Chemoil: Economic Substance, Tax Credits, and Unprofitable Ventures

I. Introduction

Advances in machine learning applied to tax law can help mitigate some of the uncertainty around the application of antiavoidance doctrines, such as the economic substance doctrine. As we will show, machine learning can also inform tax litigation strategy in cases involving the application of the economic substance doctrine.

The economic substance doctrine permits courts to examine whether the taxpayer’s use of tax benefits is consistent with congressional intent, even when all the technical statutory and regulatory requirements for claiming a tax benefit are satisfied. If it applies, the economic substance doctrine denies tax benefits arising out of transactions that fail to effect a meaningful change to the taxpayer’s economic position.

The judicial examination of a transaction’s economic substance has roots in the seminal 1930s decision in Gregory¹ and was developed further in the 1970s with the Supreme Court’s holding in Frank Lyon.² Motivated by circuit splits in its application, Congress codified the economic substance doctrine in 2010 with the enactment of section 7701(o). In general, a codified, more rules-based approach to tax legislation tends to allow taxpayers to compute their taxable income with greater predictability. Nevertheless, judicial antiavoidance doctrines can serve to close gaps between the legislative intent behind specific tax provisions and the literal requirements of tax legislation. The codification of the economic substance doctrine is one example of ongoing efforts to balance the trade-offs between greater predictability and equitable application of the law.

In this month’s Blue J Predicts column, we examine the parties’ economic substance arguments in the ongoing litigation in Chemoil.³ The case involves tax refunds denied by the IRS for excise tax credits related to alcohol-fuel mixtures. The topic discussed in Chemoil bears a close relationship to last month’s installment of Blue J Predicts,⁴ in which we correctly predicted that the D.C. Circuit in Cross Refined Coal⁵ would rule for the taxpayer, holding that a business venture that was guaranteed to be unprofitable pretax (and became profitable only after tax credits) could still be considered a bona fide partnership for federal income tax purposes. Similarly, the Chemoil dispute provides an opportunity to explore the relationship between the economic substance doctrine and unprofitable transactions that are rendered economically viable by tax credits.

Chemoil raises a particularly interesting point of contention. Transactions that generate specific excise tax credits (which are designed to encourage otherwise unprofitable activities) may appear to run afoul of the economic substance doctrine, which considers whether transactions have independent economic significance setting aside the tax benefits derived from the transactions. Thus, there is a tension between the congressional intent to provide an incentive for otherwise unprofitable activities and the congressional intent behind the economic substance doctrine, which, by its nature, denies tax benefits that arise out of transactions with no economic effects apart from their federal tax implications. Moreover, if a taxpayer cannot avoid application of the economic substance doctrine to the relevant transactions, the dispute raises the further question of whether transactions whose profitability is derived solely from excise tax credits should still be considered as having economic substance because of the nature of how these tax credits operate to encourage otherwise economically unviable yet ostensibly congressionally endorsed activities.

Unlike an appeals decision, parties before a court of first instance do not have the benefit of presumed findings of fact or conclusions of law. Instead, they must contend with uncertainties regarding sufficiency of the evidence and different characterizations of material facts. This makes understanding the legal significance of each factor more important and allows tax practitioners to focus their arguments on characterizations likely to matter most to courts.

Chemoil Corp.’s responding submissions were not available at the time of this writing. Based solely on the characterization of facts in the government’s motion for summary judgment, Blue J’s economic substance model predicts with 83 percent confidence that a court would find that the transactions in question would not have economic substance. Accordingly, the machine-learning analysis in this installment focuses on exploring the potential positions that Chemoil could adopt in response to this economic substance analysis. In doing so, we investigate the likelihood of various outcomes depending on whether the district court adopts the government’s characterization of facts, what we expect to be the taxpayer’s characterization of facts, or some combination in between.

II. Background and History

The dispute is a tax refund suit filed by Chemoil against the government to recover more than $6 million in excise tax credits that have been disallowed by the IRS. Not only has the IRS denied Chemoil’s claim of alcohol-fuel mixture credits under section 6426(b), but it also assessed a penalty of more than $4 million under section 6675, with interest, for what it characterizes as excessive tax credits claimed without reasonable cause.

