УДК338.2 JELO10
AGENT-BASED MODELING OF THE DEVELOPMENT OF A COMPLEX TAX ECOSYSTEMWITH SUPTECH AND REGTECH IMPLENMENTATION FOR
BEPS CASE
АГЕНТНОЕ МОДЕЛИРОВАНИЕ РАЗВИТИЯ СЛОЖНОЙ НАЛОГОВОЙ ЭКОСИСТЕМЫ В СЛУЧАЕ РАЗМЫВАНИЯ НАЛОГОВОЙ БАЗЫ ПРИ ВНЕДРЕНИИ SUPTECH И REGTECH
ILYINSKY ALEXANDER ORCID 0000-0002-7803-9146
ИЛЬИНСКИЙ АЛЕКСАНДР ИОИЛЬЕВИЧ
Financial University under the Government of Russian Federation Финансовый Университет при Правительстве Российской Федерации
Keywords: Complex adaptive systems, agent-based modeling, digital tax ecosystem, decision making, BEPS, tax evasion
The challenges of digital globalization facing the tax system of the Russian Federation raise the question of a fundamental change in the architecture of tax administration. In the near future, we will face the processes of deep tax robotization, the massive development of online services and individual chat bots, the creation of digital systems for managing the loyalty of taxpayers and the massive transformation of buyers into freelance tax inspectors [1]. Much attention will be paid to improving communication between tax services and taxpayers, taking into account their personal preferences, developing a tax culture and changing the behavior of taxpayers. Supervisory and regulatory bodies and agencies are transitioning to innovative fintech technologies SupTech (Supervisory Technology) and RegTech (Regulatory Technology) through digitalization of reporting and management of big data processing to improve the efficiency of supervisory institutions and regulators [2].
The Bank of Russia published an order on the massive implementation of high-tech supervision and regulation of SupTech and RegTech in April 2021 [3]. This document deals with the development of a data-centric approach, a massive transition to reporting in XBRL-format, automated monitoring of transactions in real time and digital assessment of
Ключевые слова: Сложные адаптивные системы, агентное моделирование, цифровая налоговая экосистема, принятие решений, BEPS, уклонение от налогов
Вызовы цифровой глобализации, которые стоят перед налоговой системой Российской Федерации, ставят вопрос коренного изменения архитектуры налогового администрирования. В ближайшее время мы столкнемся с процессами глубокой налоговой роботизации, массового развития онлайн сервисов и индивидуальных чат-ботов, создания цифровых систем управления лояльностью налогоплательщиков и массовой трансформацией покупателей во внештатных налоговых инспекторов [1]. Большое внимание будет уделяться совершенствованию
коммуникаций между налоговыми сервисами и налогоплательщиками с учетом их персональных предпочтений, развитию налоговой культуры и изменению поведения налогоплательщиков. Происходит переход надзорных и регуляторных органов и агентств на инновационные технологии финтеха SupTech (Supervisory Technology) и RegTech (Regulatory Technology) с помощью цифровизации отчетности и управления обработки больших данных для повышения эффективности работы надзорных институтов и регуляторов [2].
Распоряжение о массовом внедрении высокотехнологичного надзора и регулирования SupTech и RegTech Банк России опубликовал в апреле 2021 года [3]. В указанном документе речь идет о развитии датацентричного подхода, массового перехода к отчетности в в XBRL-
business risks. On the other hand, in all tax jurisdictions there is anerosion of the tax base and there are attempts to shift profits to offshore zones and tax havens. This makes the issue of joint effective measures of tax services to combat against this phenomenon relevant, and international tax experts agreed to consider proposals in two pillars, as a consensus solution to tax problems associated with digitalization [4]. Any system of tax administration is a complex adaptive system, which consists of a huge number of interacting individuals and legal entities, a large number of tax institutions and extensive regulatory documents that determine the modes of functioning of the entire system.
