community of Interest on Tax Gap, an informal network of tax gap analysts, in response to a survey on tax gap estimations conducted in 2023:
Annex Table 11.A.3. Tax gap approaches, data and methodological adjustments
Annex Table 11.A.4. Number of bottom-up methodologies by data sources
Annex Table 11.A.5. PIT gap measurement: Type of gap estimation conducted
Annex Table 11.A.6. PIT gap measurement: Approaches and methods used
Annex Table 11.A.7. CIT gap measurement: Type of gap estimation conducted
Annex Table 11.A.8. CIT gap measurement: Approaches and methods used
Annex Table 11.A.9. VAT gap measurement: Type of gap estimation conducted
Annex Table 11.A.10. VAT gap measurement: Approaches and methods used
Annex Table 11.A.11. Excise gap measurement: Type of gap estimation conducted
Annex Table 11.A.12. Excise gap measurement: Approaches and methods used
Annex Table 11.A.13. Other gap measurement: Type of gap estimation conducted
Annex Table 11.A.14. Other gap measurement: Approaches and methods used
Source: FTA 2023 survey on tax gap estimations.
Source: FTA 2023 survey on tax gap estimations.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that use bottom-up approaches.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct PIT gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct PIT gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct CIT gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct CIT gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct VAT gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct VAT gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct Excise gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct Excise gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct other gap measurement.
Source: FTA 2023 survey on tax gap estimations.
Note: The table only contains data from jurisdictions that conduct other gap measurement.
Source: FTA 2023 survey on tax gap estimations.
← 1. All data in this chapter is based on the 2023 survey responses (updated as of July 2024) from the members of the OECD Forum on Tax Administration’s Community of Interest (COI) on Tax Gap, an informal network of tax gap analysts. All percentages are calculated based on the data from the 28 jurisdictions that replied to the survey and estimate a tax gap. Jurisdictions with tax gap programmes that did not reply to the survey are out of scope for this chapter.
The data in this chapter may differ from the International Survey on Revenue Administration (ISORA) results since the two surveys were conducted during different timeframes and the questions were worded slightly differently. For example, the COI survey asked whether a tax administration estimates the tax gap for the different tax types, while ISORA asked whether the tax administration or any other government agency produces periodic estimates of the tax gap for the different tax types.
← 2. Although the Netherlands Tax Administration estimates the total gap for each entity, it is still possible to establish the estimation of the tax gap of a specific tax revenue based on these random audits. However, it is a strategic decision to focus on the total tax gap of each entity.