Introduction and Background
Every person liable to taxation under this Act shall, on or before the twenty-eighth day of February in each year, without any notice or demand, deliver to the Minister a return, in such form as the Minister may prescribe, of his total income during the last preceding calendar year. (
Canada 1917, § 7.[1])
Section 7.(1) of the
Income War Tax Act (
Canada 1917) established Canada's first personal income tax and enshrined self-assessment as the method by which taxpayers would inform the government of their tax liability.
1 At the time, all personal income taxes were due as lump-sum amounts payable with an annual information return.
From the outset, the burden of having citizens file tax returns was noted, even though fewer than 10 percent of Canadians completed tax returns (
di Matteo 2017). Over time, the number of Canadians who were required to pay taxes grew, as did the complexity of the tax code, increasing the burden placed on tax filers or leading them to enlist professional help.
Estimates of the costs of compliance in Canada date back to
Vaillancourt (1989), who included detailed estimates of the time spent in the various activities involved in compiling a return. Later work by Vaillancourt updated and elaborated on his earlier work (Grine & Vaillancourt 2023;
Vaillancourt et al. 2013,
2015). Until the vast improvement in digitization in recent years, efforts to reduce compliance costs had centred on simplification of the tax code. Given significant advances in digitization and information sharing for tax administration,
2 our primary interest in this study is in the potential to make use of existing information flows to reduce the administrative burden and out-of-pocket costs to taxpayers with simple tax returns and limited ability to pay for professional tax products or services. We do not assume any changes to current tax measures.
Another important feature of Canada's tax policy is that several important social benefits are available only to eligible Canadians who file a tax return. Those benefits are either administered directly by CRA or by other government agencies using information provided by the CRA. This means that not filing a return can cost non-filers a significant amount in foregone income-tested transfers, public subsidies, or access to public services.
3 Our primary interest in this study is the capacity of the CRA to derive accurate estimates of income for the purposes of program and cash transfer administration, separate from the function of tax collection. In fact, many Canadians with low income who rely most on income-tested benefits may not owe any income taxes at all.
Having the Tax Agency Complete Tax Returns
The prospect of having a tax administration agency initiate and complete tax returns has been studied in the past. In referring to this idea, we use the term
tax agency reconciliation (TAR).
4 Not only would TAR reduce the compliance costs borne by taxpayers, but it would also increase the likelihood that the social benefits delivered through the tax system reach eligible Canadians who currently do not file a return (
Robson & Schwartz 2020). Where returns could be completed under a TAR system using existing sources of information available to the tax agency, the administrative burden would also not increase for the tax agency or third parties.
In the United States, starting in the late 1980s and extending through the 2000s, the possibility of “return-free” filing led the US federal government to commission several reports on its feasibility (
US Department of the Treasury 2003;
US Government Accountability Office 1996). Academic studies also appeared (
Gale 2001;
Gale & Holtzblatt 1997;
Goolsbee 2006). Indeed, one US state, California, introduced return-free filing for an invited sample whose state tax returns were thought to be simple (
Bankman 2005,
2008), as did Quebec (
Vaillancourt et al. 2011, 63–72). The California and Quebec programs were maintained for only a few years, despite positive reviews from their users. Other US states experimented with various forms of return-free filing (
Holtzblatt 2007, 334–346).
Laurin and Dahir
Laurin and Dahir (2022) begin, as do we, with the presumption that the goal of TAR is to ensure that as many eligible Canadians as possible receive the social benefits that are available only to those who file a tax return. That presumption is supported by the 2020 Speech from the Throne that “committed the federal government to ‘work to introduce free automatic tax filing for simple returns to ensure citizens receive the benefits they need’” (Payette 2020, as cited in
Laurin and Dahir 2022, 1).
