In my last blog, I discussed my concerns about the IRS’s current use of math and clerical error authority under IRC §§ 6213 (b) and (g). This week, I’ll take a closer look at the proposal to expand math error authority to “correctible errors,” and explain why I think Congress should not grant such expansion in two of the three instances. Although the Treasury Department recently scored this proposal as raising a significant amount of revenue, the concerns I raise in these two blog postings should make folks think deeply about whether we should raise revenue at the expense of taxpayer rights.

The Sufficiency of Documentation Can Be Ambiguous and Difficult to Explain

The “correctible error” proposal contains a broad grant of authority to the IRS to use summary assessment procedures where a required form or schedule is not attached to the return.  However, it is unclear from the proposal whether these procedures will be used to deny benefits due to a lack of sufficient documentation, as opposed to no documentation at all.

A real-world example from a few years ago illustrates why this distinction matters.  Congress authorized the IRS to use math error authority to deny the First-Time Homebuyer Credit (FTHBC) to taxpayers who did not attach a “settlement statement,” as required.  (IRC § 6213(g)(2)(P)(iii))  Initially, the IRS accepted a settlement statement as sufficient only if it showed all parties’ names and signatures, the property address, sales price, and date of purchase. After learning that not all states required a settlement statement to include a complete address or both parties’ signatures, the IRS reversed its position. In fact, the IRS’s handling of FTHBC issues in the 2011 filing season delayed processing over 128,000 returns, by IRS estimates, and led to a sharp increase in related TAS cases (from 669 through April 30 of fiscal year (FY) 2010 to 4,299 for the same period in FY 2011).

Clearly, the use of math error authority in this circumstance would have been unwise. To make this and other determinations about the sufficiency of a settlement statement, an IRS employee had to read papers attached to the return and explain any problems to the taxpayer (or summarily assess the liability without providing a good explanation, thereby violating the taxpayer’s right to be informed). Accordingly, I recommended the use of math error authority only when a return does not contain a document that purports to be a settlement statement (i.e., a simple yes/no determination) and leaving the facts-and-circumstances determination of the sufficiency of the settlement statement to normal deficiency procedures.

A related problem arises from the differences between e-filed returns and paper returns. Running counter to Congress and the IRS’s efforts to increase e-filing, taxpayers required to provide documentation to substantiate a return position generally must file paper returns. A much-needed investment in the IRS’s systems to allow taxpayers to file required documentation electronically instead of on paper would go a long way toward improving tax compliance while still preserving taxpayer rights. The IRS has processes for handling incomplete paper returns and could develop similar ones for e-filed returns. If an incomplete return were e-filed, the IRS could simply reject it at the outset, alerting the taxpayer or preparer immediately that more information is needed and allowing the taxpayer to cure the defect and re-file the return.

The proposal to expand math error authority (or “correctible” errors) in this context is like the tail wagging the dog and is driven by the IRS’s 20th century technology. We should be designing tax administration looking forward, not backward.

Government Databases Can Be Unreliable for Tax Purposes, Such That Accurate Returns May Appear Inconsistent with Third-Party Data

The “correctible error” proposal authorizes the IRS to use summary assessment authority where the information provided by the taxpayer does not match the information in government databases. I have long recommended the IRS not use math error authority to correct discrepancies between information shown on a return and information from government databases that are not sufficiently reliable for tax purposes. For example, the IRS has the authority to assess math errors against Earned Income Tax Credit (EITC) returns that are inconsistent with the Federal Case Registry of Child Support Orders (FCR) database – where a person listed as a noncustodial parent in the FCR database claims the child. (IRC § 6213(g)(2)(M)) However, the IRS has not done so (correctly, to my thinking) because a study, which Congress mandated be undertaken with my office, showed that the FCR was not sufficiently reliable for purposes of verifying a child’s residence. The study found that up to 40 percent of the cases selected solely based on FCR data were incorrect. (This data comes from an unpublished Treasury report.) Thus, while the FCR is useful for identifying questionable returns and selecting them for audit, it is not appropriate as a basis for summarily denying a credit or exemption.

Moreover, applying data collected for nontax purposes to tax claims is akin to relying on the addresses shown in a telephone directory to deny the home mortgage interest deduction. Even if virtually all of the entries in a directory were accurate, they were compiled for a different purpose, do not disprove eligibility under the tax law, were compiled at a prior date and may not be current, and should not deprive a taxpayer of a due process right to present his or her own facts.

The definition of what constitutes a “government database” is itself problematic. The “correctible error” proposal has been touted as reducing EITC improper payments, but it is unclear to me how it can do that unless “government databases” include the IRS’s Dependent Database (DDb), a compilation of business rules and different datasets. Each return that claims a dependent or other family-status benefit (like the EITC) is run through the DDb. While some of the underlying data is reliable (e.g., Kidlink, which contains Social Security Administration information linking a child’s Social Security number (SSN) to its mother’s SSN, and in many instances, the father’s SSN), other data – like the FCR – are unreliable.

The DDb has value — it is a collection of circumstances from which the IRS is inferring the likelihood of error. But it is not a binary (yes/no) determination that makes it suitable for summary assessment authority. The Taxpayer Advocate Service (TAS) has seen instances where a taxpayer’s return has broken nearly all of the rules contained in the DDb and the taxpayer is still eligible for the exemption or credit claimed. The results derived from the DDb are probabilistic in nature. It is unprecedented to give the IRS summary assessment authority based on some unstated probability that it is correct. To undermine taxpayers’ right to petition the Tax Court based on a probability is equally unprecedented.

My concerns about the unreliability of IRS “government databases” are founded in experience.  For example, in FY 2016, the IRS delayed over 2.1 million refunds based on its filters as part of the Taxpayer Protection Program, yet 53 percent of those refunds were later deemed to be “false positives.” I am concerned that if the IRS rejects returns with valid refund claims or adjusts returns using math error-like procedures, it may prevent taxpayers from receiving the refunds to which they are entitled. Inconsistencies between a return and data that is not sufficiently reliable or determinative may indicate the IRS should do further research or initiate an audit, but should not automatically trigger summary assessment procedures, which unnecessarily burden taxpayers and the IRS.

For these reasons, I recommended in 2011, and again in my 2014 report, that Congress:

• Confine the IRS’s use of math error authority to instances that are not factually complex,
• Permit the IRS to use math error authority only in conjunction with databases that are reliable and accurate,
• Restrict math error authority in situations with a high abatement rate, and
• Require the Department of the Treasury, in consultation with the National Taxpayer Advocate, to evaluate and report to Congress on whether any proposed expansions satisfy these criteria.

I also recommended that the report should analyze the burdens and benefits of the proposed use of math error authority, considering downstream costs such as those for audit reconsideration and TAS intervention, and rigorously analyze the proposed expansions for accuracy and suitability. The Government Accountability Office has proposed similar safeguards. As noted above, Congress mandated a similar study before the effective date of the IRS’s math error authority to address FCR data mismatches, a study that the IRS would not have undertaken without the mandate.

I am certainly not against the use of math or clerical error. In many instances, it is a useful and efficient tool to correct unambiguous errors. It can be used appropriately – that is, without undermining taxpayer rights and the due process protections afforded by pre-payment access to the United States Tax Court – where a statutory provision imposes age limits or lifetime dollar caps, or where there are truly relevant and accurate government databases that do not raise facts-and-circumstances determinations. Congress has and can continue to legislate specific grants of summary assessment authority in these areas. But as these two blog postings have shown, expanding the IRS’s summary assessment authority beyond these necessarily narrow boundaries raises serious issues relating to due process and taxpayer rights concerns.  For almost nine decades, since 1926, Congress has consistently expressed concern about granting the IRS unbridled authority in this area, and it should continue to do so.