Seventh Circuit Affirms Spouse Is Not So Innocent on Appeal

I. Introduction

Clients often ask tax practitioners what their prospects of success are in litigation. Although the practitioner may have an opinion on the issue, the chances of success are often difficult if not impossible to quantify accurately and reliably. The question becomes even more complicated at the appellate stage because standards of review come into play, and the evidence to draw on is generally limited to what has already been considered by the lower court. New technology is bringing about change. Practitioners can now leverage machine learning systems trained on data from all other relevant decisions to assess the strength of their appeals on the merits. Consequently, tax practitioners and clients can together make data-driven decisions about whether to appeal and, if they do so, to formulate an optimal strategy.

In this article, we examine Rogers,¹ a case about innocent spouse relief that was recently decided by the Seventh Circuit. We use this case to illustrate how Blue J’s machine learning technology could have been used by the appellant’s counsel to assess the likelihood of success on appeal and the key factors required to succeed. The appeal in question relates to two Tax Court decisions denying a wife innocent spouse and equitable relief under section 6015(b) and (f), respectively. Both decisions were affirmed by the Seventh Circuit. Blue J’s machine learning technology predicted with greater than 95 percent confidence that based on the facts accepted by the lower court, the decisions would stand. We next provide a brief background of the facts of the case, the applicable law, and the decisions of the Tax Court and the Seventh Circuit.

II. Background

A. Mr. and Mrs. Rogers

John and Frances Rogers have been married for approximately 50 years. John is a Harvard-educated tax attorney with over 40 years of experience. John’s aggressive tax planning strategies eventually attracted the attention of the IRS — and were discredited on multiple occasions by the Tax Court and the Seventh Circuit.² Frances is a highly educated woman with a master’s degree in biochemistry, an MBA, a doctorate in educational administration, and a law degree. Frances mostly worked as an educator but stepped in to act as an administrator for her husband’s law firm in 2009 when he was incapacitated because of alcoholism.

For the tax years at issue, the Tax Court held that the Rogerses, who filed jointly, had unreported income and inappropriate business deductions.³ Frances sought and was denied relief as an innocent spouse under section 6015.

B. Section 6015 Relief

Married couples who file joint returns are jointly and severally liable for taxes due under section 6013(d)(3). However, the code provides relief from that tax liability under some circumstances — and Frances sought innocent spouse and equitable relief under section 6015(b) and (f), respectively.

1. Innocent spouse relief.

Innocent spouse relief is only available if all the conditions set out in section 6015(b)(1) are met. For the purposes of Frances’s appeal, the two most important requirements were: (1) that the spouse requesting relief establishes they did not know, and had no reason to know, that there was an understatement of tax when signing the return; and (2) taking into account all the facts and circumstances, it would be inequitable to hold the requesting spouse liable for the deficiency.

Regarding knowledge, if the requesting spouse had actual knowledge or a reason to know of the understatement of tax, then they are not eligible for innocent spouse relief. Facts and circumstances considered in determining whether the requesting spouse had reason to know are outlined in reg. section 1.6015-2(c). These overlap, to some extent, with those set out in the Seventh Circuit’s decision in Resser.⁴ Relevant facts and circumstances include the requesting spouse’s level of education and involvement in the family’s financial and business activities, the presence of unusual or lavish expenses beyond the family’s norm, and the culpable spouse’s deceitfulness about the family finances. Courts balance all the relevant factors to decide if the spouse requesting relief had reason to know about the understatement.

The equity requirement similarly includes a facts and circumstances analysis, which is guided by reg. section 1.6015-2(d). One particularly relevant factor is whether the requesting spouse significantly benefited from the understatement.

2. Equitable relief.

If innocent spouse relief is unavailable under section 6015(b), a spouse may still be granted equitable relief under section 6015(f) if it would be inequitable to hold them liable after consideration of all the facts and circumstances. Rev. Proc. 2013-34, 2013-43 IRB 397, sets out three guidelines for evaluating requests for equitable relief: section 4.01 lists threshold conditions that a requesting spouse must satisfy; section 4.02 sets out a test for a streamlined determination to grant relief; and if the taxpayer is not entitled to a streamlined determination, section 4.03 sets out a nonexclusive list of factors to consider when determining whether it would be inequitable to hold the spouse jointly and severally liable. The nonexclusive list of factors includes:

• the marital status of the spouses;

• the economic hardship to the requesting spouse if relief is denied;

• whether the requesting spouse had knowledge or reason to know of the item giving rise to the understatement;

• whether either spouse had a legal obligation to pay the tax liability;

• whether the requesting spouse significantly benefited from the understatement;

• whether the requesting spouse complied with income tax laws in subsequent years; and

• whether the requesting spouse was in poor mental or physical health.

Abuse of the requesting spouse by the other spouse is also significant and can negate the presence of other factors.⁵ Abuse can be physical, psychological, sexual, or emotional, and can include efforts to control, isolate, humiliate, and intimidate the requesting spouse, or to undermine the requesting spouse’s ability to reason independently.⁶

C. Tax Court Decisions

The 2018 Tax Court decision involved tax years 2003, 2005-2007, and 2009.⁷ Frances was denied innocent spouse relief under section 6015(b) for all those years.⁸ The court held that Frances knew or had reason to know of the understatements of tax, and it was not inequitable to hold her liable. The court further held that Frances was not eligible for equitable relief under section 6015(f). Although there was some debate about whether the threshold requirements for equitable relief under Rev. Proc. 2013-34 were met, the court found equitable relief was unavailable based on the multifactor facts and circumstances test. Although not directly discussed in the equitable relief analysis, it is noteworthy that Frances was represented by her husband in the trial proceeding and successfully fought the IRS’s attempt to remove him as counsel based on a conflict of interest.⁹

The 2020 Tax Court decision dealt with tax years 2010-2012,¹⁰ and this time, Frances was represented by independent counsel.¹¹ Nine additional witnesses were called to testify on Frances’s behalf, but the Tax Court again denied all relief. Little had changed between the 2018 and 2020 decisions, except that Frances arguably became even more involved with her husband’s business. The court found that Frances had even more reason to know about the understatements of income based on the long history of the couples’ tax troubles. The court also denied equitable relief under section 6015(f). This time, the court conceded that the couple did not live a lavish lifestyle, contrary to what it found in the 2018 decision, but this did not alter the outcome.

D. Appellate Decision

The appeals of both the 2018 and 2020 Tax Court decisions were decided together. Frances’s brief told the story of an abused woman who was dominated by her husband and who had no knowledge of or say in the financial matters of the household. Virtually all elements of the Tax Court’s decisions were criticized — the brief alleged many factual and legal errors in the court’s analyses. In an opinion handed down without oral argument, the Seventh Circuit affirmed the Tax Court decisions in their entirety.¹² We next consider how Blue J’s machine learning technology would have predicted this outcome with high confidence and which factors were most influential in driving the predicted result.

III. Machine Learning Analysis

A. Blue J’s Prediction

Blue J uses machine learning to assess and model a data set of approximately 350 cases that decide whether innocent spouse, equitable, or other relief is available under section 6015. The factors used in Blue J’s predictions are those that judges refer to and rely on in their decisions. After a user inputs all the relevant facts of their case, Blue J’s model predicts how likely it is that a court would grant innocent spouse or equitable relief based on those stipulated facts and circumstances. Blue J also attaches the degree of confidence it has in that prediction.

Based on the facts as found by the Tax Court in Frances’s case, Blue J would have predicted with greater than 95 percent confidence that neither innocent spouse nor equitable relief would have been granted on appeal. Frances faced an uphill battle — this was not a close case.

The innocent spouse and equitable relief analyses necessarily require courts to assess the evidence, make findings of fact based on that evidence, and weigh those facts against each other to conclude on whether relief is appropriate. Several facts were not in dispute between the parties in the Rogerses’s case, such as Frances’s extraordinarily high level of education and her review of the tax returns each year. But several other factors were disputed throughout the proceedings.

B. Disputed Facts: Section 6015(b)

Some of the main points of contention regarding whether Frances knew or had reason to know about the understatement of tax in the innocent spouse relief analysis included her:

• business and tax experience;

• involvement in the family businesses;

• involvement in household finances;

• knowledge of previously filed incorrect tax returns; and

• being deceived by her husband.

1. Business and tax experience.

Frances’s business and tax experience was a main point of contention in these proceedings because it affected whether she knew or had reason to know about the understatement of tax. The Tax Court rejected her argument that at best she only had a nominal understanding of finance, tax, and accounting.¹³ The court found her assertions not credible because she was highly educated, had taken tax courses in law school, and had managed and participated in significant business dealings — including planning a subdivision development and running her husband’s law firm in 2009.¹⁴ The Seventh Circuit found no clear error in the Tax Court’s findings, emphasizing that Frances’s education, particularly her tax courses, undercut the credibility of her argument.¹⁵

2. Involvement in family business.

The level of Frances’s knowledge and involvement in her husband’s business activities was also in dispute. The Tax Court did not believe Frances’s claim that John controlled their finances. To the contrary, the court determined that she was actively involved in significant business dealings especially after he was being treated for alcoholism in 2009.¹⁶ The Seventh Circuit found no clear error in the Tax Court’s findings.¹⁷

3. Involvement in household finances.

The Tax Court further found that Frances participated in the household finances; paying bills, making purchases from their joint bank accounts, making deposits, and endorsing checks from joint accounts for both personal and business expenses.¹⁸ Frances also reviewed the couples’ joint tax returns every year, asked questions about them, and was aware they paid little or no income tax even though they earned a lot of money. The Seventh Circuit affirmed the Tax Court’s findings.¹⁹

4. Knowledge of previously filed incorrect tax returns.

Another factor in dispute was the degree to which Frances knew of prior tax compliance issues, which could have alerted her to a problem with the returns involved in these cases. The Tax Court noted that the number of tax years at issue indicated to Frances that there was a pattern of noncompliance and that the IRS’s “relentless” attack on her husband’s tax shelters should have been a red flag.²⁰ The court thought that Frances should have suspected there would be problems with their joint returns for the years at issue, indicating there was reason to know about the understatements. Frances’s position on appeal was that she was repeatedly assured by John, an accomplished tax attorney, that their tax returns were accurate and completely separate from the returns that were problematic in previous tax years.²¹ The Seventh Circuit did not specifically address this factor when affirming the Tax Court’s findings in their entirety.

5. Deceit.

Deception is also a factor that is considered in the knowledge analysis. The Tax Court held that there was nothing in the record indicating that John tried to deceive or hide anything from Frances that could negate any knowledge she might have had about the understatement of tax. The court found that Frances had complete access to their financial information. She had reviewed and asked questions about the returns. She had access to bank accounts, credit card statements, and other financial records. And she had accompanied her husband to many business dinners and meetings.²² On appeal, Frances did not deny having access to all the financial information but instead argued that without an explanation of the complex tax issues in the returns, the information was virtually meaningless to her.²³ The Seventh Circuit found that the Tax Court did not clearly err when it concluded there was no evidence that John deceived her.²⁴

Table 1: Scenario Testing Disputed Facts for Innocent Spouse Relief

	Business Experience	Involved in Family Business	Involved in Household Finances	Aware of Previously Incorrect Returns	Deceit	Innocent Spouse Relief
Scenario 1 (all favorable to Frances)	No	No	No	No	Yes	Relief
Scenario 2	No	No	No	No	No	No relief
Scenario 3	No	Yes	No	No	Yes	No relief
Scenario 4	No	No	Yes	No	Yes	No relief
Scenario 5	No	No	No	Yes	Yes	No relief
Scenario 6	Yes	No	No	No	Yes	Relief

C. Testing the Facts: Section 6015(b)

To ascertain whether any of the contentious facts were actually determinative of the innocent spouse relief result in this case, we used Blue J’s machine learning technology to systematically test the effect of changing each factor, separately and in combination.

To see the effect of the disputed facts on the outcome, we start with a prediction based on the Tax Court’s findings of fact, which were affirmed by the Seventh Circuit.²⁵ Based on these facts, Blue J’s model predicts with greater than 95 percent confidence that innocent spouse relief is unavailable to Frances.²⁶ Holding all other factors constant, we found that changing the disputed facts from those found by the Tax Court to those more favorable to Frances, one at a time, did not have a significant effect on the prediction. Therefore, not one of these factors independently, taken in isolation, could have changed the outcome for Frances.

We then tested whether finding all contentious facts in favor of Frances would have resulted in her being granted innocent spouse relief (see Scenario 1 in Table 1). We found that if Frances’s arguments on all five disputed factors were accepted, Blue J predicts with 89 percent confidence that she would have been granted innocent spouse relief.

We then tested the effect of systematically reverting each one of the five disputed facts to reflect the findings accepted by the court, one at a time (see scenarios 2-6 in Table 1 below). We discovered that in all scenarios, except Scenario 6, reverting even one disputed fact was sufficient to lead to the likely denial of innocent spouse relief. Interestingly, although the parties argued about Frances’s level of business experience at length, Scenario 6 shows that the same result is achieved whether the court finds that Frances had business experience or not. Blue J’s data confirms that this factor is generally weak.

We conclude that no single factor of those in dispute in the innocent spouse relief analysis was controlling in this case. Rather, all factors considered in combination, except for business experience, would have had to have been determined in Frances’s favor to lead the court to conclude that innocent spouse relief was warranted.

D. Disputed Facts: Section 6015(f)

Two main facts in dispute relevant to the equitable relief analysis were whether Frances was abused by John and whether she would suffer economic hardship if relief from joint and several liability were not granted.

1. Abuse.

Frances claimed throughout these proceedings that she had been psychologically abused by her husband, making her unable to challenge the tax returns, negating any knowledge she had about the understatements, and making it inequitable to find her liable.²⁷ The Tax Court expressed some sympathy for Frances’s situation but did not believe she had been abused by her husband.²⁸ The court was not convinced that John exercised control over Frances to a degree that would have prevented her from challenging the tax returns, specifically pointing to Frances’s trial testimony that she did not fear for her safety.²⁹ In fact, the court found that she was able to exercise control over financial decisions and did review the tax returns at issue with her husband, asking questions along the way. Although the court acknowledged that John’s behavior humiliated and isolated Frances, it was not convinced that his behavior rose to the level of abuse claimed.

On appeal, Frances stressed the narrative that she had been controlled and emotionally abused by John, especially during the years when he was struggling with alcoholism. Specifically, Frances claimed that her husband put her down and was generally dismissive of her, especially regarding financial and tax matters.³⁰ The Seventh Circuit did not specifically comment on Frances’s abuse claims but upheld the Tax Court’s decision denying equitable relief as a whole.

We may speculate that Frances’s abuse claims may have been difficult for both courts to accept because she allowed her husband to represent her during the trial for the Tax Court’s 2018 decision and said in an affidavit that there was no conflict created by his representation. This is speculative, but had the evidence and narrative of abuse been presented from the beginning by independent counsel, perhaps Frances’s allegations of abuse would have appeared more credible to the court.

2. Economic hardship.

The Tax Court assessed Frances’s economic hardship argument as part of its equitable relief analysis. As Rev. Proc. 2013-34 indicates, to show economic hardship, Frances had to show that if she was liable to pay the understatement of tax, she would be unable to pay reasonable basic living expenses. The Tax Court found in 2018 that Frances would not suffer economic hardship because the financial documents she submitted showed that her assets were greater than the deficiencies at issue, and there was some debate about whether there were other undisclosed assets as well.³¹ In 2020 the Tax Court similarly found that Frances had not shown economic hardship.³²

On appeal, Frances pointed to evidence supposedly introduced during the trial for the Tax Court’s 2020 decision to the contrary.³³ But, as noted earlier, the Seventh Circuit upheld the Tax Court’s decision denying equitable relief as a whole, and therefore, much as with its decision regarding abuse, it did not specifically address economic hardship.³⁴

E. Testing the Facts: Section 6015(f)

We consider abuse and economic hardship to determine their relative importance in the court’s equitable relief analysis. We begin with our baseline prediction with greater than 95 percent confidence that equitable relief is unavailable, adopting the facts as found by the Tax Court and affirmed by the Seventh Circuit.³⁵

We next test the importance of the abuse factor. Contrary to what the Tax Court found, we alter the scenario by accepting Frances’s argument that her husband abused and controlled her to prevent her from knowing about or addressing the tax liability and that she feared retaliatory abuse. If that abuse were present, Blue J’s prediction reverses. Blue J’s machine learning technology predicts with 94 percent confidence that Frances would be granted equitable relief.

Once again beginning with our baseline prediction, we then test out the likely effect of accepting Frances’s argument that she would be unable to meet reasonable basic living expenses if she were forced to pay the tax liability. Blue J predicts that equitable relief would be unavailable, and the confidence in that prediction remains greater than 95 percent.

Therefore, the abuse factor is determinative, whereas the economic hardship factor is somewhat insignificant in this particular set of facts and circumstances. If Frances were able to convince the court of abuse, equity would be on Frances’s side.

IV. Conclusion

We have seen that Frances’s chances of successfully appealing the 2018 and 2020 Tax Court decisions to obtain innocent spouse or equitable relief were very low. Blue J’s data suggests that almost all the disputed facts regarding Frances’s knowledge of the understatement would have had to be resolved in her favor to reverse the decision of the Tax Court denying innocent spouse relief. This would have involved the Seventh Circuit finding error in the Tax Court’s conclusions on these points, many of which involved the weighing of evidence and assessing credibility. Blue J’s machine learning model further suggests that a finding of abuse would have likely reversed the court’s conclusion that equitable relief was unavailable.

We have demonstrated how using machine learning technology after receiving a trial decision can help counsel determine their client’s chances of success on appeal. This not only includes reviewing the overall prediction but also identifying the factors that are most relevant, which is important for determining the most effective strategy on appeal.