The subject transactions involved the sale of large volumes of a mixture of alcohol and fuel for which the credits were claimed. Section 6426 was enacted to create incentives, in the form of an excise tax credit, for the production and sale of alternative fuel mixtures. The government challenges the refund on grounds that the transactions lacked economic substance and Chemoil failed to satisfy the statutory requirements for claiming the credit.

On July 8 Chemoil filed a motion for partial summary judgment asking the U.S. District Court for the Southern District of New York to find that:

• the economic substance doctrine should not apply to the transactions involving the claimed credits in dispute or, in the alternative, find that excise tax benefits can be considered in determining whether the transactions at issue had economic substance;

• Chemoil’s alcohol-fuel mixture could have been used as fuel and thus satisfied the requirement under section 6426(b)(3); and

• the IRS failed to satisfy a requirement to levy a section 6675 penalty.⁶

The U.S. attorney’s office filed a response in opposition to Chemoil’s partial summary judgment motion on September 16.

The government initially filed a motion for summary judgment against Chemoil on July 8, but the materials were placed under temporary seal, and an amended version was filed August 10. There, the government seeks summary judgment, asking the court to find that:

• Chemoil is not entitled to the tax credits at issue because (1) the transactions lacked economic substance and (2) Chemoil did not satisfy several statutory requirements for claiming the credits; and

• Chemoil is not entitled to a refund of the penalty.

Chemoil has not yet filed a response to the government’s motion.

The following discussion focuses on the economic substance component of the dispute in Chemoil’s partial summary judgment motion and the government’s response. However, it will become evident that findings regarding the parties’ arguments under section 6426 (which concern whether Chemoil sold the alcohol-fuel mixture for use as a fuel and thus qualified for the excise tax credit) may affect the economic substance analysis.

III. Summary of the Parties’ Positions

The parties disagree on virtually all aspects of the economic substance analysis, including (1) whether the economic substance doctrine should apply; (2) if so, which version of the doctrine should apply; and, finally, (3) how the economic substance doctrine should apply.

A. Should the Doctrine Apply?

Chemoil asks the court to find that the economic substance doctrine is not relevant to the transactions at issue. In support of its position, Chemoil cites the Joint Committee on Taxation’s technical explanation (blue book) of the revenue provisions that introduced the codification of the economic substance doctrine under section 7701(o), which states:

The explanation includes examples of various credits, including the low-income housing credit, the renewable energy and refined coal production credit, the new markets credit, the rehabilitation credit, and the renewable energy credit. Chemoil argues that the section 6426(b) credit, which seeks to encourage the production and use of alcohol fuels, is no different from the noted examples. Finally, Chemoil quotes the district court in Altria,⁸ which held: “The economic substance doctrine simply has no application if it is clear that a claimed deduction is within the intent of a provision of the Code.” It also quotes the Ninth Circuit in Sacks,⁹ which stated: “Absence of pre-tax profitability does not show ‘whether the transaction had economic substance beyond the creation of tax benefits,’ . . . where Congress has purposely used tax incentives to change investors’ conduct.”

On the other hand, the United States, in its opposition to Chemoil’s motion for partial summary judgment, maintains that the economic substance doctrine almost always applies. In support of that position, the government relies on the Second Circuit’s decision in BNY Mellon¹⁰ for the proposition that there are no “categorical exemptions” from the application of the economic substance doctrine and that the doctrine exists to give courts a “second look” to ensure that a particular use of tax benefits complies with Congress’s purpose in creating that benefit. The government also relies on the Federal Circuit’s decision in Alternative Carbon,¹¹ which applied an economic substance analysis to disallow a tax credit for alternative fuel mixtures under the section 6426 provision at issue in Chemoil. The Federal Circuit rejected the taxpayer’s argument that the alternative energy tax incentives should be excluded from the economic substance doctrine.

The government also characterizes the blue book statement referenced by Chemoil as a stray statement that deserves minimal weight. It contends that blue book statements are not a legitimate tool of statutory construction because they are commentaries on recently passed tax laws and thus do not inform the decisions of lawmakers who voted in favor of the law. Moreover, the government argues, the blue book doesn’t instruct courts to ignore the economic substance doctrine just because the taxpayer is trying to claim an energy tax credit.