Laurin and Dahir (2022) estimate the proportion of tax returns that are basic—receiving income only from sources that report directly to the CRA, claiming only those federal deductions for which the required information is already available to the CRA from third parties and only those tax credits whose eligibility requirements are knowable to the CRA. They estimate that 32 percent of all filers have basic tax returns.
5 without fundamental tax policy reforms aimed at tax simplification, a broad-based automatic tax-filing system is unlikely to reduce compliance costs compared to the current auto-fill system: the majority of preliminary tax assessments prepared by the tax agency still would require further modification and optimization by tax filers. (13)
Goodman et al.
Goodman et al. (2023) use two methods of estimating whether the IRS could successfully calculate the tax liability for each tax filer. In their item-based approach, Goodman et al. define success as the absence of any features of the tax return that would cause the IRS to miscalculate tax liability. They identify 22 failure situations, including, for example, the tax filer claiming a tax credit or a tax deduction that could not be known to the IRS on the basis of the information available to it. This item-based approach is similar to that which we use in our own analysis of the Longitudinal Administrative Database.
Because they have the information returns,
Goodman et al. (2023) can also simulate the tax liability of each taxpayer and compare it with the actual tax liability reported by the taxpayer. That is, they can replicate the outcome of a TAR process, simulating what the IRS might do if TAR were instituted for all US taxpayers. By doing so, they are “the first to simulate pre-populated returns in the U.S. context” (806).
6 Success using this tax liability approach is defined as a simulated tax liability that is within a predefined dollar amount of the tax liability actually reported by the taxpayer.
Using the tax liability and the item-based approaches,
Goodman et al. (2023) estimate that 42 percent and 48 percent, respectively, of US tax returns could be successfully pre-populated and processed by the tax agency. They also note that the success rate declines with income, that most of those for whom pre-population is unsuccessful have only one failure situation, and that success is more common among taxpayers who are single and young and have no dependents. Among those for whom pre-population is a success, “43 to 44 percent used a paid preparer when filing” (
Goodman et al. 2023, 806).
Barriers to Successful TAR
The extant literature highlights several key barriers to a successful transition from a system of self-assessment to TAR. Here, we briefly note these and return to them later, in the “Conclusion” section, to discuss them in the current Canadian context.
Can TAR Be Offered to All Tax Filers?
Laurin and Dahir (2022) observe that TAR would work for only a minority of tax filers in Canada. They conclude, therefore, that “fully automatic tax filing for all appears to be a very blunt instrument to get to the narrower problem of non-filing eligible benefit recipients” (20). Instead, they argue for delivering the relevant benefits outside the tax system.
7 Goodman et al. (2023) estimate that pre-population would also be successful for just fewer than half of US tax filers but note that their “analyses can inform policymakers considering more limited approaches to implementing pre-population” (818).
Missing Personal Information
Accurate TAR of returns and ensuring that tax filers receive the pre-populated returns (to allow them to correct or dispute information if they so wish) requires that the tax agency have accurate information about the taxpayers’ family composition as well as accurate contact information. Laurin and Dahir (2022, 13) note that to complete even basic returns, the CRA would need accurate information on the age of taxpayers, their marital status, and the ages and number of their dependents. This may be more challenging in cases in which a person has never filed a return in the past.
Mismatches between Third-Party Information and Taxpayer-Reported Information
Table 2 in
Goodman et al. (2023, 815) lists the proportion of their sample with each of their 22 failure situations. Three of the five most common failure situations are mismatches between the information provided on the information returns and the information reported by the taxpayer. These mismatches involve (1) self-employment income, (2) wage income, and (3) family composition between 2018 and 2019.
Tax Filer Concerns
Tax filers who expect to receive tax refunds or to receive benefits delivered through the tax system might not be willing to wait for the pre-populated returns. In Canada, most third parties must currently provide information to the CRA and to tax filers by the end of February following the year to which the information applies. This timing issue was a central barrier cited in the US reports of the 1990s and 2000s when it seemed that pre-populated returns might not be available until after the April tax deadline.
