Tax Notes logo

Using Machine Learning to Evaluate the Existence of a Trade or Business: Olsen

Posted on Feb. 28, 2022
Christopher Yan
Christopher Yan
Benjamin Alarie
Benjamin Alarie

Benjamin Alarie is the Osler Chair in Business Law at the University of Toronto and the CEO of Blue J Legal Inc. Christopher Yan is a senior legal research associate at Blue J Legal.

In this article, Alarie and Yan examine how machine learning can be used to assess the strength of the taxpayer’s position in the appeal of the Tax Court’s decision in Olsen.

I. Introduction

The need to determine whether a taxpayer is engaged in a “trade or business” arises frequently. Although the term “trade or business” is used throughout the IRC and the associated regulations in the Code of Federal Regulations, it is not defined explicitly. In some ways this is surprising. Crucially, establishing that there is a trade or business affects whether a taxpayer can deduct business expenses. Moreover, the phrase “trade or business” is relied on for many other taxpayer rights and obligations. Erring about whether a taxpayer is carrying on a trade or business can have wide-ranging financial consequences for deductions, credits, exemptions, disqualifications, and penalties, among other things.

In last month’s installment1 of Blue J Predicts, we reflected on our predictive analyses and evaluations of tax issues over the past year and described how advances in computing power and machine learning are being applied to evaluate and analyze the strength of tax positions for those issues.

We continue our predictive analysis and evaluative work in this month’s column, examining how machine learning can be used to assess the strength of the taxpayer’s position in the appeal of the Tax Court’s decision in Olsen.2 The Notice of Appeal was submitted September 7, 2021. The appellants’ opening brief was filed January 13. At the time of this writing, the respondent’s brief has not been submitted but is expected to be filed March 7 in accordance with a motion to extend the time to file.

Although multiple issues led to the Tax Court’s original determination in Olsen, we focus primarily on the question of whether the taxpayers were engaged in a trade or business. However, a finding on this issue is not necessarily determinative of their entitlement to the deductions and credits sought because they may also qualify on an alternative basis. There is another path to success for the taxpayers; theoretically, they could prevail on all the other issues on appeal and end up entitled to the tax benefits. Still, the taxpayers’ failure to establish that they were engaged in a trade or business is a significant obstacle for qualifying for the credits and deductions in dispute.

Based on the facts accepted by the Tax Court and the arguments advanced in the appellant’s opening brief, Blue J’s algorithm predicts with greater than 95 percent confidence that the Tenth Circuit will find that the taxpayers were not engaged in a trade or business. We appreciate that a predictive analysis with that high a degree of confidence is unusual for a matter on appeal; we therefore also examine why this may be the case.

II. Overview of Dispute

Olsen involves a dispute over whether Preston and Elizabeth Olsen were entitled to claim depreciation deductions and energy tax credits for purchases of lenses ostensibly intended for solar projects. The solar projects were later held to be a part of a tax shelter scheme. The promoters of the scheme represented that the Olsens would be able to claim energy tax credits and depreciation deductions for any lenses they purchased, effectively neutralizing any federal tax liability arising from the earned income.

The tax shelter scheme was eventually shut down in an indirectly related decision3 in which the Tenth Circuit upheld a district court decision that found the tax shelter to be unlawful and ordered disgorgement of $50 million in gross receipts from the scheme. The Olsens have nonetheless maintained that the deductions and credits claimed for the lenses under the solar projects should be allowed.

To qualify for and use the deductions and credits as claimed, the Olsens must establish three things. First, they must show that they were engaged in a trade or business for the lenses or that the lenses were held for the production of income. Second, the Olsens must show that the lenses were placed in service during the years for which the amounts were claimed. Finally, for the Olsens to apply the losses and credits against regular income, the losses and credits on the claimed activity cannot be connected to a passive activity.

In its decision, the Tax Court sided with the IRS and held that the Olsens were not entitled to the claimed tax benefits and ruled against them on all subissues. In doing so, the Tax Court found that (1) the Olsens’ activities and extent of involvement did not constitute a trade or business (rejecting the alternative possibility that the property in question was held for the production or collection of income), (2) the lenses were never placed in service, and (3) the investment activity was a passive activity. It is noteworthy that the Olsen decision adjudicated the first of more than 200 cases involving investors who participated in this tax scheme.

