An Empirical Analysis of Tax Evasion among Companies Engaged in Stablecoin Transactions

Abstract

This research investigates the relationship between stablecoin usage and tax evasion. We present a model that includes variables related to transactions such as intensity, frequency, environment on-chain (P2P) vs. off-chain (IntraVasp), and company characteristics such as age, sector, and size. Our model was empirically tested using a logistic regression based on data from the Brazilian Federal Revenue Service (Receita Federal do Brasil (RFB)) in 2021. This novel approach aims to understand the tax behaviours associated with stablecoin use in corporate financial practices. Our results indicate that the intensity, frequency, environment of transactions (specifically IntraVasp and P2P transactions), age, sector, and size are factors significantly associated with tax evasion behaviour. However, we found no evidence to suggest that firms engaging in only P2P transactions have a higher propensity for tax evasion than those engaging only in IntraVasp transactions. Our findings reveal that younger and medium-sized companies with intensive use of stablecoin, with high stablecoin transaction frequency, engaging in IntraVasp and P2P transactions, and belonging to the service sector are more likely to evade tax. Therefore, our research provides a detailed understanding of how digital financial practices with crypto assets (blockchain-based technology) intersect with corporate tax strategies, which can offer valuable insights for regulators, industry practitioners, and policymakers.

1. Introduction

Blockchain technology, which comprises a decentralised digital ledger that records transactions across many computers, has revolutionised digital transactions, ensuring security and transparency (Zhang et al. 2023). Within this innovative ecosystem, stablecoins, a form of crypto asset, have emerged as a significant development (Allen et al. 2022; Baur and Hoang 2021; Li et al. 2024; Pernice et al. 2019). Unlike other crypto assets, generally known for their volatility, stablecoins aim to maintain stable values by staying pegged to a reserve asset, such as the USD, gold, or other crypto assets (Allen et al. 2022; Hampl and Gyönyörová 2021; Li et al. 2024; Pernice et al. 2019; Waerzeggers et al. 2023; G.-J. Wang et al. 2020). This stability has made them an attractive transaction medium and a bridge between traditional fiat currencies and crypto assets, enhancing their utility in the digital or traditional economy (Li et al. 2024; Tarasova et al. 2020). For example, companies involved in international trade can use stablecoins or decentralised finance to facilitate cross-border payments (Chainalysis 2023; Harvey and Rabetti 2024; IMF 2020). In addition, those who operate in environments with unstable banking systems or economies with high inflation may prefer to carry out transactions in stablecoins or other crypto assets (Chainalysis 2023; IMF 2020; Waerzeggers et al. 2023). Additionally, as a store of value, stablecoins present an opportunity for investors to preserve capital within the volatile crypto markets, acting as a safe haven during periods of high fluctuation (Baur and Hoang 2021; Grobys et al. 2021; Travkina et al. 2022; G.-J. Wang et al. 2020).

While stablecoins have legitimate benefits, companies engaging with stablecoins may have specific reasons to bypass traditional financial oversight. Considering that stablecoin payments can resemble cash payments (Moura de Carvalho et al. 2022; Son et al. 2022), and cash payments are strongly associated with tax evasion since they are not traceable and facilitate the under-reporting of sales, in addition to their association with usage in illegal activities (Aurazo 2020; Immordino and Russo 2018; Madzharova 2014), stablecoin payments might similarly facilitate tax evasion. One scenario is a company receiving stablecoins in exchange for goods or services and intentionally omitting these transactions from their taxable records. In addition, the ease of transferring stablecoins across borders without the typical financial oversight or regulatory reporting requirements further complicates the ability of tax authorities to track, trace, and tax these assets effectively.

Recent studies have addressed tax evasion with crypto assets. For instance, Cong et al. (2023) utilised off-chain data (transactions executed internally on a centralised platform) to demonstrate that crypto traders harvest tax loss as an alternative to non-compliance, especially during increased IRS tax scrutiny. On the other hand, De Simone et al. (2024) found that traders evade taxes by pledging their assets on DeFi lending platforms. Furthermore, a comprehensive study by Meling et al. (2024), which merged de-anonymised trading data of crypto assets with individual tax returns, survey data, and records from tax enforcement actions from the Norwegian Tax Administration, revealed widespread tax non-compliance among crypto investors. This is notable even among those trading on exchanges that provide identifiable data to tax authorities. The study also highlighted that males and younger and urban individuals are more likely to invest in crypto assets. However, despite the prevalence of stablecoin transactions in the crypto asset market (Chainalysis 2023), no study has investigated the tax evasion behaviour of companies engaging in stablecoin transactions. This gap in research is particularly concerning regarding whether factors—such as a company’s size, age, sector, or the intensity, frequency, and environment of transactions—might be associated with tax evasion behaviour. Our research addresses these gaps by empirically analysing the tax evasion behaviours of small and medium enterprises (SMEs) engaged in stablecoin transactions in Brazil during 2021, based on confidential RFB data.

To achieve this goal, we applied quantitative methodology through logistic regression. We explored six key questions to deepen our understanding of the factors associated with tax evasion behaviours in the stablecoin market. First, we investigated whether the intensity of stablecoin use is associated with a higher propensity for tax evasion. Second, we analysed if companies engaged in high-frequency stablecoin transactions are more likely to evade taxes than those with low-frequency transactions. Third, we considered whether companies engaging in on-chain transactions are more prone to tax evasion than those involved only in off-chain transactions. Fourth, we assessed the relationship between a company’s age and its propensity to engage in tax evasion. Fifth, we examined if service sector companies are more likely to evade taxes than those in the commercial sector. Finally, we explored whether medium-sized companies have a higher probability of tax evasion than smaller ones. The main results reveal that younger and medium-sized companies with intensive use of stablecoin and high stablecoin transaction frequency, engaging in off-chain and on-chain transactions and belonging to the service sector, are likelier to evade taxes.

