Paul L. Caron

Wednesday, May 4, 2022

Blue J Predicts With 86% Confidence Debt-Equity Decision In Tribune Media

Benjamin Alarie (Osler Chair in Business Law, University of Toronto; CEO, Blue J Legal) & Kathrin Gardhouse (Senior Legal Research Associate, Blue J Legal), The Debt-Equity Distinction and Tribune Media, 175 Tax Notes Fed. 593 (Apr. 25, 2022):

Tax Notes Federal (2020)In this article, Alarie and Gardhouse use the Blue J debt-equity predictor to analyze part of the Tax Court’s recent decision in Tribune Media [v. Commissioner, T.C. Memo. 2021-122 (Oct. 26, 2021)]. ...

Common law debt-equity characterization depends on the synthesis of more than a dozen factual and circumstantial elements. In real-world situations, with so many considerations in play, ambiguity is endemic. The threshold challenge for taxpayers, the IRS, and, ultimately, the courts is to determine the most appropriate characterization for a given financing, all things considered.

This is particularly difficult to do in cases that are close calls, in which there are balanced sets of factors alternately favoring debt and equity. Those cases often lead to judicial squinting to identify distinctions. This can magnify slight differences that, in the ordinary course, would unlikely be influential, let alone dispositive. Indeed, when there is a balanced set of factors in play, it can be difficult to reach reliable conclusions and produce compelling reasons. ...

In cases in which a difficult judgment must be made with reasonably balanced factors, it can be worthwhile to garner the “wisdom of crowds” to base one’s analysis on the entirety of the case law. Fastidiously collecting training data, and identifying the facts and circumstances of past cases along with the resulting debt or equity characterization, provides a data-rich foundation to train a machine-learning model to reliably and accurately assess the likelihood that a decision-maker would reach a characterization of debt or equity.

Blue J has done just that by assembling a detailed data set of debt-equity decisions from 1956 on. The Blue J debt-equity model yields 95.6 percent agreement with the decisions of the courts. It has been thoroughly back-tested against historical case law and, for the past few years, has been making accurate predictions of new debt-equity cases as they are decided. It is being used as a teaching tool to inform debt-equity analyses in leading university tax law courses and programs. Practitioners increasingly leverage Blue J’s debt-equity predictor to produce evaluations of the strength of novel situations involving new variations of facts and circumstances, many of which have never been directly judicially tested.

To illustrate the insights possible with machine learning, we use the Blue J debt-equity predictor to analyze part of the decision recently rendered in Tribune Media. Tax Court Judge Ronald L. Buch discussed and assessed the 13 Dixie Dairies factors in Tribune Media, analyzing whether the obligation ought to be properly characterized as debt or equity. Tribune Media was not a particularly close call, as the court (and Blue J’s model, with 86 percent certainty) concluded that the obligation in question was an advance on account of equity, with seven factors pointing in the direction of equity, three factors weighing toward debt, and three neutral factors. But as any tax lawyer with experience in analyzing the debt-equity distinction will remind us, an accurate assessment on the merits is not a simple matter of counting factors on each side of the ledger.

Indeed, analysis leveraging the Blue J debt-equity model shows that a different position on only two of the Dixie Dairies factors (one of which was characterized by the court in Tribune Media as neutral) would have flipped the outcome of the debt-equity characterization (with 77 percent confidence). This would not have been a long shot; on these two factors, the taxpayer had a prima facie reasonable prospect of success. However, the text of the decision does not disclose that these two factors could have made such a difference. In fact, as we will explore, there are curious differences between the decision’s determination of the factors’ significance and the importance suggested by Blue J’s model based on the accumulated wisdom from the decisions of judges in past debt-equity cases. We attribute this gap, at least in part, to the rhetorical challenge of producing compelling reasons in close cases. ...

We can conclude from the findings we have set out that a court’s explicit guidance about the relevance of a legal test’s factors should be received with caution. We are more inclined to confidently follow what judges do in terms of placing weight on specific factors, rather than what judges say they might do in terms of assigning weight to factors. Machine-learning tools can leverage an extensive database of previously decided cases and produce accurate and reliable predictions of how courts will decide new debt-equity cases. The main advantage of using machine learning is that it can overcome the rhetorical challenge that courts face in producing compelling reasons in borderline cases. Human intuition and experience alone are not able to reliably reveal the precise role factors in a legal analysis have played in hundreds of previously decided cases. Machine learning can. Human experts then serve an indispensable role in leveraging machine-learning insights for even more powerful tax advice. These ongoing machine-learning developments promise to continue to materially improve analysis, planning, and dispute resolution in the debt-equity context.

Prior TaxProf Blog coverage:

Scholarship, Tax, Tax Analysts, Tax Scholarship | Permalink