III. Background

The Olsens are husband and wife. Although they jointly filed tax returns and are co-parties in the action, most of the relevant activity involves Preston Olsen (hereinafter the taxpayer).

In 2009 the taxpayer met with a promoter of the solar energy tax scheme, which involved the purchase of solar lenses to obtain tax benefits. The tax scheme used investment tax credits and accelerated depreciation associated with the solar lenses. The solar lenses were intended to be used for concentrating solar power to generate electricity. The investor would purchase the lenses and, by owning those lenses, be eligible for ITCs and accelerated depreciation deductions. These credits and deductions could then be applied to reduce or potentially eliminate the taxpayer’s federal tax liability.

On July 23, 2009, the taxpayer signed an agreement to purchase $60,000 worth of lenses but was required to make only a 30 percent down payment. The balance of the payment would be due only if and when the project began generating electricity. The contract also provided that the down payment would be refunded if the lenses were not installed by a specific date. On the advice of the promoter, the taxpayer purchased the lenses through the newly created PFO Solar LLC, of which the taxpayer was the sole member. The agreement also designated “LTB” (multiple limited liability companies formed by the promoter) as being responsible for ensuring the installation of the lenses, although LTB itself was not a party to the agreement.

Although the taxpayer did not purchase any lenses in 2010, he carried forward unused credits from 2009. In May 2011 the promoter sought further investment to purchase more lenses and revised the incentive structure by reducing payment required to 10 percent of the down payment. The taxpayer declined to purchase additional lenses until he was able to ascertain his tax liability at the end of the year. Once he was able to do so in December 2011, the taxpayer purchased an additional $49,000 worth of lenses.

The terms of the second purchase were substantially similar to the original agreement with a new purchase structure and were executed concurrently with an agreement to lease the lenses back to LTB. Rent payments would start only if and when lenses began producing revenue from generating electricity, which never occurred, resulting in no rent payments during the course of the lease agreement. Substantially similar purchase and lease agreements were entered into in 2012, 2013, and 2014, resulting in aggregate nominal purchases totaling $242,000.

Consequently, the taxpayer claimed depreciation expenses and ITCs to eliminate the income tax liability from income earned at his day job as a law firm associate. The IRS disallowed all deductions and credits on the basis that the lenses were not used in a trade or business, were not held for the production of income, and were not ever placed in service.

In summary, during the tax years at issue, the taxpayer purchased solar lenses from a corporation that was formed by the promoter and simultaneously leased the lenses back to a separate group of LLCs that were also formed and controlled by the promoter. The taxpayer then claimed ITCs and depreciation deductions in connection with the solar lenses he purchased to zero out his tax liability.4

IV. Tax Court Analysis on ‘Trade or Business’

A taxpayer is permitted to claim ITCs only under the relevant sections for depreciable property. Property is depreciable under section 167 when the property is either used in a trade or business or held for the production of income. Thus, a finding that the solar lenses are not depreciable would disentitle the taxpayer to the depreciation deduction and the ITCs.

The Tax Court relied on several principles regarding whether the taxpayer was engaged in a trade or business from prior decisions, including the following:

  • “To be engaged in a trade or business, the taxpayer must be involved in the activity with continuity and regularity and . . . the taxpayer’s primary purpose for engaging in the activity must be for income or profit.”5

  • To satisfy the trade or business requirement, the taxpayer must show extensive business activity over a substantial period.6

  • Sporadic activities do not rise to the level of a trade or business.7

  • Managing one’s own investments does not constitute a trade or business.8

  • Determining whether a taxpayer is engaged in a trade or business requires an examination of all the facts in the case.9

The taxpayer changed the description of the trade or business in Schedule C for the relevant tax years. The taxpayer claimed that he was engaged in a “solar energy” business for 2010-2012. This was then changed to “equipment rental services” for 2013-2014. Thus, the Tax Court’s trade or business analysis considered the characterization of each business separately.