Finally, the government asserts that Chemoil’s argument about the blue book fails even on its own terms because Chemoil’s activity was not the type that the credit was intended to encourage. Although the credit was intended to provide an incentive to add alcohol to gasoline as an alternative fuel, the government characterizes Chemoil’s alcohol-fuel mixture as ethanol with a “splash” of gasoline to generate tax credits. In other words, the government contends that Chemoil is trying to claim a credit that was designed to encourage the addition of an alternative fuel (such as alcohol) to gasoline by instead adding a negligible amount of gasoline to alcohol, such that the addition of gasoline did not affect the characterization of the end product as ethanol under applicable specifications. The government states that the counterparties (that purchased the mixture) were concerned only that Chemoil’s tax-motivated addition of gasoline not exceed the relevant ethanol specifications’ limits on the amount of gasoline that could be added.

The government’s characterization of Chemoil’s mixture as ethanol with a dash of gasoline also raises the question whether the mixture was produced “for use as a fuel” as required by section 6426 to claim the credit, since the mixture could not be used as fuel in its current form. There is a disagreement of statutory interpretation as to whether “for use as a fuel” requires the mixture to be usable as fuel or for the mixture to be usable as fuel downstream after undergoing further processing.

B. Which Economic Substance Doctrine?

Chemoil argues that if the economic substance doctrine does apply, the statutory economic substance doctrine under section 7701(o) alone should apply. In its submissions, Chemoil explains the history behind the codification of the doctrine, which was necessitated out of a circuit split over the way the test was applied. It concludes that accordingly, as a matter of law, the codified economic substance doctrine, rather than the common law doctrine, applies to transactions since section 7701(o) was enacted on March 30, 2010. Moreover, Chemoil asserts that the codification of the economic substance doctrine in section 7701(o) effectively supersedes the common law insofar as how the doctrine is applied.

Section 7701(o)(1) states that a transaction shall be treated as having economic substance only if:

A closer examination of Chemoil’s later submissions reveals that it prefers the application of the statutory economic substance doctrine because “Federal income tax effects” refers to taxes imposed under subtitle A of the IRC. Therefore, the tax benefits from excise tax credits under section 6426, which do not fall under subtitle A, can arguably be considered in determining whether the transactions at issue had economic substance because the credits have no “Federal income tax effects.”

The government, however, is quick to address this distinction and advocates for the application of the common law economic substance doctrine rather than the codified doctrine. The government notes that the statutory version doesn’t apply because section 7701(o) applies only to income taxes under subtitle A of the code. Section 7701(o)(5)(A) defines the term “economic substance doctrine” to mean: “the common law doctrine under which tax benefits under subtitle A with respect to a transaction are not allowable if the transaction does not have economic substance or lacks a business purpose.” (Emphasis added.)

The government asserts that the partial codification does not affect the determination of whether the economic substance doctrine is relevant to a transaction and that, as stated in section 7701(o)(5)(c), that determination “shall be made in the same manner as if [section 7701(c)] had never been enacted.” Accordingly, the common law economic substance doctrine, as elucidated in BNY Mellon,¹² ought to apply, according to the government.

The common law doctrine considers whether the taxpayer had an objectively reasonable expectation of profit from the transaction, apart from tax benefits. Tax benefits under the common law test would thus include the tax credits at issue under section 6426 (rather than being restricted to tax benefits under subtitle A under the statutory version). The government maintains that this is a sensible interpretation because the income tax under subtitle A, to which the doctrine applies under the statute, also excludes the consideration of tax benefits under subtitle A.

C. Should the Credits Affect Profitability?

As discussed, Chemoil contends that the application of the statutory economic substance doctrine permits the consideration of excise tax credits when determining whether a transaction had economic substance. The government asserts that the common law economic substance doctrine applies, which would exclude the consideration of excise tax credits in that determination.

Outside the issue of which version of economic substance ought to apply, Chemoil relies on several Second Circuit cases, including Long Term Capital Holdings,¹³ that say that an absence of economic substance beyond the creation of tax benefits is not necessarily sufficient to conclude that the transaction was a sham. However, those cases don’t consider the effect of excise tax credits. So Chemoil relies on two cases outside the Second Circuit in greater detail.