Tax filers might also worry about the motivations of the tax administration. On the basis of the two previous government reports and their own survey of taxpayers, the
US Department of the Treasury (2003) lists a number of reasons why taxpayers might not favour pre-populated returns, beyond the desire to get refunds as soon as possible. Summarizing the concerns of their survey respondents, they write, “The primary barriers to a return-free system were concerns about giving the government too much control over taxpayers’ lives and questions regarding how problems would be resolved with the IRS” (37).
Resistance from the Tax Preparation Industry
There is little doubt that TAR threatens businesses involved in tax preparation, including firms that prepare tax returns and those that sell tax preparation software.
Ventry (2010) details the efforts of Intuit, the owner of TurboTax software, to undermine the California ReadyReturn system.
Elliot and Kiel (2019), writing for ProPublica, document the lengthy efforts of Intuit to forestall efforts to use technology across the United States, whether in the form of TAR or by creating its own free software. The tax preparation industry has itself identified government policy to provide free services or to automate filing as a risk to its business model (
Intuit 2023, 16).
Synthesis of the Literature and the Present Study
The main conclusion we draw from the past literature on TAR is that, even without any changes to the tax code, TAR should be possible for an important share of low- and modest-income tax filers. We begin our empirical work by discussing our analysis using de-identified tax return data in the Longitudinal Administrative Databank (LAD). We then briefly discuss supplemental findings from the CRA's published administrative data. We conclude with some observations and suggestions as governments in Canada proceed with commitments to simplify or automate tax filing for lower-income Canadians.
Data and Methods for the Longitudinal Administrative Databank Study
Our goal in this study is (1) to estimate the share of Canadian families with simple returns that could be initiated and completed by CRA and (2) to describe the sources of complexity that appear on the tax returns of the Canadian families whose returns could not currently be successfully completed by the CRA.
Data from the Longitudinal Administrative Databank
The analysis in this article is based on the LAD,
8 which is a 20 percent sample drawn from the Statistics Canada (STC) database known as the T1 Family File (T1FF). The T1FF is constructed by STC, primarily using three administrative databases provided by the CRA. STC combines (1) individual tax returns (i.e., T1 returns), (2) employers’ statements of remuneration paid (i.e., T4 slips), and (3) a file documenting the receipt of federal child benefits. Together, these files allow STC to impute the existence of family units, “matching by social insurance number, family name, and postal code while accounting for age, sex, and marital status” (
STC 2023b). Each imputed family unit consists of the individual tax filer; their spouse or partner, if any; and any unpartnered children of any age, all living in the same dwelling.
9 Unattached individuals with no dependent children are counted as a family of one person. The T4 records allow some working but non-filing spouses or partners of tax filers to be identified in the LAD using third-party records, and the federal child benefits file allows any children in the family to be identified if both parents have filed a return (which is a condition of benefit payment).
STC (2023b) writes that “when complete, the [T1FF] data is approximately 96 percent of the population.”
10 The information for each individual tax filer is then augmented by information about their family, if any, allowing STC to impute family-level variables, such as the receipt of various forms of income by family members and the use of various provisions of the tax system by family members.
The LAD has been created specifically for the purpose of enabling research that requires large volumes of administrative data (
STC 2021). The first LAD was created in the 1980s, based on tax returns for 1982. Once a person is selected for the LAD, the individual remains in the sample and is picked up each year from the T1FF as long as they appear in the T1 file in that year. Individuals selected for the LAD are linked across years by a unique LAD identification number, thus creating a longitudinal profile of each individual in the sample. It is possible to link about 95 percent of the individual tax filers in each year's LAD to their previous tax returns. The LAD is augmented each year with a sample of new tax filers so that it still consists of approximately 20 percent of tax filers every year. Reflecting increases in the Canadian population and changes in the incidence of tax filing, the 20 percent sample has grown over the years, starting with 3.2 million individuals and reaching 5.8 million in 2019. In our analysis, we use only tax information pertaining to the 2019 tax year and variables describing 2019 characteristics. This was the most recent year available at the time of writing.