A. Solar Energy Business, 2010-2012

For the solar energy business claimed in Schedule C of the taxpayer’s returns for 2010-2012, the Tax Court found that he failed to establish that he engaged in the solar energy business. In particular, the Tax Court noted the following:

  • The taxpayer did not possess the skills, education, or experience to conduct the business.

  • The taxpayer’s activities primarily consisted of: (1) writing checks to the promoters, (2) signing a few forms and documents each year, and (3) engaging in email correspondence with the promoters.

  • For the purported business, the taxpayer: (1) kept no business records, (2) received no gross income, and (3) reported no expenses apart from depreciation and legal fees he was unable to substantiate.

  • PFO Solar LLC, which was formed to engage in the purported business: (1) had no business records or employees, and (2) did not maintain its own bank account.

  • The taxpayer’s responsibilities concerning PFO Solar LLC merely involved: (1) renewing the LLC each year, (2) maintaining PDF copies of documents, and (3) trying to determine how many lenses to purchase.

  • The taxpayer had trouble explaining the nuts and bolts of the project and made observations that were contrary to someone engaged with continuity and regularity in a genuine trade or business. For example, the taxpayer admitted that “people ask me what it is specifically that they will be purchasing, and I don’t know,” and also commented that the solar equipment “looks a little like junk.”

The Tax Court also noted that the taxpayer did not use the lenses to conduct any business activity but leased the lenses back to an entity controlled by the promoter, who managed and maintained the lenses, leaving the taxpayer with “nothing to do.” The Tax Court specifically considered that any success of the activity would not be derived from the taxpayer’s own trade or business but from that of the promoter’s companies.

Putting aside questions regarding the extent (or lack) of the taxpayer’s involvement in the purported use of the lenses, the Tax Court also questioned whether the lenses were actually used for any trade or business activity. For example, the Tax Court commented that the taxpayer could not establish that he had ever seen the lenses at all, nor was there any evidence that his lenses were installed on a tower or even unloaded from a pallet.

Finally, the Tax Court found that the taxpayer’s primary purpose in purchasing the lenses was to benefit from tax savings, not to derive profit from the conduct of a genuine business enterprise. In doing so, the Tax Court adopted the definition of profit to mean “economic profit independent of tax savings.”10

B. Equipment Rental Business, 2013-2014

For the equipment rental business claimed on the taxpayer’s returns for 2013-2014, the Tax Court found that his course of conduct was entirely inconsistent with that of a person engaged in an equipment rental business. The Tax Court described the arrangement as what was basically a sale-and-leaseback transaction. The arrangement was problematic to the Tax Court for a multitude of reasons, including the fact that:

  • the taxpayer had agreed to purchase the lenses from the promoter with pre-drafted purchase and lease terms; and

  • the terms of the lease agreement provided that the lenses would be leased for a mere $150 per year, and the lessee never made a rent payment because the lenses never produced revenue.

In other words, the taxpayer was essentially leasing the lenses to the promoter’s affiliates for free, negating any intention of income or profit.

Concerning the taxpayer’s failure to establish any continuity and regularity of rental activity, the Tax Court noted that:

  • the taxpayer did not advertise or promote the rental business;

  • the taxpayer did not seek or have any customer other than LTB (the LLCs controlled by the promoter); and

  • the taxpayer never possessed any of the lenses, nor could he identify them at the site.

Finally, in a manner similar to the solar energy business, the Tax Court highlighted that:

  • the taxpayer showed no regular or active involvement in any rental activity;

  • the taxpayer supplied no evidence that PFO Solar had a bank account, books and records, business plans, or marketing strategies; and

  • the taxpayer kept no records of time logs, calendars, or diaries to show any time devoted to the business.

Based on all this, the Tax Court found that the activity plainly did not rise to the level of a trade or business and cited a decision11 that disallowed depreciation deductions when taxpayers had failed to show active involvement in a rental business.

V. Taxpayer’s Position on Appeal

The taxpayer now appeals the Tax Court’s findings and conclusions on virtually all the issues leading to the disallowance of the tax credits and depreciation deductions. This includes an appeal of the Tax Court’s finding that the taxpayer was not engaged in a trade or business.