The relevance of this study stems from the increasing prominence of stablecoins in cryptocurrency transactions within Brazil (RFB 2023) and other emerging countries (Chainalysis 2023), suggesting a global trend towards their use in electronic commerce, international remittances, and as gateways into the broader crypto asset market. This rise in stablecoin utilisation underscores the need for comprehensive research to understand the tax behaviours associated with these transactions, mainly since data gaps currently prevent regulatory bodies from fully grasping the extent and nature of stablecoin usage in developing economies (Financial Stability Board—FSB 2024). Additionally, the scarcity of robust evidence regarding tax evasion within the crypto asset sector (Baer et al. 2023; Meling et al. 2024) makes this study particularly crucial. By shedding light on how stablecoins are incorporated into business practices, the findings of this study also offer valuable insights for policymakers and researchers focused on enhancing stablecoin auditing and compliance frameworks.

In addition, considering that tax-evasive practices are one of the main problems of emerging and transition economies (Pedroni et al. 2022) and that they deprive the state of billions of dollars that could be used for education, healthcare, infrastructure, or social services (Korndörfer et al. 2014), our research complements studies on tax evasion in emerging economies. For example, Clemente et al. (2021) investigated the differences observed in the rate of tax evasion between Global North and South countries, with a particular focus on Brazil, by comparing critical parameters of their tax systems. Their findings pointed to shortcomings in the tax systems of Latin American countries as creating an avenue for high tax evasion. Additionally, McGee et al. (2022), in a study about attitudes towards tax evasion of sample populations in Brazil, Russia, India, and China, identified that Brazilians show less opposition to tax evasion than Chinese and Indian respondents, albeit showing more opposition than Russians. Notably, Brazilian opposition to tax evasion has diminished over time. Therefore, our study may inspire the strategies of governments and policymakers towards preventing tax evasion schemes, which can increase public revenue.

This article is structured as follows: First, the conceptual background addresses the basic concepts of stablecoin, the environment of stablecoin transactions (on- and off-chain), the Crypto Assets Transactions Report in Brazil—IN 1888/2019, and the Crypto Asset Reporting Framework—CARF. Second, we formulate the research hypothesis. Next, the data sample, measurement of tax evasion, and empirical model are presented. Then, the results are presented and discussed. The last section presents the main conclusions, limitations, and future works.

2. Conceptual Background

2.1. Stablecoin

Stablecoins represent a significant innovation within digital currencies, building directly upon the foundational technologies of blockchain and smart contracts. Stablecoin values are pegged to more stable assets, like fiat currencies, to counteract the high volatility typically associated with cryptocurrencies. There are three main types of stablecoins: fiat-backed, crypto-backed, and algorithmic (Moura de Carvalho et al. 2022; Grobys et al. 2021; Oefele et al. 2024).

The fiat-backed stablecoins are the most common and straightforward. They are pegged to traditional fiat currencies such as the USD, EUR, or JPY, typically at a 1:1 ratio. They maintain reserves of these currencies as collateral to ensure the value remains stable (Grobys et al. 2021; Hampl and Gyönyörová 2021). Examples of fiat-backed stablecoins include USDC and Tether (USDT), which promise easy conversion between the digital token and the fiat currency it represents. Fiat-backed stablecoins offer stability and ease of use, but there are issues associated with their structure and reliance on traditional financial systems. One primary concern is their centralisation; unlike decentralised crypto assets, fiat-backed stablecoins depend on centralised entities to manage the fiat reserves that back them (Catalini et al. 2022). This centralisation introduces potential risks, including the possibility of mismanagement or fraud by the controlling entity. Another critical issue is the need for transparency and regularity in audits. Ensuring stablecoins are fully backed by fiat reserves requires transparent and frequent audits (Arner et al. 2020). Liquidity risks are also associated with fiat-backed stablecoins, particularly during market stress or financial panic. If many users attempt to redeem their stablecoins simultaneously, enough liquid assets might not be available to promptly satisfy these demands (Catalini et al. 2022). This situation can lead to delays in redemption and impact the stablecoin’s ability to maintain its tie to the fiat currency. Lastly, the dependence on the traditional banking system to hold and manage fiat reserves introduces several risks, mainly those associated with the custodial bank’s financial health (Galati and Capalbo 2024). This dependence can pose significant risks, suggesting a need for careful oversight, rigorous management practices, and strict regulatory compliance to maintain the advantages of fiat-backed stablecoins.

Crypto-backed stablecoins are another type where the value is secured by other cryptocurrencies rather than fiat money (Moura de Carvalho et al. 2022; Grobys et al. 2021). These stablecoins are over-collateralised to account for the volatility of the backing assets (Catalini et al. 2022); for instance, if a stablecoin is pegged to the USD but backed by Ethereum, the amount of Ethereum held as collateral will be worth more than the number of stablecoins issued. This buffer helps to absorb fluctuations in the underlying crypto’s market value, ensuring the stability of the stablecoin. DAI is an example of a stablecoin backed by crypto that can be minted with sufficient collateral in other cryptocurrencies (Catalini et al. 2022).