11The information in the LAD is organized into four levels of aggregation: individual, couple, child, and family. Because many tax credits and tax deductions take the income or characteristics of an individual's partner (and, less frequently, other family members) into account, we believe that sources of complexity are best measured at the family level. Our analysis, therefore, uses only family-level data.
12 To avoid double counting the family information, we retain only one record per family. We also apply family weights provided by STC, which allow its 20 percent sample to be representative of the whole set of families found in the T1FF.
Measuring Simple versus Complex Returns
The LAD enables us to observe, for each imputed family, the types of income that the family received in 2019. From their T1 information, we can also observe the use of any of the various federal tax credits and tax deductions. Depending on the elements that are included in the family's returns, we can therefore estimate which returns could have been completed by the CRA using third-party information. The CRA receives information from employers on wages, taxable benefits, union dues, registered pension plan contributions, Canada or Quebec Pension Plan contributions, and Employment Insurance contributions. This means that the CRA, using only the T4 information slip provided by employers, could complete the return of an individual whose only source of income was wages and who did not claim any credit or deduction. We would classify such an individual as having a simple return.
However, the CRA also gets information about a wider range of income sources.
13 For example, it receives slips about public benefits from provincial and federal governments (T5007, T4E, and T4EQ slips). It also receives information from financial institutions that report savings and income in registered accounts, dividends, and interest income above $50 in any given year (T4A and T4RSP slips). The CRA does not have third-party information about whether individuals make charitable donations or whether they have moving expenses, which can be claimed as tax credits and deductions, respectively. We presume, however, that the CRA can automatically apply a series of characteristics-based credits for individuals. These include the basic amount, the spousal or equivalent-to-spouse amount; the Canada employment amount; and the amounts for age, dividends, pension income, employment insurance premiums paid, and provincial parental insurance plan premiums. These credits are all based on information the CRA can derive from information slips or the date of birth of the filer where a previously completed return exists in their records.
14We thus assume that an individual whose sources of income are all covered by these information slips and who claims only the simple credits listed has a simple return that could be completed by CRA. This is equivalent to the item-based approach used by
Goodman et al. (2023) to assess the potential for successful creation of pre-populated returns.
15Given that our unit of analysis is the imputed T1FF–LAD family, we assess simplicity at the family level. A family's tax situation is considered simple only if all members have simple returns (or no returns, in the case of children). Our estimates of the simplicity of families’ tax returns are thus conservative because one individual with a complex tax situation raises the tax return complexity for other family members who might, individually, have simple returns.
Because the LAD is based on information from actual tax records, it shows what families included in their tax returns, including elements for which the CRA does not get information slips. Using the information on this extensive set of possible credits, deductions, and sources of income, we can also study families with complex tax returns in depth. We call these credits, deductions, and sources of income for which the CRA does not have third-party information “sources of complexity” and divide them into three groups: (1) income sources, such as self-employment or capital gains; (2) deductions, such as moving or childcare expenses; and (3) credits, such as those for medical expenses or charitable donations.
We take each of our LAD records and calculate the number of sources of complexity that are present in the returns of that specific family, overall and in any one of the three groups of sources of complexity. This provides us with a measure of the depth of complexity by looking at the share of families that have different numbers of sources of complexity in their returns. We also list the sources of complexity per family, which gives us a measure of the concentration of complexity in any given source or group of sources. We can thus go beyond estimating only whether families have simple or complex returns; we can also find out how close to being simple most complex returns are and identify the main sources of that complexity.
The LAD also provides us with information about age, gender, income level, and main income sources of families. This allows us to study how tax return simplicity varies across different types of families. In our analysis, we define total income as the sum of all income, from all sources, before taxes and before refundable tax credits (such as the Canada Child Benefit).
Results of the Analysis of the Canada Revenue Agency Data
In
Table 4, we present four-year means for the 2015 to 2018 tax years. Percentages calculated from other lines in the table are in bold for ease of reference.
Using the CRA's administrative data, we find that two-thirds of all personal income tax returns filed by Canadians are simple according to the agency. Among the 10.5 million taxpayers with a total personal income below $25,000, the share of simple returns was even higher at 77.8 percent. Note that these results are for individual income rather than the family income we use in our main study and so are not directly comparable with our main results.
The CVITP program completes an important but relatively small number of returns each year (fewer than 800,000 per year, on average), representing just 2.9 percent of the 27 million annual tax returns received by the CRA. Among the 27 million personal tax returns filed, 15 million are EFILEd by a tax preparer outside the CVITP system. Although some of these might be prepared for free by friends and family who are tax professionals or through no-charge services provided by some commercial preparers (see
Appendix C in the online supplement), we expect that the majority will have been filed by a professional in exchange for a fee paid by the tax filer. Among these 15 million returns, we find that a majority, 9.6 million (or 59 percent), are categorized as simple by the tax agency. When looking only at the 17.8 million simple returns received by the agency, up to 49.4 percent are prepared by a commercial tax preparer and likely incur a fee. This is similar to, but somewhat higher than, the results for the United States reported by
Goodman et al. (2023).
Finally, we calculate that there are up to 4.6 million returns where the tax filer has a simple tax situation (a low personal income below $25,000) and has perhaps paid a fee to a commercial tax preparer to file a return that could likely have been accurately completed by the CRA. These represent 17 percent of all 27 million tax returns filed with the CRA. Eliminating these tax compliance costs, particularly if these filers are submitting a return primarily to ensure eligibility for income-tested benefits and services, could represent a major cost savings to households.
24Conclusion
Like Laurin and Dahir (2022), we estimate that the CRA could, if so directed, initiate and complete tax returns for a minority of all Canadian tax filers. Providing TAR to that minority would reduce administrative burdens and improve access to the important benefits that tax filing confers.
Our results make clear that the CRA could likely implement TAR for nearly one-third (29 percent) of Canadian families, most of whom would be concentrated in the lower end of the income distribution and would include between 60 and 88 percent of families relying on provincial social assistance. This is not equivalent to a universal system of TAR for all tax filers. In fact, even in countries with return-free taxation, some individuals are still required to complete annual returns (
Laurin & Dahir 2022). We do not believe that an inability to offer TAR to middle- and upper-income Canadians with more complex tax situations should be reason to deny it to those who would benefit most. Tax filers with simple tax situations are most likely to be lower- and modest-income Canadians who have the least ability to pay a tax preparer and the most to gain from the many income-tested benefits and services that require a T1 return. We would argue that the government and its tax agency should not let the perfect be the enemy of the good. The CRA could, and should, move toward completing tax returns for as many Canadians as it can.
With the agreement and participation of provinces (that administer social assistance), the CRA could make use of T5 information slips and other information already collected by social assistance programs to complete the returns of persons in receipt of provincial social assistance.
25 Provinces should be responsible for getting the consent of social assistance clients to share their data and use it in this way. These kinds of information-sharing consents are already a common aspect of provincial social assistance systems.
The literature review in this article identified several potential barriers to the adoption of TAR, even for a subset of all taxpayers. Some of these barriers have become less salient with advances in digital technology. Other barriers remain important. We discuss them here in the current Canadian context.
Potential delays in processing returns in a TAR system (while the tax agency waits for third party information and then populates simple returns) have been raised as a concern. Taxpayers awaiting a refund might object to waiting for the CRA to prepare their return. In Canada, individual tax filers must already wait for their third-party information slips to initiate a return. If one assumes that the CRA has resources that enable it to use the most advanced data processing technologies, delays due to processing of simple TAR returns are unlikely to be a concern.
26Mismatches between third-party information on file with the CRA and a taxpayer's own records have been raised as a concern with TAR (
Goodman et al. 2023). We note, however, that the tax agency already has the power to assess taxes owed with or without an information return and that mechanisms exist for individual taxpayers to review and, if necessary, challenge third-party information on file with the CRA.
27Tax filer trust in the CRA to correctly apply tax measures may also be a concern regarding TAR. When choice or discretion is used in applying certain credits (such as pension income splitting or transfers of claims for medical expenses), we recognize there is potential for taxpayers to mistrust the discretion of the tax agency. When returns require little or no choice to be accurately completed, however, this concern may be harder to justify. It is also important to note that the CRA already retains the right to reject claims for certain tax measures, and, in the current system, some adjudicated credits see as many as one in five claims subjected to back-and-forth communication between a tax filer and the CRA before being decided in the tax filer's favour (
Tedds & Robson 2023). To be fair and accountable, a TAR system must permit tax filers to review, correct, and dispute the assessment of the CRA, consistent with the rights of tax filers in the current system.
Industry objections to the introduction of TAR include the argument that the industry already provides eligible tax filers with free options to complete their returns with limited administrative burden, potentially reducing demand for TAR.
28 Those objections will no doubt persist. In Canada,
Tax-Filer Empowerment Canada (2023a), the lobbying arm of the Canadian tax preparation industry, argues that
Canadian tax filers are best served by a system that keeps the role of the tax collector, investigator, auditor and enforcer, separate. The separation of these roles enshrines the right of Canadians to independently prepare and file their income taxes with impartial and client-focused services provided by our members, while enabling the CRA to focus its efforts on assessing tax returns and taking enforcement action when necessary.
Despite industry concerns, we note that the IRS is piloting a web application called Direct File that permits truly free filing options (
IRS 2023). This could serve as a pathway to resurrect the Canadian Digital Services’ aborted efforts to create a similar service for the CRA (
Robson and Schwartz 2021). Given that the Government of Canada has now committed to “pilot a new automatic filing service that will help vulnerable Canadians who currently do not file their taxes receive the benefits to which they are entitled” (
Canada 2023, 38–39), we anticipate that industry objections will accelerate.
Others have considered the effects of complex tax measures for the prospects of TAR and have concluded that general tax simplification is a necessary pre-condition for success. This can be taken to mean cancelling tax measures and raising effective tax rates. However, our results have identified the specific tax measures that most frequently present obstacles to TAR for most Canadians. These are claims for medical expenses, charitable donations, education credits, and self-employment income. Furthermore, our results suggest that the majority of tax filers in Canada make use of just one or two of these tax measures. We therefore reject the suggestion that these tax measures would need to be eliminated to enable TAR for an important share of Canadians. Indeed, countries that have return-free taxation have found administrative mechanisms to enable taxpayers to benefit from similar credits or deductions without the requirement of a complete tax return (
HM Revenue and Customs 2023;
Inland Revenue 2021;
Revenue 2023; Skat 2023; Saktterverket 2023). A critical review of those administrative mechanisms (such as enhanced third-party reporting, digital reporting for specific credits, and integrating more credits into payroll withholding rates) is outside of the scope of this study but could yield useful options for consideration. Still, we recognize that some Canadians will continue to be required, or may prefer, to submit a personal income tax return.
The 2020 Speech from the Throne committed the federal government to work toward automatic filing for Canadians with simple returns (
Canada, Parliament 2020).
29 More recently, the 2023 federal budget has promised a pilot project to offer an automatic filing service, beginning with vulnerable Canadians. The government of Quebec announced an almost identical pilot project as part of its 2023 budget. Quebec operates its own income tax system and administration in parallel with the federal income tax system. We welcome these developments with cautious optimism. The results presented in this article will be relevant to discussions about the direction, scope, and details of potential pilot initiatives and policy changes thereafter.