First, Chemoil relies on the Ninth Circuit’s decision in Sacks,¹⁴ which reversed the IRS’s victory in the Tax Court. The court of appeals found that “tax advantage such as Congress awarded for alternative energy investments is intended to induce investments which otherwise would not have been made” and that “if the government treats tax-advantaged transactions as shams unless they make economic sense on a pre-tax basis, then it takes away with the executive hand what it gives with the legislative.”¹⁵ Thus, the Sacks transactions’ lack of pretax profitability was not evidence of abuse but rather evidence that the transactions were consistent with Congress’s purpose.

Second, Chemoil relies on the Tax Court’s opinion in Cross Refined Coal,¹⁶ which was affirmed by the D.C. Circuit in August, after Chemoil filed its materials. The Tax Court in Cross Refined Coal allowed the refined coal tax credit because the market, unassisted by credits, was not producing refined coal on the scale Congress thought to be beneficial. Accordingly, the incentive offered by after-tax profits was the precise purpose of the credit, so the absence of pretax-credit profitability did not mean that the activity lacked economic substance — rather, that fact was the reason for the credits.

The government distinguishes the cases relied on by Chemoil. First, it argues that Sacks is distinguishable because the taxpayer paid fair market value for the panels, had a genuine business purpose for installing solar water heaters, and demonstrated the possibility of long-term profits (aside from tax benefits).¹⁷ Unlike that taxpayer, Chemoil entered into contracts with significant pretax losses, with no reasonable prospect of profitability, and lacked any legitimate business purpose other than the creation of tax benefits.

Second, the government distinguishes the D.C. Circuit’s decision in Cross Refined Coal (which was available by the time its materials were filed), which had a “practical economic effect” and involved an activity that Congress specifically sought to encourage. By contrast, Chemoil had not shown it had any legitimate purpose in buying ethanol (sometimes at above-market prices), adding gasoline that its counterparties did not want, and selling the ethanol at below-market prices. In short, the government argues that Chemoil engaged in “essentially wasteful transactions that are beyond any reasonable understanding of the activity Congress sought to promote” by engaging in these transactions in an attempt to create tax benefits. In the government’s view, there is no reason to believe that Congress sought to encourage mixtures that were sold to fulfill contracts for the purchase of fuel-grade ethanol rather than an alcohol-fuel mixture that met the same specifications as if Chemoil had added no gasoline at all and created a product that could not be used as fuel without further mixing or processing.

Remarkably, Chemoil is not seeking a declaration from the court to find that the transactions at issue have economic substance; instead, it seeks to frame arguments around that characterization by asking the court to include the benefits derived from the excise tax credits in the profitability element of the economic substance test. The government, on the other hand, not only wishes to exclude those benefits but seeks a ruling, in its own separate motion for summary judgment, that the transactions be disregarded because they lack economic substance. This asymmetry means that without access to Chemoil’s responding briefs to the government’s motion, only the government’s characterization of the facts is available at the time of this writing.

IV. Applying Machine Learning

Recognizing the complex history of the economic substance doctrine, Blue J’s machine-learning algorithm is trained on a database of 250 cases that invoke the doctrine. The database includes common law cases that predate codification. At a high level, most of the facts and circumstances of these cases involve information about the origin of the transaction, the structure of the transaction, the economic effects of the transaction, the effects of the transaction other than profit, and the taxpayer’s risk level. Collectively, those broad categories are represented by 25 different facts and circumstances that are used to analyze new economic substance scenarios.

Among those 25 factors, Blue J’s data reveal three that are particularly salient for courts and appear to be influential in determining the results in economic substance doctrine cases: (1) pretax profits; (2) transaction costs; and (3) transacting at values other than FMV. Although only the government’s position on economic substance was available at the time of this writing, Chemoil’s briefs suggest that it would dispute the government’s characterization of transaction costs and whether the excise tax credits should be considered pretax profits.

Before we examine how the parties’ positions differ on these factors, we outline what they reflect.