Algorithmic stablecoins, the third category, attempt to maintain their peg through a working mechanism managed by algorithms and smart contracts without relying on traditional collateral (Briola et al. 2023; Lee et al. 2023; Morgan 2023). However, this model carries significant risks, as illustrated by the Terra Luna collapse (Briola et al. 2023; Cho 2023; Lee et al. 2023). Terra’s stablecoin (UST) was meant to stay pegged to the USD algorithmically but failed, leading to a dramatic loss in value and severe market consequences.

Stablecoins emerge as a critical innovation in the digital asset space, offering a bridge between the volatility of cryptocurrencies and the stability of fiat currencies (Li et al. 2024). Whether fiat-backed, crypto-backed, or algorithm-driven, each type of stablecoin leverages the unique benefits of blockchain technology, smart contracts, and innovative economic models to provide users with stable value in the digital realm (Moura de Carvalho et al. 2022).

2.2. On- and Off-Chain Stablecoin Transaction

A simplified bifurcation was established in our scientific investigation of stablecoin transactions, classifying transactions as either on-chain or off-chain (Moura de Carvalho et al. 2022). On-chain transactions are recorded on the blockchain and involve a peer-to-peer (P2P) exchange. This encompasses decentralised transactions that can occur directly between companies or individuals or between a company or individual and decentralised finance platforms. On-chain transactions are characterised by transparency, as they are permanently recorded on a distributed ledger. However, while the transaction record itself is public, the participants’ identities are often pseudonymous, providing a level of privacy and making it challenging for auditors to trace and verify the parties involved in the transactions (Moura de Carvalho et al. 2022).

Off-chain transactions, by contrast, refer to those that are conducted internally within a centralised crypto asset1 exchange (known as a Virtual Asset Service Provider, VASP, as denominated by the Financial Action Task Force [FATF] (2021)) and not transmitted to the blockchain (AICPA 2022). In this paper, we classified these transactions as ‘IntraVasp’. These IntraVasp transactions do not have the same level of transparency as on-chain transactions. Nevertheless, they are conducted under the auspices of a centralised entity that can require user identification for compliance with regulatory standards, such as Know Your Customer (KYC) and Anti-Money Laundering (AML) procedures. Thus, IntraVasp transactions, while less public, can be subject to oversight, and the participants are known to the VASP. IntraVasp transactions can provide a clearer picture of the participants involved, enabling auditors to identify and verify transaction details more easily (Moura de Carvalho et al. 2022).

Our study categorises companies based on their engagement in either or both transaction types. Companies engaging in IntraVasp and P2P transactions are considered to be utilising the full spectrum of what stablecoins can offer in terms of transactional flexibility. This dual-use approach suggests that these companies are navigating both regulated and less regulated spaces, maximising the benefits of stablecoins’ versatility. It reflects a comprehensive use of digital financial systems, exploiting the advantages of both the transparency and trust offered by blockchain technology and the convenience and familiarity of traditional financial services provided by VASPs.

2.3. Crypto Assets Transactions Report in Brazil—IN 1888/2019

The data regarding company transactions with stablecoin (IntraVasp and P2P) were obtained from the RFB. Since August 2019, the RFB has accessed transaction histories from local crypto asset exchanges in Brazil to combat money laundering and tax evasion. RFB Normative Instruction N° 1888, dated to 3 May 2019, represents an important regulatory framework in Brazil by establishing guidelines for declaring transactions with crypto assets to the RFB. This regulation aims to increase transparency and fiscal control over transactions with virtual assets, such as Bitcoin, Ethereum, and stablecoin, which have gained popularity and significant economic relevance in recent years.

This regulation requires VASPs in Brazil to disclose specific transaction data, including information about the parties involved. IN 1888 details the obligation to provide monthly information by two main agents: VASPs that operate and are domiciled for tax purposes in Brazil; and individuals or legal entities resident or domiciled in the country that carry out transactions with crypto assets on exchanges abroad or directly with another party via P2P. Brazil’s VASPs must report all transactions carried out IntraVasp to the RFB monthly. Individuals or legal entities must report when the volume of transactions in the month exceeds BRL 30,000.00 (thirty thousand reais2) carried out in VASP domiciled abroad or P2P transactions (RFB 2019).

The scope of Instruction 1888 extends to a wide range of operations with crypto assets. This expansion includes a diverse list of transactions such as purchase and sale, exchange, donation, transfer of crypto assets to and from the exchange, temporary assignment (rent), dation in payment, issuance, and other operations involving transferring crypto assets (RFB 2019). This specification aims to cover all possible forms of financial movements that can occur in the crypto asset ecosystem. The mention of ‘other operations that involve the transfer of crypto assets’ serves as an open clause, allowing the RFB to maintain the relevant regulations even in the face of the emergence of new types of transactions. This flexibility is crucial in a rapidly evolving environment, such as crypto assets, where new operations and uses for crypto assets are constantly being developed. For each transaction, the RFB is informed of the following details: type of transaction, identification of the parties involved, types of crypto assets involved, number of units transacted per type of crypto asset, and value of the transaction in BRL.