The taxpayer appeals the Tax Court’s trade or business determination on two fronts. First, he contends that the Tax Court erred in disallowing his rental activity status as a trade or business solely because he wanted to (or did) obtain tax benefits. Instead, when considering whether the transaction was profitable or profit-oriented, the Tax Court ought to have taken tax benefits into account.

The taxpayer examines the history and the purpose of the ITC and argues that tax credits play a role to encourage activity and should not be disallowed in a transaction in which a taxpayer undertakes the type of activity the credit was intended to encourage. The taxpayer also cites a Ninth Circuit opinion,12 which held that a taxpayer was entitled to ITCs from a sale and leaseback of solar water heating equipment even though the transaction would not be profitable unless tax benefits were taken into account.

The taxpayer also relies on the Tax Court opinion in Cross Refined Coal,13 which followed Sacks to find economic substance even though the operation became profitable only after accounting for refined coal credits. Although not mentioned in the taxpayer’s appeal brief, that decision is still under appeal. Coincidentally, we predicted that the IRS’s appeal of Cross Refined Coal would likely be unsuccessful in our inaugural installment of Blue J Predicts.14

Second, the taxpayer argues that the activity of purchasing and leasing solar equipment is, by itself, sufficient to constitute a trade or business in the context of rental and leasing activity. The taxpayer relies on Cooper,15 which involved taxpayers who purchased energy equipment on a leveraged basis and leased the equipment back to a party related to the vendors. In Cooper, the Tax Court held that the taxpayer entered into transactions to make a profit and that the sales were bona fide multiparty transactions. Most importantly, the Tax Court in Cooper stated, “Once the equipment was leased to Coordinated, nothing further remained for petitioners to undertake.”

In the context of the Tax Court’s decision, the taxpayer here is essentially claiming that the Tax Court imposed too onerous a requirement for the extent the taxpayer is required to be involved for the activity to constitute a leasing trade or business. From the taxpayer’s perspective, elements of the activity — such as the leveraged purchase basis, the sale-leaseback arrangement, the claiming of deductions and tax credits for the full purchase price of the equipment, and the fact that installation of the equipment was undertaken exclusively by the lessee — are all not sufficient to preclude a finding of trade or business in the context of rental activity.

Now that we have an understanding of the taxpayer’s position on appeal, let’s examine how we can use machine learning to assist with evaluating the strength of the taxpayer’s position.

VI. Legal Technology Insights

While a deceptively simple test on its face, the question whether a taxpayer is engaged in a trade or business has been the subject of extensive litigation. There is a vast array of possible activities and a taxpayer’s varying extent of involvement in these putative trade or business activities. Advances in computing power and machine learning are particularly well suited to handle the sizable body of case law and the fact-intensive inquiry required to get insight into this legal question.

The facts and circumstances used in Blue J’s predictions are the kinds that judges refer to and rely on in their decisions. After a user inputs all the relevant facts of the case, Blue J’s model produces a prediction of how likely it is that a court would rule that a taxpayer is engaged in a trade or business based on those stipulated facts and circumstances. Blue J also discloses the degree of confidence it has in that prediction.

Blue J uses machine learning to build a model based on a data set of more than 700 decisions that consider whether a taxpayer is engaged in a trade or business. At a high level, most of the factors considered by Blue J’s algorithm concern information about (1) the type of activity; (2) whether the taxpayer engaged in businesslike practices; (3) the taxpayer’s expertise and expectations; (4) the activity’s income and loss history; and (5) the continuity, regularity, time, and effort spent on the activity. Collectively, these five broad categories of considerations are further split into indicia of trade or business, ultimately captured by 23 different factors.

A. Framing of Trade or Business on Appeal

Recall that the Tax Court opinion considered the trade or business determinations of the solar energy business and the equipment rental services business separately, based on the taxpayer’s different filings in 2010-2012 and 2013-2014, respectively. But rather than following the Tax Court’s format, the taxpayer frames the appeal by asking the Tenth Circuit to consider whether the taxpayer’s “rental activity is a trade or business.” By framing the activity under consideration as a rental activity, it may appear at first blush that the taxpayer is only appealing the Tax Court’s trade or business finding in relation to the equipment rental services business and not the solar energy business. However, the taxpayer’s appeal includes paragraphs from the solar energy business, suggesting that the taxpayer is appealing both findings, although this is not explicitly stated.