We represent the tax ecosystem as an interconnected and interdependent network structure consisting of subjects and objects of different nature (people, institutions, digital objects and platforms, a system of legal norms and law enforcement practices) to finance public spending. The tax ecosystem is immersed and interacts with the external socioeconomic environment. Any changes in the interaction of participants in the tax system can lead to the emergence of new, often unforeseen, modes of behavior of the entire aggregate system, to the manifestation of emergent properties of a complex adaptive system. External shocks and changes in a broad socio-economic context lead to the emergence of dynamic responses in the tax ecosystem, and the nonlinearity of the system itself does not allow the use of the linear principle of superposition of responses to various types of shocks and external influences. The complex structure of the tax ecosystem, its significant nonlinearity and heterogeneity of the population of tax agents create currently insurmountable obstacles to the use of classical micro- and macroeconomic models.
As an alternative modeling method for overcoming the indicated disadvantages of neoclassical models, the agent-based modeling method has become widespread in social, economic, and environmental studies [5]. Agent-based modeling is a powerful simulation method using autonomous digital objects that interact with each other and the environment, exchange information and make decisions, that is, perform targeted actions. These digital objects are called agents. With the help of sets of interacting
формате, автоматизированном мониторинге операций в реальном масштабе времени и цифровой оценке хозяйственных рисков. С другой сторону, во всех налоговых юрисдикциях отмечается размывание налоговой базы и наблюдаются попытки вывода прибыли в офшорные зоны и налоговые гавани. Это делает вопрос о совместных эффективных мерах борьбы налоговых служб с этим явлением актуальным, и международные налоговые эксперты согласились рассмотреть предложения по двум направлениям, в кчестве консенсусного решения налоговых проблем, связанных с цифровизацией [4]. Любая система налогового администрирования представляет собой сложную адаптивную систему, которая состоит из огромного числа взаимодействующих физических и юридических лиц, большого числа налоговых институтов и обширной нормативно-правовой документации, определяющей режимы функционирования всей системы.
Рассмотрим налоговую экосистему как взаимосвязанную и взаимозависимую сетевую структуру, состоящую из субъектов и объектов различной природы (люди, институты, цифровые объекты и платформы, система правовых норм и практики правоприменения) для финансирования государственных расходов. Налоговая экосистема погружена и взаимодействует с внешней социально-экономической средой. Любые изменения взаимодействия участников налоговой системы могут приводить к возникновению новых, часто заранее непредвиденных, режимов поведения всей совокупной системы, к проявлению эмерджентных свойств сложной адаптивной системы. Внешние шоки и изменения в широком социально-экономическом контексте приводят к возникновению динамических откликов в налоговой экосистеме, а нелинейность самой системы не позволяет использовать линейный принцип суперпозиции откликов на различные виды шоков и внешних воздействий. Сложная структура налоговой экосистемы, ее существенная нелинейность и гетерогенность популяции налоговых агентов создаёт непреодолимые в настоящее время препятствия использования классических микро- и макроэкономических моделей.
В качестве альтернативного метода моделирования для преодоления указанных
heterogeneous agents, models of artificial communities are constructed, which convey the main properties of the studied complex adaptive systems. The tax ecosystem is modeled by the behavior of a population of heterogeneous tax agents (taxpayers, tax officers, auditors, etc.), which are described by a set of internal variables, interactions between agents are determined by a set of fixed rules.
Interactions of agents fit into a complex socioeconomic environment of behavioral rules, social norms and tax law enforcement. Thus, the use of agent-based models is due to their ability to investigate the heterogeneous behavior of large populations and complex tax ecosystems. Analysis of foreign literature has shown the widespread and high efficiency of agent-based modeling methods to improve the efficiency of national tax systems in combating tax evasion. Such work is carried out on the orders of the national tax authorities. Currently, there is a huge number of agent-based models for the functioning of tax systems in various jurisdictions, a modern overview of which can be found in the monograph [6]. A sophisticated agent-based model of the tax system, developed by the Rand Corporation at the request of the US Internal Revenue Service, made it possible to determine the possible consequences of changes in the efficiency of public spending financing [7].