For comparison purposes, the scope of information in IN 1888 is similar to that of the definitions in the Crypto Asset Reporting Framework (CARF). The CARF is a model of rules developed and approved by the OECD (2022) for collecting information on crypto assets transacted through intermediaries intending to automatically exchange this information between jurisdictions. We identified similarities between the Brazilian (IN 1888) and the CARF models: the definitions of the crypto asset are aligned; concerning transactions, both scopes cover purchase and sale transactions, exchange, donation, transfer of crypto assets to the exchange, withdrawal of crypto assets from the exchange, dation in payment, and other operations that imply transfers of crypto assets. However, there are differences between the Brazilian and CARF models. One difference between the models is the granularity of the information provided. IN1888 adopts the transaction-by-transaction approach. In CARF, the information is provided in aggregate form by type of crypto asset for each user. Another difference is that only IntraVasp information is provided in the CARF. In the Brazilian model, if there is no intermediary, the information must be provided by the user of the crypto asset domiciled or residing in the country, thus reaching the so-called P2P transactions.

Therefore, Brazil’s Normative Instruction 1888/2019 established a regulatory framework for overseeing crypto asset transactions within the country, predating similar global initiatives such as the OECD’s CARF3. Despite these differences, with some adaptations, researchers or regulators can use data from tax administrations in countries that join CARF to reproduce and compare the results obtained in our research.

3. Hypothesis Development

Tax evasion is an individual’s illegal and deliberate act to reduce tax obligations (Alm 2012). This can involve under-reporting income, profits, and gains, overstating deductions, failing to file tax returns, concealing assets, engaging in fraudulent transfers, or using complex offshore structures to hide taxable income. According to Economic Deterrence Theory, individuals and organisations make rational decisions based on a cost–benefit analysis, weighing the potential benefits of evading taxes, such as increased profits and cash flow, against the costs, which include the probability of being detected, the severity of penalties, and the legal and reputational risks (Allingham and Sandmo 1972). In the digital economy, the obscurity of transactions involving stablecoins or other crypto assets may reduce perceived detection risks, potentially encouraging evasion. Meling et al. (2024) found that 80% of the investors trading on domestic crypto exchanges fail to declare their cryptos, even though these exchanges share identifiable trading data with the Norwegian Tax Administration. However, in line with the Economic Deterrence Theory and building on the premise that tax-loss harvesting using crypto assets also indicates tax reporting compliance, Cong et al. (2023) found improvements in tax compliance for crypto assets due to enhanced IRS enforcement.

To develop effective strategies to mitigate tax evasion and increase compliance and risk perception, the tax administration must understand the behaviour of taxpayers engaging in transactions with stablecoins or other crypto assets. In this sense, based on empirical analysis of RFB data, we formulated hypotheses, described below, to verify whether the intensity, frequency, and environment of stablecoin transactions, together with company characteristics, such as size, age, and sector, can be significantly associated with evasive tax behaviour. The proposed hypotheses aim to systematically explore these relationships, providing a comprehensive view of the factors influencing potential tax evasion in the context of stablecoin usage.

3.1. Intensity

The hypothesis related to the intensity of stablecoin use offers a nuanced understanding of how the depth of engagement with stablecoin can influence corporate tax behaviours. The ratio between stablecoin purchases and sells and bank deposits indicates the intensity of stablecoin use by companies. Supposing that a significant portion of a company’s financial transactions are directed towards acquiring stablecoin, it is reasonable to infer that it is used for commercial transactions, such as paying suppliers. The same reason may indicate that companies receive stablecoin payment for their products or services. In this case, ‘buying’ stablecoin would be equivalent to receiving it in exchange for goods or services. Intensive use of stablecoins could facilitate financial transactions that still need to be fully captured in traditional banking or reporting systems, potentially leading to tax evasion due to under-reported revenues or taxes. The relationship between the intensity of stablecoin use and the propensity to evade taxes is based on the premise that stablecoins offer a digital, borderless, and potentially less regulated medium of exchange.

Any other traceable payment method makes evasion more complicated (Immordino and Russo 2018). Adhikari et al. (2021), using the information available on Form 1099-K, various tax returns, and IRS information reports, developed an index of the intensity of payment card use equal to the ratio of receipts via card credit and total receipts. The authors’ research hypothesis was that in locations where the use of payment cards by consumers is high, companies receive a greater part of their revenue through payment cards and, therefore, have a greater part of their revenue reported to the IRS through Form 1099-K. Hence, companies that use credit cards more intensively are less prone to tax evasion.

Based on these assumptions, this study intended to test hypothesis 1:

H1. 

The intensity of stablecoin use is associated with the propensity for companies to engage in tax evasion behaviours.

3.2. Frequency

Expanding on the frequency of stablecoin transactions offers an opportunity to explore how the regularity of engagements with digital currencies impacts corporate tax behaviour. This dimension can reveal insights into operational practices and their potential alignment with tax evasion strategies. The rationale behind this hypothesis is that frequent transactions may reflect higher volumes of crypto activity and suggest a more systemic integration of stablecoins into the company’s financial practices, potentially facilitating tax evasion strategies through complexity and transaction volume. Conversely, less frequent (low-frequency) transactions might be sporadic or opportunistic and do not align with attempts to minimise tax liabilities in less transparent ways. To investigate this economic phenomenon, we hypothesised:

H2. 

Companies engaged in high-frequency stablecoin transactions are likelier to evade taxes than those with low-frequency stablecoin transactions.