Perhaps the framing is a deliberate attempt by the taxpayer to seek a determination on whether the purported rental activities constituted any sort of trade or business, rather than being confined to the Tax Court’s framing of whether the taxpayer engaged in a solar energy business (which, admittedly, seems like a higher bar to meet). Thus, it remains to be seen whether the Tenth Circuit will follow the format of issues as outlined in the Tax Court’s opinion or whether it will adopt the taxpayer’s framing in making an overall determination on the trade or business issue.

Still, in the spirit of evaluating the strength of the taxpayer’s appeal, we will adopt the taxpayer’s framing and treat the taxpayer’s business as a single activity when considering whether the activity constituted a trade or business (despite the inconsistency in the taxpayer’s characterizations of the type of business against which it was seeking tax benefits).

B. Applying Machine Learning

As noted, the taxpayer contends on appeal that the Tax Court misapprehended the law by adopting the narrow definition of profit to include only “economic profit, independent of tax savings,” from the court in Surloff.16 Instead, the taxpayer implores the Tenth Circuit to follow the reasoning in the Sacks court, which held that “profits of the enterprise must be considered to include tax credits,”17 and thus to broaden the definition of profits to include tax benefits.

Blue J’s trade or business algorithm considers the profitability of an activity as a factor in a few ways. The algorithm first asks whether an activity has experienced at least one profitable year since inception and, if so, then inquires about the frequency of those profits and whether the profits were because of unusual or nonrecurring items.

When the Tax Court’s opinion was released, Blue J’s algorithm originally predicted with over 95 percent confidence that a court would rule that the taxpayer did not engage in a trade or business after considering all the factors. Blue J made this prediction based on the Tax Court’s finding that the taxpayer did not experience any profitable years since inception.

However, even if we adopt the most favorable version of the taxpayer’s appeal position to include tax benefits as part of profits and we assume that each of the tax years was profitable, Blue J’s algorithm predicts with 86 percent confidence that a court would still likely rule that the taxpayer did not engage in a trade or business. This suggests — contrary to the taxpayer’s assertion on appeal that the Tax Court disallowed the activity as a trade or business solely because he wanted to (or did) obtain tax benefits18 — that there are other strong factors present in the case that militate against a finding of trade or business. An examination of the Tax Court’s opinion supports this assertion, given that it took issue with a significant number of findings (summarized in the bulleted points earlier) aside from profitability and benefit-seeking behavior.

While holding all other factors constant, Table 1 illustrates the effect of profitability on Blue J’s prediction on whether a court is likely to find that the activity constitutes a trade or business.

Effect of Profitability on Likelihood Activity Constitutes Trade or Business

Taxpayer Profitability

Frequency of Profitability in Last Few Years

Blue J’s Evaluation

Not profitable


> 95% not trade or business



95% not trade or business



94% not trade or business



93% not trade or business



91% not trade or business



86% not trade or business

The algorithm’s revised (albeit less confident) prediction that a court would still find that a taxpayer did not engage in a trade or business — even if we assume full-time profitability — captures the reality that many other problematic aspects of the activity, aside from profitability, raise concerns about whether the activity is a trade or business. For example, the algorithm took issue with other factors, such as the fact that PFO Solar had no records or employees and no independent bank account, the taxpayer possessed no specialized knowledge or skills in the business, and the scope of the taxpayer’s activity was limited and infrequent, etc. Blue J’s algorithm attempts to quantify these risks by representing the totality of these factors in the form of a confidence number that is informed by how judges have behaved in similar circumstances in the past.

To summarize, the taxpayer asks the Tenth Circuit to reverse the Tax Court’s conclusion on trade or business based on the profits issue. However, Blue J’s algorithm suggests that even if the Tax Court erred in failing to consider credits as part of profits, this error alone is unlikely to be sufficient for the Tenth Circuit to reverse the Tax Court’s finding to conclude that the taxpayer was engaged in a trade or business in light of all the other factors.

