Two recent Tax Court cases show us that while the §6751(b) supervisory approval requirement does apply to a tax penalty mechanically applied by a human employee it does not apply to the same penalty mechanically applied by a computer. As a result, two similarly situated taxpayers get treated differently. One gets penalized and the other does not. It is an understandable result, but not a sensible one. To me, it shows the incoherence of the statute.
In Andrew Mitchell Berry and Sara Berry v. Commissioner, T.C. Memo, 2021-42 (Apr. 7, 2021), Judge Marvell holds that a §6662(b)(2) understatement penalty is invalid without proper supervisory approval when proposed as a matter of routine in a 30-day letter issued by a Revenue Agent. In contrast, Anna Elise Walton v. Commissioner, T.C. Memo. 2021-40 (Mar. 30, 2021) (Judge Urda) explains why supervisory approval is not required for the very same penalty if it is first proposed in a computer-generated CP2000 notice, issued without any human involvement. Both the 30-day letter and the CP2000 notice serve the same function, to encourage the taxpayer to engage with the IRS to ensure the accuracy of their returns. Yet the penalty proposed in one requires 2 humans to approve and the penalty proposed in the other requires no human approval.
These cases are straightforward applications of the statute. They are unremarkable in their conclusions that human-proposed penalties need human review but computer-proposed penalties do not. That is what the statute indeed says. However, what makes them worth your time is that they demonstrate the strange interaction of penalty statutes and tax administration. Here we have two equally culpable (or innocent, take your pick!) taxpayers, but only one gets hit with the same mechanically-computed penalty and that solely because of the difference in how the penalties are first proposed. The difference is between what is routine and what is automatic. It’s a difference created by how the IRS operates, the language of the statute, and the Tax Court’s interpretation of that statute. And it’s a difference that makes little sense, at least to me. I think there is a better distinction to be made.
If you are already a tax penalty jock and know how incoherent the system is, you do not need this lesson. Otherwise, I invite you to dive into the details below the fold.
Law: The Mechanical Nature of the Substantial Understatement Penalty
What’s the point of penalties? Well, according to the IRS, their purpose is to promote voluntary compliance by “(1) defining standards of compliant behavior; (2) defining consequences for noncompliance, and (3) providing monetary sanctions against taxpayers who do not meet the standard.” Penalty Handbook, IRM 220.127.116.11. Over the years Congress has massively expanded the number and severity of penalties. In 1955 there were about 14 penalties. Currently there are more than 140. IRM 18.104.22.168.1 (11-25-2011). See generally, Penalty Study in the National Taxpayer Advocate 2012 Annual Report To Congress, Vol. 2 at p. 151. Not all penalties serve all three of the purposes set out in the IRM. I want to pay particular attention to the §6662(b)(2) substantial understatement penalty, because that is the common penalty in today’s cases.
The §6662(b)(2) penalty connects most strongly to that first purpose listed in the IRM, the norm-defining purpose. Professor Michael Doran has a lovely exposition on that first purpose in his article Tax Penalties and Tax Compliance, 46 Harv. J. on Legis. 111 (2009). Worth your time reading. The basic idea is that our system of taxation depends in the first instance on taxpayer’s submitting accurate (not just non-fraudulent, but accurate) returns. Yet the complexity that Congress loves to add to the laws sometimes makes it really hard to be accurate.
The substantial understatement penalty in §6662(b)(2) defines the tolerance for inaccuracy. It imposes a 20% penalty for any understatement of tax that exceeds the greater of $5,000 or 10% of the tax that was supposed to be reported. So that’s the nominal wiggle room, the tolerance. And it’s a purely mechanical calculation.
Where the norm-defining nature of this penalty truly comes into play, however, is in the decidedly non-mechanical ways taxpayers can avoid the penalty. First, a taxpayer can avoid the penalty if the understatement results from a position that had either “substantial authority” (if the position was not disclosed on the return) or, even more leniently, a “reasonable basis” (if disclosed on the return). See §6662(d)(2)(B). Second, the penalty won’t apply to any portion of the understatement if the taxpayer shows “there was a reasonable cause for such portion and that the taxpayer acted in good faith with respect to such portion.” §6664(c)(1). That is, if the taxpayer can show they behaved reasonably under the circumstances, the IRS will either not propose or will remove a proposed substantial understatement penalty.
Thus, the mechanical calculation is just a first cut; it creates a presumption that the taxpayer’s return is outside the norms of accuracy tolerances. The taxpayer, however, can overcome that presumption.
To get past the mechanical application of the penalty, a taxpayer needs to engage with the IRS. The burden is on the taxpayer to show “substantial authority” “reasonable basis” or “reasonable cause.” The regulations tell us, not surprisingly, that what constitutes reasonable cause depends on the facts and circumstances, but the over-arching inquiry is “the extent of the taxpayer's effort to assess the taxpayer's proper tax liability.” Treas. Reg. 1.6664-4(b)(1). That being so, it makes sense why one sees this issue frequently litigated when the taxpayer and the IRS dispute whether a taxpayer had good cause. At that point, a court will decide. But the point remains: the taxpayer must engage.
Even as it promiscuously promotes penalties, Congress expresses contradictory concerns that the IRS should not impose penalties as a matter of routine but should “make a correct substantive decision in the first instance rather than mechanically assert penalties with the idea that they will be corrected later.” H.R. Conf. Rep. No. 101-386, at 661 (1989)(emphasis supplied), as quoted in National Taxpayer Advocate's 2012 Annual Report to Congress 2012, Vol. 2, note 34.
The idea of a correct substantive decision in the first instance is most strongly expressed in §6751(b). Subsection (b)(1) creates a supervisory approval requirement. As interpreted by the Tax Court, the requirement is that before the first IRS communication of a proposed penalty to the taxpayer, the supervisor of the person who first proposes the penalty must have approved of the proposed penalty in writing. Belair Woods, LLC v. Commissioner, 154 T.C 1 (2020).
Subsection (b)(2) provides exceptions to the supervisory approval requirement: “(A) any addition to tax under section 6651, 6654, or 6655”; or “(B) any other penalty automatically calculated through electronic means.” In Walquist v. Commissioner, 152 T.C. 61 (2019), the Tax Court held that penalties proposed in an 30-day letter issued by the Automated Correspondence Exam system were “automatically calculated through electronic means” within the meaning of §6751(b)(2)(B) because the 30-day letter had been issued without any human being putting eyeballs on it.
Today’s pair of cases show why that holding—as reasonable as it is standing alone—creates incoherence in the penalty regime. Let’s take a look.
Facts of Berry v. Commissioner
In 2013 Mr. Berry Sr. and his son Mr. Barry Jr. were in the construction business, though an S Corporation called Phoenix. The 2013 returns of the taxpayers and Phoenix were selected for audit. It does not appear that the taxpayers provided information to the Revenue Agent (RA) assigned to the returns. The opinions says “As part of the audit RA Chavez issued information document requests to petitioners requesting their accounting records for 2013, but petitioners did not respond.” As practitioners know, IDRs are a routine part of examination. See IRM 22.214.171.124.2. Taxpayer do not have to respond or engage, but they then bear the risks of non-response, which include having the RA draw negative inferences from the missing information.
RA Chavez drew many negative inferences! He inferred from the records he could find that Phoenix: (1) failed to report over $500,000 of income; (2) erroneously deducted $121,000 for Mr. Berry Jr.’s car-racing activity; (3) erroneously deducted $9,000 spent to rent space to work on Mr. Berry Jr.’s race car; and (4) erroneously deducted $19,000 for truck and car expenses. He also determined that Mr. Berry Sr. faked a Schedule C by improperly characterizing certain payments from the Phoenix to him as sole prop. income, when Phoenix had reported those as wages, and then putting erroneous deductions on the Schedule C. Nothing readers have not seen before in practice.
When RA Chavez went to calculated the resulting deficiency in tax, it was way more than the greater of $5,000 or 10% of the tax that was supposed to be reported. So, as a matter of routine, he put the §6662(b)(2) substantial underpayment penalty into the Revenue Agent Report (RAR) that he sent to the taxpayers on April 11, 2016 as an attachment to a 30-day letter. It was electronically signed by his supervisor. But the supervisor did not sign a penalty approval form until May 6, 2016.
The taxpayers made no response to the 30-day letter. Exam eventually issued the NOD on August 18, 2016. The taxpayers timely filed their petition in Tax Court. Only then, as part of the pre-trial process, did the taxpayers finally supply their accounting records.
Too little too late. Judge Marvel sustained all of RA Chavez’s negative inferences. However, she threw out the §6662(b)(2) penalty because the IRS had stipulated that the penalty was not approved until after the RAR and 30-day letter were issued. Writes Judge Marvel: “because the managerial approval did not occur until after the penalties were initially determined and communicated to petitioners, respondent cannot meet his burden of production. Petitioners, therefore, are not liable for the accuracy-related penalties.” Op. at 23.
Facts of Walton v. Commissioner
Ms. Walton, a respected (and highly paid) consultant on corporate governance, executive leadership and development, got snagged by the IRS Automated Underreporter Program (AUR). Here’s why.
Ms. Walton had started her own business in 2015. In January 2016, she turned to her accounting firm of 20 years to prepare her 2015 tax return. The process took until September 2016. Along the way various miscommunications resulted in her return omitting approximately $169,000 of payments reported on two 1099’s. While she missed those, the AUR did not. On October 17, 2017, the AUR system spit out a computer-generated letter CP 2000 proposing an additional tax liability of $62,500 and a §6662(b)(2) penalty of $12,500. Again, the initial calculation of the §6662(b)(2) penalty was purely mechanical.
Just like the Berrys, Ms. Walton failed to respond to the CP2000. She did not engage, and the IRS then sent her an NOD. She then filed a petition in Tax Court, arguing that she had reasonable cause for her inaccuracy. Judge Urda was unimpressed. He faulted her for not paying closer attention to the return that was filed, writing “even a cursory review of her return would have revealed the omission of $169,426—over 32% of her total nonemployee compensation for that year.”
Judge Urda also took time to explain why there was no requirement here for the IRS to show that the §6662(b)(2) penalty had received supervisory approval. “Because the penalty was determined mathematically by a computer software program without the involvement of an IRS examiner, we conclude that the penalty was ‘automatically calculated through electronic means.’ Respondent thus was not obligated to comply with the supervisory approval requirement.” Judge Urda then cited to Walquist.
Lesson: The Incoherence Of §6751(b)
As both written and interpreted, §6751(b) undermines the ability of the §6662(b)(2) penalty to enforce tax preparation norms of behavior. That is because the statute attempts to distinguish between what is routinely done by humans and what is automatically done by computer. That’s a bad distinction.
Congress wrote §6751 as an anti-abuse statute predicated on an outdated view of IRS tax administration as a series of individualized IRS-taxpayer interactions. HA! A cynic might put it this way: sure, in the old days individual IRS employees put the screws to individual taxpayers, but now ex-ante programming decisions put the screws to entire classes of taxpayers, automatically! The notion that the computers treat all taxpayers the same is false. Taxpayers who respond are treated very differently than taxpayers who are so unfortunate as to not be able to respond in the precise way in the tight time periods they are given.
IRS computers abuse more taxpayers than do IRS employees. That is because IRS computers use automatic bright-line rules and those rules always assume the worst of taxpayers. For example, mis-matched 1099’s are presumed to represent unreported income. In the collection world, the Automated Collection System (ACS) always assumes that all taxpayers can pay but just won’t. See Lesson From The Tax Court: Using CDP To Stop The Collection Train, TaxProf Blog (Oct. 15, 2018).
The better distinction may be between penalties proposed without taxpayer input and penalties proposed after taxpayer input. That is, responsive taxpayers are distinguished from non-responsive taxpayers. A non-responsive taxpayer is non-responsive, whether that non-response is to a computer-generated missive or human-generated missive. What is required for the non-responsive goose should be the same as what is required for the non-responsive gander. Similarly, responsive taxpayers should get the same rules, regardless of whether what prompts their engagement is computer-generated or human-generated.
Here’s how that would work. The §6662(b)(2) penalty is initially calculated by a formula. It’s a mechanical declaration of when a taxpayer’s return exceeds the basic tolerance for accuracy. Whether included as part of a CP2000 or 30-day letter, the penalty is automatically calculated by through electronic means. The only difference is that no human presses a button to generate the calculation in the CP2000 process and a human does press a button to generate the calculation in the audit process.
Once a taxpayer provides evidence of substantial authority, reasonable basis, or reasonable cause, however, the decision to impose the penalty is no longer automatic or mechanical. At that point it is a human judgment call. So that is the point where it makes sense for the supervisory requirement of §6751(b) to kick in.
This distinction has two consequences. First, a taxpayer should not be able escape the penalty by being non-responsive. Second, however, neither should the IRS be allowed to avoid the §6751(b) requirement just because the initial penalty is spit out by a computer. The IRS has a duty to properly apply penalties and what triggers that duty for the §6662(b)(2) penalty is the engagement of the taxpayer and the resulting need for an IRS employee to exercise judgement about the appropriateness of the penalty. Until then, there is no reason to treat an RAR or 30-day letter differently than a CP2000. In neither situation is the employee exercise a judgment about the appropriateness of the penalty: it’s just a mechanical first cut.
The IRS has actually operationalized this distinction between responsive and non-responsive taxpayers. In the Automated Correspondence Exam part of the IRM, IRM 126.96.36.199.2(5), the IRS instructs examiners that a supervisor must affirmatively approve the ACE penalties before the NOD goes out when a taxpayer is responsive to the computer-generated penalty proposals.
As applied here, neither taxpayer engaged with the IRS before the NOD was sent out. Ms. Walton did not respond to the CP2000 letter. The Berrys did not respond to the equivalent 30-day letter with the RAR attached. For this penalty, the distinction between the computer-calculated penalty and the human-calculated penalty should make no difference because neither assertion was the result of human judgment about the norm of behavior being enforced by the penalty. Application of the $5,000/10% rule is automatic.
In contrast, if either of the taxpayers had engaged, it would have then be necessary for the supervisory approval requirement to kick in. Thus, if Ms. Walton had responded to the CP2000, that would now trigger the need for human judgment. Likewise, if the Berrys had engaged with the RA after receiving the 30-day letter to show why the penalty was inappropriate, that is when the RA would need to exercise judgment about the penalty and obtain supervisory approval to put it in the resulting NOD.
None of today's taxpayers lived up to the norm enforced by the §6662(b)(2) penalty: reasonable behavior by the taxpayer to ensure an accurate return. These taxpayers should be treated alike. That they were treated differently makes no sense to me, even though one cannot fault the Tax Court’s interpretation of the statute.
It’s just an incoherent statute.
Bryan Camp is the George H. Mahon Professor of Law at Texas Tech University School of Law. He invites readers to return to TaxProf Blog each week for a new Lesson From The Tax Court.