Tax Gap in Canada: A Conceptual Study

Chapter 4

How Can the Tax Gap Be Used in Tax Administration?

As outlined above, a tax gap estimate should ideally be useful both for a tax administration to improve compliance and for taxpayers to understand what the administrator is doing and how well. Estimating the tax gap may also provide insight into potential weaknesses in existing tax policy and legislative frameworks which can provide the basis for discussions aimed at improving compliance between the tax administration and the relevant legislative body.

Tax gap estimation has benefits and shortcomings that are outlined below.

Four people sitting at a conference table viewing a screen

4.1 Usefulness for Tax Administration Purposes

Tax gap estimation is not a panacea for combatting non-compliance. However, it can play a role in helping tax administrations improve their understanding of the health of the tax system, as well as, providing guidance on which areas may benefit from increased attention.

a. Insight into the Overall Health of the Tax System

Tax gap estimates can provide a general sense of a tax administration's performance at promoting and enforcing compliance. Provided that an estimate is accompanied by a detailed description of the methodology used to arrive at the estimate, publication of a tax gap estimate can improve the openness and transparency of the tax administration.

Where the tax gap estimates are available over a period of time based on the same methodology, they can provide information on whether non-compliance is increasing or decreasing, and highlight the state of voluntary compliance with the tax system. This can provide a degree of insight into the effectiveness of tax laws, tax administration services, and enforcement activities.

However, as a rule, tax gap estimates should be analyzed in conjunction with other information and intelligence on non-compliance to provide a more complete view of the health of the tax system. In general, tax gap estimates, on their own, are poor indicators of a tax administration's short-term performance because the size of the tax gap is not only determined by taxpayer compliance but also by factors beyond the tax administration's control (for example changes to tax policy, economic cycles, and other factors).

Bottom-up estimates, in particular, are problematic from a performance monitoring perspective. This is because bottom-up estimates only capture known and identifiable sources of risk and cannot take into account those that are unknown (for example a new tax evasion scheme that the tax administrator is not yet aware is in use). Therefore, a small tax gap arrived at using bottom-up methodology may not mean there is truly a small tax gap. In addition, a bottom-up tax gap estimate may increase by virtue of having discovered a new form or area of non-compliance; in such a case, the tax gap would not have actually increased but a previously unknown area of non-compliance would simply now be identifiable and quantifiable.  

Interestingly, the very act of publishing a tax gap estimate may have an impact on compliance behaviour. Some commentators suggest that in circumstances where a tax gap is estimated to be low, indicating that most taxpayers comply with the tax rules, other taxpayers may be encouraged also to comply. However, the opposite may also be true. That is to say, a large estimated tax gap could potentially also discourage compliance among taxpayers if they perceive this indicator to imply a tendency towards non-compliance in the system in general.

The GST/HST tax gap estimate released today illustrates some of these issues. The estimate suggests an overall GST/HST gap of 5.6 percent of total theoretical GST/HST liability. Further, the estimate demonstrates that tax gap estimates are impacted by factors unrelated to non-compliance: the fluctuations in the GST/HST gap over the period covered by the study can largely be explained by the timing of when input tax credits are claimed, which may be influenced by changes to GST/HST rate or by provinces joining the HST framework.

Other compliance measurement tools also provide insight into the success of a tax administration in improving compliance. For example, the OECD's Forum on Tax Administration in its 2014 report "Measures of Tax Compliance Outcomes" discusses the concept of "tax assured" which is a measure of the proportion of the revenue base where there is confidence that the taxpayer is complying with all tax obligations, and in particular that information reported on tax returns is reliable. According to the OECD, "measuring where the tax system is working well provides the right incentives to ensure that the tax administration system is designed to get the right revenue outcomes the first time."

b. Understanding the Composition and Scale of Non-Compliance

Tax gap estimates, used in conjunction with other assessment tools, can also enhance the value of the intelligence held by a tax administrator on the sources of non-compliance within the tax system. By attempting to quantify non-compliance, tax gap estimates can provide a mechanism for comparing compliance among different aspects of the tax system which could help the administration understand the areas of greatest risk within the system.

While tax gap estimates are not particularly well suited to pinpointing specific cases of non-compliance (i.e. estimates cannot identify a particular taxpayer who may be engaging in tax evasion), they can act as a guide for tax administrators in a variety of more general ways. In particular, information gained through tax gap estimation:

  • may be used to refine risk assessment tools to guide tax administrators towards certain aspects of the tax system for further scrutiny;
  • may help to inform management decisions and operations strategies including decisions relating to resource allocation; and,
  • may contribute to identifying areas where new rules are needed to prevent non-compliance, and/or allow the tax administrator and policymakers to respond to non-compliance more effectively.

4.2 Limitations of Tax Gap Estimation

While tax gap estimates provide some insight into non-compliance within the tax system, they also raise a number of concerns about their precision, timeliness, and cost, as well as the impact that data collection may have on taxpayers. These issues are discussed in more detail below.

a. Precision and Margin of Error

It is widely acknowledged by tax administrators and policy officials, including those who estimate the tax gap, as well as commentators that tax gap estimation is imprecise. Further, the degree of imprecision may often be difficult to quantify.
Two key sources of error arise in the tax gap estimation process:

  • Sampling errors arise because data is drawn and characteristics are estimated from a smaller population, or sample, rather than the whole population. For example, in the context of tax gap estimation, data from random audits (described in more detail later in this paper) are often used to estimate the tax gap among a given population. Similar to survey research, audit results of a sample used for tax gap estimation may not be fully representative of the issues that would be apparent if audits were performed on the whole taxpayer population. 
  • Systematic errors arise from errors in the assumptions used in calculating the estimate or other errors related to the data used resulting in an estimate that is either too low or too high. For example, top-down methodologies rely on the assumption that national accounts data covers all economic activities. However, if certain concealed economic activities are not captured by the national accounts, the revenue loss caused by those activities would not be covered by the tax gap estimate.