This analysis should not be interpreted as a criticism of the taxpayer’s appeal strategy. Appellants’ decisions to prioritize one aspect of an appeal over another are often a reflection of constraints created by immutable facts or unfavorable findings of the Tax Court that are subject to a highly deferential standard of review. Rather, this is an example of when using tech-enabled solutions to engage in legal research can allow tax professionals to quantify risk and realistically assess the strengths and weaknesses of their positions.

C. The Role of the Attorney

Aside from the algorithm’s ability to dynamically quantify the significance of each factor in a given legal question, an attorney is faced with the even more challenging task of evaluating the likelihood that a court will accept a position or characterization advanced by a client. Blue J’s algorithm is unable to evaluate the sufficiency of a client’s evidence to support a particular characterization.

Although Blue J’s interface attempts to mitigate this problem by directing the user to how other courts have considered and applied the law as far as this factor is concerned, it remains the tax professional’s responsibility to exercise skill and judgment to make that determination. This is precisely why a party’s briefs and submissions mainly revolve around the characterization of a set of facts within a legal framework rather than disputing which factors ought to be more significant.

Let us consider the example of the profitability factor from the previous section. Even though profitability was not a determinative factor given the facts of this scenario, an attorney may encounter a file with a different set of facts in which profitability is indeed likely to be the swing factor. While the respondent’s briefs are not available at the time of this writing, the taxpayer is likely going to receive pushback on the characterization of tax benefits as profits. The taxpayer’s contention that profits ought to include tax benefits is susceptible to challenge in several ways. For example, the respondent could argue the following:

  • The Ninth Circuit in Sacks did not explicitly find that the taxpayer carried on a trade or business, only that the sale-leaseback arrangement was not a sham.

  • Cross Refined Coal is distinguishable on the basis that the activity of coal refinement is inherently unprofitable and that the credits sought were production credits rather than credits received by owning an eligible asset.

  • Cross Refined Coal and Sacks are further distinguishable because the taxpayers bore the risk of loss or failure, whereas the taxpayer in Olsen did not bear any risk because the down payments for the lenses were refundable if production quotas were not met and the balances for the lenses were not paid.

  • The discussion of profit is a red herring, because the taxpayer’s activity did not merely lack profit — it lacked gross income of any sort.

  • Cases involving the consideration of tax credits when evaluating profits involved credits that supplemented income to make the activity profitable, not credits that consistently constitute 100 percent of the taxpayer’s receipts from the activity for all tax years.

  • Not only were the taxpayer’s activities not profitable, but the underlying business operated by the promoter was inherently problematic.

Some of these responses have the potential to undermine the characterization of profitability in a way that falls entirely outside the scope of the algorithm’s predictive capabilities. Machine learning can only identify the factors that matter and tell you how much they matter based on what has come before. This dynamic creates an opportunity for synergy between technology and the tax professional, in which machine-learning tools can identify the significance of factors to focus on most while the tax professional crafts arguments in support of those characterizations once the most significant factors are identified.

D. Research Using Cataloging Method

One of the underappreciated aspects of machine-learning and predictive model development is the process by which the data are collected to train these models. This involves translating each decision on a legal issue from raw unstructured case data into structured data that can then be used to train a model and develop the algorithm. In other words, every single case on a legal issue that is used to develop a model must be assessed on all the factors considered by the algorithm. This process results in the creation of a comprehensive catalog of all relevant factors and metadata for every case in our system.

This presents an interesting use case that can be illustrated by the taxpayer’s second set of trade or business arguments on appeal. Recall that the taxpayer argued that the Tax Court imposed too onerous a bar on activities for leasing. The taxpayer did so by identifying many similarities between the taxpayer’s activities and those in Cooper, which also involved taxpayers who purchased energy equipment on a leveraged basis and leased the equipment back. The Tax Court in Cooper ruled in favor of the petitioners, and the taxpayers here in Olsen are merely asking for the same treatment.