• Pretax profits: Courts have demonstrated an interest in whether a taxpayer expects a pretax profit from a transaction. The expectation of a pretax profit, particularly if higher than the expected tax benefit, makes it empirically more likely that a court will conclude that a transaction has economic substance.

• Transaction costs: Courts are interested in whether a taxpayer could have fulfilled the transaction’s ostensible business purpose with lower transaction costs (for example, commissions and professional fees). If there were less expensive ways to achieve the alleged business purpose, it is less likely that a transaction will be found to have economic substance.

• Transacting at FMV: Courts also subject transactions to enhanced scrutiny when property does not trade at FMV, since it can be inconsistent with “common sense from an economic standpoint.”¹⁸

Based solely on the characterization of facts in the government’s motion for summary judgment, Blue J’s algorithm predicts with 83 percent confidence that a court would find that the transactions in question do not have economic substance (see Scenario 1 in the table). It is interesting that among the list of decisions with similar factors, Blue J’s economic substance algorithm recommended Alternative Carbon¹⁹ as a substantially similar case factually. This is consistent with the government’s heavy reliance on the court’s analysis in Alternative Carbon and its attempt to characterize the facts in Chemoil as mirroring those in that case.

The parties hotly contest whether the benefits from the excise tax credits may be considered in the application of the economic substance doctrine. Recall that Chemoil contends that to the extent that economic substance is relevant to the analysis, the statutory economic substance doctrine is the correct version of the doctrine to apply, and it thus relies on the statutory gap in section 7701(o) that excludes taxes imposed under subtitle A (and not the excise tax credits under a different subtitle). The government, on the other hand, contends that the common law economic substance doctrine applies, which means that all tax benefits, including the excise tax credits at issue, would be excluded in considering whether the transactions had economic substance.

If Chemoil successfully convinces the court to adopt its position to consider the economic effect of receiving excise tax credits in the transaction, the algorithm would still predict that the transactions in question would be found to lack economic substance, albeit at a lower confidence level of 68 percent (see Scenario 2 in the table). As such, this factor alone would still be insufficient for Chemoil to establish that its transactions had economic substance, though it serves as an improvement for the taxpayer from the algorithm’s base case of 83 percent confidence.

A second major point of contention between the parties is whether the addition of gasoline to alcohol served any purpose other than to generate tax credits. The government claims that the addition of gasoline to alcohol was intended to be immaterial to the sale and use of the product as ethanol. Instead, the mixtures were sold to fulfill contracts for the purchase of fuel-grade ethanol, not an alcohol-fuel mixture, such that the end product met the same specifications as if Chemoil had added no gasoline at all. Thus, any amount spent on adding gasoline, in the government’s view, is an unnecessary transaction cost that was not required to achieve the transaction’s stated purpose. Therefore, the government claims that the transactions did not involve beneficial activity that would have been uneconomical without the credits.

However, if Chemoil can demonstrate that the modifications made to the alcohol served a function outside of generating tax credits, it may be able to argue that the transactions could not have been fulfilled for a lower amount to achieve a stated business purpose (provided that the business purpose existed).

Should Chemoil be successful on this point of contention alone (while being unable to persuade a court to include excise tax credits in determining profitability), the algorithm would still predict that the transactions will be found to lack economic substance, albeit with a lower confidence level of 51 percent (see Scenario 3 in the table), making this an extremely close case. This would still be insufficient for the algorithm to conclude that it is more likely than not that the court would find that the transactions had economic substance. However, if Chemoil succeeds on both points of contention — convincing a court to include excise tax credits in determining profitability and demonstrating that the modifications to the alcohol served a functional purpose in line with congressional intent — the algorithm would predict, with 76 percent confidence, that the transactions do have economic substance (see Scenario 4 in the table).

Finally, the government argues that Chemoil purchased ethanol at above-market prices, paid a third party to add small volumes of gasoline to the ethanol, and then resold the resulting mixture at below-market prices. These assertions are based on contemporaneous market values computed by the government’s expert economist. This suggests that the contract prices, which were untethered to market prices, are additional pieces of evidence that Chemoil had no expectation of pretax profit.