The resulting variations in tax gap estimates can be large. Accordingly, tax gap estimates, and the changes to those estimates over time, must be interpreted recognizing their imprecision. Changes to the estimates can occur not only as a result in changes to the level of non-compliance, but also due to improvements to or deterioration of the data available, the methodology used, and a host of other factors.

b. Timeliness

Tax gap estimates are prepared using historical data. The time lag between the period to which the data refers (the study period) and the development of a tax gap estimate varies depending on the source and quantity of the data used to estimate the tax gap and the related collection and processing time for that data.

  • For example, an estimate of the personal income tax gap for a particular tax year would be based on information from tax returns filed in the following year, assessment of those returns in the year or years following, and for a robust estimate, audit and collections data related to that tax year. This information would only be available after a number of years had passed.

A comprehensive tax gap estimate may have a significant time lag from the study period to publication. For example, tax gap estimates for the United States published by the IRS in 2012, were for the 2006 tax year; estimates published in early 2016 were for the 2008 to 2010 tax years.

These time lags have an impact on how tax gap estimates should be interpreted and can be used. Tax laws may have changed and a tax administrator's techniques for improving compliance and combatting non-compliance may have evolved to address certain types of non-compliance during the time between the study period and publication. Consequently, tax gap estimates generally do not reflect the current state of compliance, health of the tax system, or impact of the tax administration on compliance, but rather lag behind it.

c. Random Audit Programs

The cost of tax gap estimation is determined by the comprehensiveness of the program and the techniques used to collect and analyze the data and to develop estimates. Tax gap estimates are particularly costly if they require a random audit program to gather data.

It is important to note that random audit programs are not only used to collect data for tax gap estimation purposes. They also help achieve a variety of other goals including evaluating the validity of existing risk assessment models and identifying new compliance issues for investigation in the future.

In the context of tax gap estimation, random audit programs are generally used to gather data to estimate tax gaps for direct taxes. Random audit programs generally involve selecting a random sample of taxfilers, intended to be representative of the entire tax-filing population, and conducting full compliance audits on the returns filed by those individuals or businesses. The main goal of a random audit program is to identify the extent of non-compliance in the sample with the aim of extrapolating that amount to estimate non-compliance in the whole population. It is important to note that random audits will not identify all instances of non-compliance, which may lead to the underestimation of the tax gap based on this approach alone.

Random audit programs subject taxfilers, many of whom may be fully compliant, to scrutiny solely for the purposes of data collection. Despite the fact that random audits can yield valuable information from an intelligence perspective, they also impose a significant burden on these taxfilers, and potentially a significant cost as well.

For these reasons, random audits are generally only carried out to the extent required to effectively calibrate risk assessment systems (for example to improve risk based audit programs) or inform analysis of particular sectors in the economy for potential non-audit (preventative) interventions.

A number of countries around the world use random audit programs of varying sizes. As noted above, both the United States and the United Kingdom, for example, use random audit programs. The IRS undertakes approximately 14,000 random audits (out of approximately 138 million individual returns) to help inform its tax gap estimate, while in 2012-13, HMRC conducted approximately 2,600 random audits for individual taxpayers (out of approximately 30 million individual taxpayers). However, when employers and corporations are considered, more than 4,000 random audits were undertaken by HMRC.

As is the case with public opinion surveys, the reliability of the estimate derived from a sample of random audits increases with the number of audits conducted. Consequently, a relatively low number of random audits leads to estimates with a low level of reliability.

Photo of paper dolls with paper dolls highlighted at random

Random Audits at the CRA

Random audit programs have been used by the CRA for a number of years beginning with the Core Audit Program (CAP) that was first implemented in 1999. The CAP was a random audit program that alternated annually between Small and Medium Enterprise (SME) audit populations and consisted of 1,500 to 1,700 audits per year. This relatively small number of randomly selected audits only allowed the CRA to produce statistically reliable national and regional non-compliance estimates.

More recently, the CAP was replaced by the SME Research Audit Program, a random audit program designed to enhance the Agency's understanding of the reporting non-compliance of the different SME audit populations (self-employed individuals and corporations) for income tax and GST.

About 4,700 SMEs were selected for full-scope audits with the objective of measuring the non-compliance rates by industry as a basis for monitoring compliance trends over time, and to validate and refine the CRA's risk assessment systems in order to improve file selection and target audit resources more effectively.

Research audits for the unincorporated small business segment of the program have been completed; the results indicate that almost half of the target population made no reporting errors and for about 78 percent of the files that were adjusted, the errors were minor. Results for the incorporated business segment are expected in the summer of 2017, and GST results are expected later.

Report a problem or mistake on this page
Please select all that apply:

Thank you for your help!

You will not receive a reply. For enquiries, contact us.

Date modified: