<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>Every taxable arrangement is subject to an anti-abuse test. Abusive arrangements are treated as not valid for tax purposes, which is similar to the treatment of artificial arrangements in civil law. The European Union has introduced in its Anti-Tax Avoidance Directive a general anti-abuse test which must be transposed into the domestic laws of Member States. Such a test has its inner structure, consisting of an elimination and requalification stage, while the elimination stage entails genuineness and a tax benefit test. The general anti-abuse test has a great potential (or scalability when speaking in the language of start-ups) of being automated and integrated into different legal application processes (such as taxpayer self-assessment systems, transactions certified by public notary or merger and acquisition deals) to discover debt push down abuses or other arrangement structures which may have abusive content. While the best method for create a reliable algorithm is a decision tree type model, the inner ambiguity of the general anti-abuse test prevents using the full benefits of automation of tax laws. The purpose of this article is to design a decision tree type model for the test and address the main challenges of such a model, both from the perspective of the clarity of concepts and the quality of input information such an engine would use.</p>
</abstract>

Every taxable arrangement is subject to an anti-abuse test. Abusive arrangements are treated as not valid for tax purposes, which is similar to the treatment of artificial arrangements in civil law. The European Union has introduced in its Anti-Tax Avoidance Directive a general anti-abuse test which must be transposed into the domestic laws of Member States. Such a test has its inner structure, consisting of an elimination and requalification stage, while the elimination stage entails genuineness and a tax benefit test. The general anti-abuse test has a great potential (or scalability when speaking in the language of start-ups) of being automated and integrated into different legal application processes (such as taxpayer self-assessment systems, transactions certified by public notary or merger and acquisition deals) to discover debt push down abuses or other arrangement structures which may have abusive content. While the best method for create a reliable algorithm is a decision tree type model, the inner ambiguity of the general anti-abuse test prevents using the full benefits of automation of tax laws. The purpose of this article is to design a decision tree type model for the test and address the main challenges of such a model, both from the perspective of the clarity of concepts and the quality of input information such an engine would use.

Legal Engineering of the Anti-Abuse Rule in ATAD: Architecture of the Regression Tree Model

TalTech Law School, Tallinn University of Technology

TalTech Journal of European Studies

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Every taxable arrangement is subject to an anti-abuse test. Abusive arrangements are treated as not valid for tax purposes, which is similar to the treatment...

Legal Engineering of the Anti-Abuse Rule in ATAD: Architecture of the Regression Tree Model

Online veröffentlicht: 15. Nov. 2021

Seitenbereich: 65 - 82

DOI: https://doi.org/10.2478/bjes-2021-0015

Schlüsselwörter
anti-abuse rule, ATAD, decision tree, legal algorithms, legal engineering, machine learning, regression tree

© 2021 Kaido Künnapas, published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Legal Engineering of the Anti-Abuse Rule in ATAD: Architecture of the Regression Tree Model

Online veröffentlicht: 15. Nov. 2021

Seitenbereich: 65 - 82

DOI: https://doi.org/10.2478/bjes-2021-0015

Schlüsselwörteranti-abuse rule, ATAD, decision tree, legal algorithms, legal engineering, machine learning, regression tree

© 2021 Kaido Künnapas, published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Schlüsselwörter
anti-abuse rule, ATAD, decision tree, legal algorithms, legal engineering, machine learning, regression tree