3.3. IntraVasp (Off-Chain) and P2P (On-Chain) Stablecoin Transactions

The environment in which stablecoin transactions occur (whether within a VASP, P2P, or a combination of both) can significantly influence the probability of companies engaging in tax evasion. IntraVasp transactions balance convenience and compliance, while P2P transactions could be preferred for their directness and potential for anonymity (Moura de Carvalho et al. 2022). Engaging in both might indicate a sophisticated approach to leveraging the strengths of each for operational or strategic benefits, including tax considerations. Therefore, the probability of evading tax can vary significantly across different transaction modes. Companies engaging exclusively in P2P transactions show the highest propensity towards tax evasion, followed by those engaged in IntraVasp and P2P transactions and those operating solely within crypto asset exchanges (IntraVasp). As such, we formulated two hypotheses to test whether the mode of stablecoin transactions affects the probability of tax evasion:

H3a. 

Companies engaging in IntraVasp and P2P transactions have a higher propensity to evade tax than those engaging only in IntraVasp transactions.

H3b. 

Companies engaging in P2P transactions have a higher propensity to evade tax than those engaging only in IntraVasp transactions.

This hypothesis stems from the assumption that the degree of regulatory oversight decreases from IntraVasp to P2P transactions (FATF 2021), thereby increasing the ease with which companies can evade taxes. These hypotheses aim to illuminate the complex dynamics between stablecoin transactions and corporate tax behaviour, contributing to the growing literature on digital currencies and fiscal governance.

3.4. Company Age

Per Parisi and Federici (2023), in a study on tax aggressiveness, it is challenging to anticipate the correlation between the age of a company and its tax behaviour. As companies age, they typically establish more robust internal controls, gain experience navigating regulatory environments, and face greater stakeholder scrutiny, which may lead to more accurate and transparent financial reporting. Moreover, older companies might prioritise maintaining their reputation and minimising legal risks, which could reduce the likelihood of under-reporting tax revenue. However, these firms may also be more skilled at hiding their financial reporting (Parisi and Federici 2023). Regarding the tax evasion literature, Pedroni et al. (2022) and Wang (2012) found that evasive corporate behaviour through sales under-reporting is negatively linked with company age. Thus, we propose the following hypothesis:

H4. 

The company’s age is associated with the propensity for companies to engage in tax evasion behaviours.

3.5. Sector

The service sector often involves intangible goods, knowledge-based services, and non-standardised pricing mechanisms, which may provide more opportunities for financial opacity or discretion in reporting revenues and taxes. In contrast, the commercial sector, characterised by tangible goods and inventory, might need more flexibility. According to Abdixhiku et al. (2017), the service industry tends to be more tax evasive than wholesale and retail companies. Pedroni et al. (2022)’s findings show that service companies are more likely to under-report their sales than commercial companies. The nature of the two sectors justifies the hypothesis that companies in the service sector have a higher incidence of tax evasion compared to those in the commercial sector:

H5. 

Companies in the service sector are likelier to evade taxes than those in the commercial sector.

3.6. Size

The literature addresses the relationship between company size and tax evasion. Two studies by Ngah et al. (2020) and Mohamad et al. (2016) found that company size significantly impacts positive tax evasion practices. However, Azrina Mohd Yusof et al. (2014) research found the opposite; they discovered a negative relationship between company size and tax non-compliance of SMEs. As SMEs become larger, they become more compliant. Additionally, Pedroni et al. (2022) research revealed that the size factor is negatively associated with revenue under-reporting, especially when measured by annual sales. However, we point out that in our study, the variable that represents the size of the company is the total bank deposits, as described in the methodology section. Thus, based on the literature, we formulated the following hypothesis:

H6. 

Medium companies are likelier to evade tax than small companies.

4. Methodology

4.1. Sample and Data

RFB has been collecting information on tax transactions involving crypto assets since August 2019. According to data from August 2019 to January 2023 (shown in Figure 1), the volume of transactions carried out by firms and individuals involving crypto assets peaked at around USD 5 billion monthly in May 2021, considering the average dollar exchange rate of USD 1/BRL 5 (Brazilian real). This study considered stablecoins such as Tether (USDT), USD Coin (USDC), Binance USD (BUSD), Gemini Dollar (GUSD), TrueUSD (TUSD), and DAI, which have gained immense popularity and account for approximately 90% of all crypto transactions in Brazil as of December 2022.

According to data from the RFB, in 2021, 26,393 companies in Brazil carried out transactions with stablecoin. We selected only domestic SMEs from the commercial and service sectors with bank deposits ranging from USD 100,000 to USD 20,000,000. Additionally, all companies in the sample were operationally active in 2023 and did not include any VASP companies. This resulted in a sample of 1858 companies.

The choice of 2021 and 2021 only is because the data were collected in April 2023. On that date, there was no complete information on bank deposits for the year 2022. By law, banks in Brazil must report the volume of deposits and withdrawals from their customers’ current accounts in the year following the year in which the banking transactions occurred. In addition, the selection criteria for our study include companies with substantive economic activities and bank deposits ranging between USD 100,000 and USD 20,000,000. This range ensures that the sample includes only companies that are not too small to be negligible nor so large that they are outliers. The study also recognises the commerce and service sectors where stablecoin use is more relevant.

4.2. Measuring Tax Evasion

Collecting sufficient and trustworthy data is a primary challenge in researching and finding solutions to evasion behaviour and related questions. Obtaining concrete, dependable, and easily observable data on evasion behaviour is difficult (Alm 2012; Korndörfer et al. 2014; Lietz 2013). Scientific research to estimate the prevalence and extent of tax evasion has used various measurement methods (Alm 2012; Korndörfer et al. 2014). Several research studies employ data from surveys by the World Bank. The questionnaire in this survey provides indirect reasonable measures of tax evasion (Abdixhiku et al. 2017; Alm et al. 2019; Pedroni et al. 2022). This measurement is derived from a single survey question: ‘Recognising the difficulties many enterprises face in fully complying with taxes and regulations, what percentage of total sales would you estimate the typical establishment in your area of activity reports for tax purposes?’. This approach is intended to be a reliable replacement as respondents often hesitate to disclose their compliance. The answer given by the respondents is believed to be influenced by their personal experiences and is, therefore, regarded as a reasonable representation of their conduct (Abdixhiku et al. 2017; Alm et al. 2019; Pedroni et al. 2022).

The most accurate source of information is obtained through direct measurement of evasion, which involves actual audits of individual returns (Alm 2012). For example, Mohamad et al. (2016) examined factors influencing tax evasion in SMEs. The dependent variable was the total amount of tax evasion obtained from Inland Revenue Board of Malaysia (IRBM) data. Using data from the IRBM, Ngah et al. (2020) measured the total amount of tax evasion activities through fraudulent financial reporting practices uncovered during tax audits. DeBacker et al. (2015) used confidential Internal Revenue Service (IRS) audit data in another study. They measured a firm’s tax evasion using the ratio of IRS-determined tax deficiency over total income. The IRS-determined tax deficiency is the amount of positive adjustment to the firm’s tax liability following the audit. However, the authors argued that the tax adjustment used in this study is not strictly tax evasion but instead the best measure for tax evasion.

Our method, in line with studies like those of DeBacker et al. (2015), Mohamad et al. (2016), and Ngah et al. (2020), utilises objective data from tax records, contrasting them with the subjective estimates of self-reported surveys. We measured tax evasion by directly examining tax declarations to authorities, specifically identifying companies that reported zero tax revenue in the sample year. A binary classification, where ‘1’ indicates revenue was declared and ‘0’ indicates none was declared, provides a clear and quantifiable metric of tax evasion derived from actual tax authority interactions, thereby offering a more accurate reflection of real-world tax behaviour than indirect methods. This measure is consistent with the fact that it is unlikely that a company’s main activities, including those involving stablecoins, do not generate any reportable revenue. This approach was integrated into our logistic regression model to study tax evasion effectively.

4.3. Empirical Model

We applied a logistic regression model, which is suitable for predicting the probability of a binary categorical dependent variable (EVASION) based on the values of the independent variables. This model is considered as follows:

$$\begin{array}{ll}\ln \left({\displaystyle \frac{P_{EVASIO{N}_i}}{1-{ P}_{EVASIO{N}_i}}}\right)=& {\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{I}\mathrm{N}\mathrm{T}\mathrm{E}\mathrm{N}\mathrm{S}\mathrm{I}\mathrm{T}\mathrm{Y}}_{\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{F}\mathrm{R}\mathrm{E}\mathrm{Q}\mathrm{U}\mathrm{E}\mathrm{N}\mathrm{C}\mathrm{Y}\left(\mathrm{H}\mathrm{i}\mathrm{g}\mathrm{h}\right)}_{\mathrm{i}}+{\mathsf{\beta }}_3{\mathrm{O}\mathrm{N}\_\mathrm{O}\mathrm{F}\mathrm{F}\mathrm{C}\mathrm{H}\mathrm{A}\mathrm{I}\mathrm{N}\left(\mathrm{P}2\mathrm{P}\right)}_{\mathrm{i}}+{\mathsf{\beta }}_4{\mathrm{O}\mathrm{N}\_\mathrm{O}\mathrm{F}\mathrm{F}\mathrm{C}\mathrm{H}\mathrm{A}\mathrm{I}\mathrm{N}(\mathrm{I}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{V}\mathrm{a}\mathrm{s}\mathrm{p}\_\mathrm{P}2\mathrm{P})}_{\mathrm{i}}\\ & +{\mathsf{\beta }}_5{\mathrm{A}\mathrm{G}\mathrm{E}}_{\mathrm{i}}+{\mathsf{\beta }}_6{\mathrm{S}\mathrm{E}\mathrm{C}\mathrm{T}\mathrm{O}\mathrm{R}\left(\mathrm{S}\mathrm{e}\mathrm{r}\mathrm{v}\mathrm{i}\mathrm{c}\mathrm{e}\right)}_{\mathrm{i}}+{\mathsf{\beta }}_7{\mathrm{S}\mathrm{I}\mathrm{Z}\mathrm{E}\left(\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{u}\mathrm{m}\right)}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}\end{array}$$

where i represents each company; β_j represents the coefficients; ε_i is the error term.

The logistic regression model presented is designed to analyse the binary outcome of tax evasion (EVASION) among companies engaging in stablecoin transactions. The model incorporates a mix of continuous and categorical variables, offering a nuanced view of how various factors may influence the chance of a company to evade taxes.

Specifically, as described in

Table 1. Variable descriptions.

Variable	Description
EVASION	A dependent categorical variable representing companies that declared zero tax revenue to the tax administration in the sample year, ‘YES’, otherwise, ‘NO’.
INTENSITY	A continuous variable representing the natural logarithm of the ratio between the purchase and sale volume of stablecoin and the volume of bank deposits.
FREQUENCY	A categorical variable representing two levels of quantity stablecoin transactions. Companies performing above 15 transactions are categorised as ‘High’ and those not meeting this threshold are categorised as ‘Low’.
ON_OFFCHAIN	A categorical variable representing three categories: ‘IntraVasp’, for companies that only carried out transactions with stablecoin through a VASP; ‘P2P’, for companies that only transacted directly with another counterparty without VASP intermediation; and ‘IntraVaspandP2P’, for companies that carried out transactions either through a VASP or directly with another counterparty (P2P).
AGE	A continuous variable that represents the company’s natural logarithm of age.
SECTOR	A categorical variable representing the economic sector to which the company belongs based on its main activity: ‘Service’ or ‘Commercial’.
SIZE	A categorical variable representing the company’s size: ‘Small’ or ‘Medium’. Due to the lack of accounting information, we performed categorisation based on the volume of bank deposits: a ‘Small’ company was below BRL 10 million (~USD 2 million); a ‘Medium’ company was above this value until BRL 100 million (~USD 20 million).

, it includes INTENSITY (the natural logarithm of the ratio of stablecoin transactions to bank deposits), FREQUENCY (the number of stablecoin transactions), ON_OFFCHAIN (the mode of stablecoin transactions, whether within a virtual asset service provider, peer-to-peer, or both), AGE (the natural logarithm of the company’s age), SECTOR (the economic sector of operation, service or commercial), and SIZE (the size of the company, small or medium, based on bank deposit volume).

5. Results Analysis

5.1. Descriptive Statistics of Variables

This section presents the descriptive statistics of the variables considered in the empirical models. Descriptive statistics offer a quantitative summary of the variables for a sample.

Table 2. Descriptive statistics of the variables.

Variable		Mean	SD	Median (IQR)	Min–Max
Continuous variables:
INTENSITY		−2.44	1.447	−2.55 (−3.325 to −1.71)	−7.03–6.84
AGE		1.51	0.871	1.39 (0.693 to 2.08)	0.00–3.91
Dummy variables:	Level	Counts	% of Total
EVASION	No	1081	58.2 %
	Yes	777	41.8 %
FREQUENCY	High	907	48.8 %
	Low	951	51.2 %
ON_OFFCHAIN	IntraVasp	442	23.8 %
	P2P	387	20.8 %
	IntraVaspandP2P	1029	55.4 %
SECTOR	Commercial	1224	65.9 %
	Service	634	34.1 %
SIZE	Medium	802	43.2 %
	Small	1056	56.8 %

SD: standard deviation; Max: maximum value; Min: minimum value; IQR: interquartile range; Q1: first quartile (25th percentile); Q3: third quartile (75th percentile).

summarises descriptive statistics of the continuous variables used in the logistic regression, such as the minimum–maximum values, median, mean, standard deviation (SD), and interquartile range. Table 2 also shows the frequency and percentage of dummy variables.

Regarding the dependent variable EVASION, within the scope of this sample, from the 1858 observations, over 41% of companies declared zero tax revenue. This expressive proportion suggests a potentially high tax evasion rate within the sample.

Concerning the independent variables, this descriptive analysis shows that INTENSITY presents a range of values with significant variability, from −7.03 to 6.84, with a mean value of around −2.44 and a standard deviation of 1.447. These values reveal the significant variability concerning the intensive use of stablecoin in the sample. The mean (−2.44) of INTENSITY is negative, demonstrating that, on average, the volume of stablecoin transactions relative to bank deposits is less than 1 on a logarithmic scale. The median (−2.55) being close to the mean demonstrates a symmetrical distribution, and the negative IQR (−3.325 to −1.71) shows that most companies operate with lower transaction volumes using stablecoin than their volumes of bank deposits. The wide range (−7.03 to 6.84) confirms that there are outliers with very high or very low stablecoin activity relative to bank deposits.

Regarding FREQUENCY, the distribution of stablecoin transactions is almost equal between the high and low frequencies. However, low-frequency transactions make up a slight majority (51.2%).

In the context of ON_OFFCHAIN, more than half of the companies (55.4%) engage in IntraVasp and P2P transactions, indicating a preference for diversified transaction methods. Those exclusively using IntraVasp or P2P are significantly fewer, suggesting a tendency towards mixed transactional behaviour.

Concerning AGE, the range from 0 (newly established companies) to 3.91 log years indicates various company ages. The variable has a mean slightly above 1.5 log-years, indicating that the average company in the sample is relatively young, as it is the logarithm of age in years. The distribution is positively skewed since the mean is greater than the median (1.39), suggesting that some relatively older companies are in the sample.

Regarding the variable SECTOR, most companies (65.9%) operate in the commercial sector. This may be due to the sector’s significant size within the overall economy or a greater inclination of commercial companies to use stablecoins.

Regarding SIZE, more than half of the companies (56.8%) are classified as small based on bank deposits. This indicates that the sample is skewed towards smaller enterprises.

5.2. Model Results Presentation and Discussion

5.2.1. Model Validation

This section presents the results of the statistical tests for the Generalized Linear Model (GLM) logistic regression to validate the model’s estimates and aspects related to the model’s efficiency and multicollinearity. In this sense,

Table 3. Generalised linear model (binomial distribution and logit link).

