JANUARY 06, 2025

By Neeraj Prasad, IRS (C&IT)

IMPROVING tax compliance and reducing the tax gap is typically one of the main objectives of a tax administration. 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. Although a tax gap may have a relatively simple definition, its estimation is complex and includes many nuances. 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.

2. Tax gap definition

Tax gap is generally defined as a difference between the potential tax revenue and actual tax revenue. 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. 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.

A small number of jurisdictions publish their overall tax gap estimates in the public domain 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 Several jurisdictions do not publish their overall tax gap, but they make one tax gap component usually, Value Added Tax publicly available.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 an irregular schedule.

Policy Gap Example: European Commission

The European Commission's annual study on the VAT gap in the EU (European Commission, 2023) 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.

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.

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.

3. Tax gap types

A tax gap can be divided based on non-compliance sources such as registration, filing, reporting or payment. 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 and payment non-compliance as they are major sources contributing to the tax gap. 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.

Non-compliance sources that contribute to the tax gap:

Registration non-compliance: When taxpayers who are required to register with the tax administration do not do so.

Filing non-compliance: When taxpayers who are required to file a tax return with the tax administration do not do so.

Reporting non-compliance: When taxpayers fail to provide complete or accurate information on their tax returns by under-reporting income or claiming deductions or credits to which they are not entitled.

Payment non-compliance: When taxpayers do not pay taxes by the payment deadline.

United States - Tax gap by source of non-compliance

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.

4. Key factors of a successful tax gap program

Data: Investing in data management and data cleaning. Ensuring data covers a target population. Accumulation of longitudinal data/analysis. Periodical revision to improve estimates' reliability.

Methodology: Selecting the most appropriate methodology tailored to the data available. Establishing clear objectives. Using both bottom-up and top- down techniques. Consistent estimation techniques over time.

Human resources: Building a multidisciplinary team with diverse expertise (i.e. in data analytics, statistics, econometrics, audit, tax policy). Avoiding frequent turnovers and keep the team stable.

Management support: Senior management endorsement. Patience in team development.

Institution orientation: Public mandate to estimate tax gap. Regular publication. Adopting transparency.

5. Tax gap methodologies

Tax gap estimation is complex and requires nuanced analysis. In general, there are two main approaches to estimating the tax gap:

- 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.

- Boflom-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.

Two main approaches to estimate a tax gap

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.

Boflom-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).

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 and 2023.

Random audit programmes are considered a high-quality method to estimate tax gaps in large populations of registered taxpayers (OECD, 2017). 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.

Random audit programmes

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. Improving compliance behaviour depends on 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.

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), 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.

On average, there are more bottom-up methodologies from random audits than from risk-based audits.

6. Non-detection multiplier

Some jurisdictions are applying a non-detection multiplier to their tax gap estimates. 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.

There are several methods to develop a non-detection multiplier as can be seen in Table. 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; Swedish Tax Agency, 2023).

Methods to estimate non-detection multiplier

Detection controlled estimation (DCE): Econometric approach based on a separate study: brings all audit cases to the same level as they were examined by the "best" examiner. Developed by researchers in late 80s for the United States. First used by the United States for the Tax Year 2001 tax gap estimates.

Secondary review by expert auditors: Audits are passed to a separate group of experts to review and estimate a non-detected tax value. Requires additional resources and is limited to the ability of experts to find non-detected taxes.

Third-party data matching: Audits are conducted without third-party information and the results are then compared to available third-party data.

Expert judgement (Delphi technique): Panel of experts estimate how much tax generally does undetected in different types of audits at an aggregated level. Requires less resources but is limited to experts' ability to estimate non- detection amount in groups of taxpayers. Is used by the United Kingdom.

Adopting others' multipliers: Adopting multipliers calculated by other tax administrations or other experts or using international ranges. May not represent a country- specific tax system and audit process. May be a good alternative in an absence of other methodologies.

Non-detection multiplier or uplift factor

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.

7. Tax gap components

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.

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.

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.

The European Commission (European Commission, 2023) and the IMF (2017) published details on main methodologies used to estimate VAT gaps.

Italy - VAT frauds and tax gap estimation

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, 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.

8. Tax gap use

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

9. Challenges related to tax gap estimation

Main challenges for tax gap estimation are usually related to data, methodology, resources and legislation. 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.

Key challenges for tax gap estimation

Tax gap estimation is not an easy exercise and it usually requires a lot of experience and diverse expertise.

Data: No random audits, small or not representative samples, data availability lags in data matching various data sources, non-detection, heterogeneous population, need of multiple data sources.

Methodology: Extrapolating from risk-based audits, difficulties in modelling complex non-compliance schema, finding appropriate methods for given data, accounting for emerging trends, limitations of top-down methods. need of multiple approaches.

Resources: limits in audit capacity, lack of tax gap experts, lack of budget, time- consuming audits, lack of internal support.

Legislation: Frequent changes in tax laws, complex tax systems.

Notwithstanding the challenges its time that a Tax Gap Analysis is publicly available for an informed engagement by all stake holder.

Reference: Tax Administration Report (2024), OECD.

[The views expressed are strictly personal.]