In the cataloging process used to develop Blue J’s trade or business module, Cooper was one of the more than 700 cases used to train Blue J’s predictive model. After running the facts of Cooper through Blue J’s model, the algorithm predicted that Cooper was engaged in a trade or business with a high degree of confidence. This may seem surprising because Blue J also predicted that Olsen was not engaged in a trade or business with a high degree of confidence. However, given the extent to which the taxpayer asserts that the cases are similar, perhaps the cases are similar in some respects but different on factors that matter enough to flip the prediction.

When reading Cooper on the Blue J platform, the cataloging method gives users a bird’s-eye view of the case and can help users identify factors that drive a wedge between the two cases to understand why they lead to different evaluations, even if two cases are superficially similar.

Blue J’s platform allows a user to identify that while the nature of the activity as a rental/leasing activity is similar, the facts of the two differ on other meaningful factors. For example, unlike in Olsen, the petitioner in Cooper: (1) had a written business plan, (2) kept contemporaneous substantiated records, (3) followed money-making practices of professionals in the field, (4) made changes in the activity’s methods and processes that improved profitability, (5) could demonstrate an appreciation in the asset’s value, (6) experienced at least one profitable year since inception, (7) generated income more frequently, and (8) hired persons to carry on the activity, etc. This suggests that the facts of Cooper might not be as similar as originally believed, especially on factors that matter to courts when it comes to a trade or business determination.

Overall, this ability to obtain a high-level view of the 23 factors that drive each case allows a party who cites a case to evaluate the similarities between a case and allows a party responding to a cited case to identify distinguishing factors when crafting responding arguments. Moreover, the predictive capabilities of the models also allow parties to surface which distinguishing factors matter most to best focus their efforts when advancing their case.

VII. Conclusion

Using a machine-learning model can provide valuable insight, especially in legal areas with a sizable body of case law and a fact-intensive inquiry. Our analysis reveals that the taxpayer’s chances of successfully appealing the Tax Court’s trade or business determination in Olsen are not high, even if we entirely accept the taxpayer’s characterization of the issues. While advances in computing power have come a long way, they are not ready to replace human decision-making, and tax practitioners remain in the driver’s seat when it comes to novel arguments and clever characterizations. Still, tax practitioners can benefit from speeding up the process of determining which arguments to focus on and leverage Blue J’s unique cataloging method to accelerate research and compare cases that may seem similar on their surface but meaningfully differ in ways that are not immediately obvious.


1 Benjamin Alarie and Bettina Xue Griffin, “Using Machine Learning to Crack the Tax Code,” Tax Notes Federal, Jan. 31, 2022, p. 661.

2 Olsen v. Commissioner, T.C. Memo. 2021-41.

3 United States v. RaPower-3, 960 F.3d 1240 (10th Cir. 2020).

4 Many facts that are not relevant for the determination of whether the taxpayer was engaged in a trade or business have been omitted (including various details involving the tax shelter scheme and deficiencies in the promoter’s operation of the solar business).

5 Commissioner v. Groetzinger, 480 U.S. 23, 35 (1987).

6 Snyder v. United States, 674 F.2d 1359, 1364 (10th Cir. 1982).

7 Groetzinger, 480 U.S. at 35.

8 Higgins v. Commissioner, 312 U.S. 212, 216 (1941).

9 Id. at 217.

10 Surloff v. Commissioner, 81 T.C. 210, 233 (1983).

11 Jafarpour v. Commissioner, T.C. Memo. 2012-165.

12 Sacks v. Commissioner, 69 F.3d 982 (9th Cir. 1995).

13 Cross Refined Coal LLC v. Commissioner, No. 19502-17 (2019) (bench opinion).

14 Alarie, Griffin, and Yan, “An Unprofitable Pretax Venture Can Still Be a Partnership,” Tax Notes Federal, June 21, 2021, p. 1951.

15 Cooper v. Commissioner, 88 T.C. 84 (1987).

16 Surloff, 81 T.C. at 233.

17 Sacks, 69 F.3d at 991.

18 Olsen v. Commissioner, No. 21-9005 (10th Cir. 2021); Appellant’s Brief, at 20 (Jan. 13, 2022).


Copy RID