This report is devoted to the construction of an agent-based model of a tax system with a set of heterogeneous agents, partially evading taxes. The neoclassical theory of Allingham and Sandmo [8] is used to construct a model of an agent-taxpayer evading taxes. An analytical solution of the equilibrium equation is obtained and the value of tax evasion is determined for the case of an upward convex power function of expected utility. The values of the absolute and relative coefficients of risk aversion Arrow-Pratt for the agent-taxpayer are found. The analysis of the sensitivity of the equilibrium solution for the exogenous variables of the model is carried out. A simple conceptual model of the tax ecosystem is proposed, which was
недостатков неоклассических моделей широкое распространение в социальных, экономических, экологических исследованиях получил метод агентного моделирования [5]. Агентное моделирование - это мощный метод имитационного моделирования с использованием автономных цифровых объектов, которые взаимодействуют друг с другом и окружающей средой, обмениваются информацией и принимают решения, т. е. совершают целенаправленные действия. Эти цифровые объекты получили название агентов. С помощью наборов взаимодействующих гетерогенных агентов строятся модели искусственных сообществ, которые и передают основные свойства изучаемых сложных адаптивных систем. Налоговая экосистема моделируется поведением популяции гетерогенных налоговых агентов
(налогоплательщиков, налоговиков, аудиторов и т. д.), которые описываются набором внутренних переменных, взаимодействия между агентами определяются набором фиксированных правил.
Взаимодействия агентов вписываются в сложноорганизованное социально-экономическое окружение поведенческих правил, социальных норм и налогового правоприменения. Таким образом, использование агентных моделей обусловлено их способностью исследовать гетерогенное поведение больших популяций и сложно организованных налоговых экосистем. Анализ зарубежной литературы показал широкое распространение и высокую эффективность методов агентного моделирования для повышения эффективности национальных налоговых систем по борьбе с уклонением от налогов. Подобные работы выполняются по заказам национальных налоговых органов. В настоящее время существует огромное число агентных моделей функционирования налоговых систем различных юрисдикций, современный обзор которых можно найти в монографии [6]. Сложная агентная модель налоговой системы, разработанная Рэнд Корпорейшен позаказу Налогового управления США, позволила определить возможные последствия изменения эффективности
финансирования государственных расходов [7].
Настоящая статья посвящена построению агентной модели налоговой системы с набором гетерогенных агентов, частично уклоняющихся от уплаты налогов. Для построения модели агента-
implemented in an agent-oriented programming налогоплательщика, уклоняющегося от уплаты language in the NetLogo integrated development налогов, используется неоклассическая теория environment. Аллинхэма и Сандмо [8]. Получено аналитическое
решение уравнения равновесия и определена величина уклонения от налогов для случая выпуклой вверх степенной функции ожидаемой полезности. Найдены значения абсолютных и относительных коэффициентов неприятия риска Арроу-Пратта для агента-налогоплательщика. Проведен анализ чувствительности равновесного решения по экзогенным переменным модели. Предложена простая концептуальная модель налоговой экосистемы, которая была реализована на агентно-ориентированном языке
программирования в интегрированной среде разработки NetLogo.
References / Библиография
1. План деятельности ФНС России на 2021 год https://data.nalog.ru/html/sites/www.new.nalog.ru/docs/analit/pg2021.pdf
2. Dirk Broeders and Jermy Prenio. Innovative technology in financial supervision (suptech) - the
experience of early users, FSI Insights, No 9, 16 July 2018, Bank for International Settlements 2018.
3. Основные направления развития технологий SupTech и RegTech на период 2021-2023 годов
https://www.cbr.ru/press/event/?id=9801
4. OECD (2020), Tax Challenges Arising from Digitalisation - Report on Pillar One Blueprint:
Inclusive Framework on BEPS, OECD/G20 Base Erosion and Profit Shifting Project, OECD Publishing, Paris, https://doi.org/10.1787/beba0634-en.
5. Wilensky, U., & Rand, W. (2015). An introduction to agent-based modeling: Modeling natural,
social and engineered complex systems with NetLogo. Cambridge, MA: MIT Press.
6. Agent-based Modeling of Tax Evasion: Theoretical Aspects and Computational Simulations. Ed.
Sascha Hokamp, Laszlo Gulyas, Matthew Koehler and H. Sanith Wijesinghe 2018, 376 р.
7. RAND's Interdisciplinary Behavioral and Social Science Agent-Based Model of Income Tax
Evasion (2019)
https://www.rand.org/content/dam/rand/pubs/working_papers/WR 1300/WR1322/RAND_WR 132 2.pdf
8. Allingham, M.G. and A. Sandmo, 1972. Income tax evasion: A theoretical analysis. J. Public Econom., 1: 323-
338. DOI: 10.1016/0047-2727(72)90010-2