Chemoil’s partial summary judgment briefs do not contain an explanation for purchasing alcohol at above-market prices or reselling resulting mixtures at below-market prices. This is understandable because in its partial summary judgment motion, Chemoil seeks only clarification on whether the economic substance doctrine applied and whether the excise tax credits should be included in an economic substance determination; it does not seek an economic substance determination in that pleading.

But if, in addition to succeeding on the two other points of contention, Chemoil or its experts can show that the purchase of ethanol and resale of the mixture were transacted at market prices, that would further bolster Chemoil’s case, and the algorithm would predict, with 82 percent confidence, that the transactions have economic substance (see Scenario 5 in the table). However, in the absence of success on other factors, this factor alone would not make a significant difference from the government’s base case.

Effect of Factors

Scenario	Pretax Profits	Could Transaction Costs Be Lower?	Non-FMV Transactions?	Predicted Taxpayer Success Rate on Merits
1	No	Yes	Yes	17% (83% no economic substance)
2	Yes	Yes	Yes	32% (68% no economic substance)
3	No	No	Yes	49% (51% no economic substance)
4	Yes	No	Yes	76% (24% no economic substance)
5	Yes	No	No	82% (18% no economic substance)

From a review of the results, it is clear that the algorithm does not assign static weights to individual facts and circumstances but rather adapts those factors dynamically based on how the other factors have been interpreted by the courts. The results also show that Chemoil faces a significant uphill battle: For it to be more likely than not that a court will find that its transactions had economic substance, Chemoil will have to win most, if not all, of its characterization disputes.

An examination of the three factors outlined in the table is just one example of how practitioners can use a machine-learning algorithm to test the range of possibilities based on a given scenario. It is important to note that the above range does not represent the entire spectrum of possibilities, especially considering that Chemoil’s full articulation of its defense that the transactions had economic substance was unavailable at the time of this writing. Moreover, using an algorithm to quantify the impact of disagreements between the parties illustrates the significance of marshaling the appropriate evidence to support a characterization, and it conserves resources and limited space to advance arguments and characterizations that matter most to courts.

V. Conclusion

We began our consideration of the issue by exploring the parties’ positions on the ground rules of whether economic substance applies, which version of economic substance applies, and how economic substance applies. Although this involved a lengthy analysis, the quantification of the risk of losing on a particular characterization by Blue J’s algorithm reveals that the parties are certainly well-served in engaging in a hotly contested dispute over how, if at all, the generation of tax credits ought to be considered in the economic substance analysis.

Even though the case law confirms that pretax profits are not necessary for a transaction to have economic substance, it is entirely understandable why Chemoil tries to claim that the doctrine should not apply to tax credits designed to encourage otherwise unprofitable activity. Chemoil would naturally benefit from avoiding the risk of having its transactions invalidated on the basis of lacking economic substance. Moreover, Chemoil’s framing of the economic substance doctrine as one that should take into account excise tax credits (when considering the phrasing of the statutory economic substance doctrine) positions it to argue that pursuing tax credits is a legitimate enterprise when done in a manner that is consistent with congressional intent. On the other hand, the government recognizes this and mounts significant resistance to each characterization while reiterating the facts it regards as likely leading to the application of the doctrine to the situation at issue.

Finally, although Chemoil has not yet had an opportunity to respond to the government’s summary judgment motion, Blue J’s economic substance doctrine model highlights that central to Chemoil’s prospects in the litigation is the ability to demonstrate that its addition of gasoline to alcohol to create mixtures was not done solely to qualify for the tax credit. Indeed, the analysis shows that Chemoil’s case will be improved significantly if it can demonstrate that the addition meaningfully changed the composition of the mixture in a way that aligns with congressional intent to increase the use of alternative fuels.

Fundamentally, because this is a contest between conflicting congressional intentions, it is difficult to predict how the case will be resolved. If the government’s characterization of the facts prevails, our machine-learning model points to a likely government win. On the other hand, if Chemoil is able to successfully counter the government’s characterizations, the case could quite reasonably go the other way. Despite the difficulty of predicting the outcome in this case, one thing we can say with confidence in light of our analysis is that insights from machine learning can be an important contributor to tax litigation strategy.