Independent Variables	OR	CI 95%	p-Value	VIF
Intercept	0.872	0.599; 1.270	0.476
INTENSITY	1.115	1.027; 1.210	0.009	1.18
AGE	0.544	0.481; 0.615	<0.001	1.03
FREQUENCY:				1.15
High—Low	1.242	0.988; 1.561	0.064
ON_OFFCHAIN:				1.03
P2P—IntraVasp	1.253	0.918; 1.709	0.155
IntraVaspandP2P— IntraVasp	1.661	1.289; 2.139	<0.001
SECTOR:				1.04
Service—Commercial	1.510	1.218; 1.873	<0.001
SIZE:				1.09
Medium—Small	2.349	1.890; 2.920	<0.001
Pseudo R2Nagelkerke = 0.182
χ2(7) = 269, p < 0.001

Note. Estimates represent the log odds of ‘Evasion = Yes’ vs. ‘Evasion = No’; OR: odds ratio; CI: confidence interval. In all inferences, a significance level of 5% was considered.

presents the results of the statistical tests carried out for the model.

The model as a whole is statistically significant (χ²(7) = 269, p < 0.001). The model’s Chi-squared value is highly significant, indicating that the set of predictors together are useful in distinguishing between companies that do and do not evade tax. The Nagelkerke’s Pseudo R-squared value of 0.182, while not particularly high, is typical for logistic regression models in social sciences and suggests that the model explains a modest but meaningful proportion of the variance in tax evasion.

Another perspective on analysis is the absence of multicollinearity. It was checked by calculating the Variance Inflation Factor (VIF). Table 3 summarises the VIF for all independent variables in this study, indicating no multicollinearity between the variables for this model.

Thus, considering the joint analysis of the statistical tests performed, including the VIF check, the Chi-squared value, and the Pseudo R-squared, the results suggest that this model explains the dependent variable for this GLM logistic regression in the context of these sample data. However, the model does not necessarily imply causation, instead revealing patterns that could warrant further investigation by policymakers and economists.

5.2.2. Empirical Model Results and Discussions

This section seeks to answer the research hypotheses based on the estimation results of the GLM models. The objective was to evaluate the relationship between the binary dependent variable EVASION and the independent variables related to the companies’ characteristics that used stablecoin in 2021. The predictors in the model provide insight into how different attributes of companies correlate with their likelihood of tax evasion. In reporting the logistic regression analysis findings according to Table 3, we interpreted the significance of each variable to address the stated hypotheses as follows.

The companies’ engagement in stablecoin transactions relative to their bank deposits, measured by INTENSITY, has a statistically significant positive association with the odds of tax evasion. With each log-unit increase in INTENSITY, the odds of a company evading taxes (declaring zero tax revenue) increase by 11.5% (p = 0.009). Given that the p-value is below 0.05, we are confident that this result is not due to random chance. Hypothesis H1 posits that a significant positive relationship exists between the intensity of stablecoin transactions (INTENSITY) and the likelihood of a company engaging in tax evasion. The results from the regression analysis support this hypothesis. Thus, Hypothesis H1 is accepted. This relationship suggests that as companies integrate stablecoin transactions more intensively into their financial activities, the complexity and opacity of these transactions may provide greater opportunities or incentives to minimise tax liabilities through evasion. The positive relationship with tax evasion could be explained by the anonymity and lesser regulatory oversight often associated with crypto assets, which may facilitate tax evasion behaviours. Economically, companies with higher stablecoin transactions might seek to minimise transaction costs or bypass certain financial restrictions, which could correlate with a propensity to under-report revenues.

Concerning the FREQUENCY factor, companies with a high number of transactions have 24.2% higher odds of evading tax than those with a low number of transactions. However, this result is only marginally significant (p = 0.064). This finding leads us to reject H2. High-frequency stablecoin transactions may indicate greater risk appetite or a more assertive approach to financial management, which could align with a higher willingness to engage in tax evasion.

The ON_OFFCHAIN variable represents the transaction methods used by companies, with three possible categories. For Hypothesis H3a, the results indicate that companies engaging in both off-chain (IntraVasp) and on-chain (P2P) transactions (IntraVaspandP2P) are significantly more likely to evade tax, with an odds ratio of 1.661 and a p-value of less than 0.001. This finding leads us to accept H3a. The significant positive relationship here might be explained by the increased complexity and reduced traceability when companies engage in centralised and decentralised financial transactions. Companies using VASPs and P2P might have more opportunities to obscure financial transactions, potentially facilitating tax evasion.

However, the results do not support Hypothesis H3b, which suggests that companies transacting exclusively P2P transactions have a higher propensity for tax evasion than those who engage exclusively in IntraVasp transactions. The odds ratio for P2P compared to IntraVasp was 1.253 with a p-value of 0.155, indicating no statistically significant difference. Thus, H3b is rejected. The lack of significance could be due to a smaller sample size or less pronounced effects. Although not significant, the positive direction of the relationship suggests that P2P transactions, known for their decentralisation and potential anonymity, might theoretically be associated with tax evasion.

Hypothesis H4 suggests an association between company age (AGE) and the propensity to engage in tax evasion. The results confirm this relationship. As a company’s age increases, the odds of evading tax decrease substantially (about 45.6% for each unit increase in the logarithm of age), suggesting that older companies are less likely to evade taxes. This is statistically significant (p < 0.001). Therefore, Hypothesis H4 is accepted. These results corroborate those of Pedroni et al. (2022) and Wang (2012). Pedroni et al. (2022) and Wang (2012) found that companies that under-report sales (commit tax evasion) have a negative correlation with company age. Older companies may have more established financial processes and greater scrutiny from tax authorities, leading to a negative association with tax evasion. They may also have a reputation to maintain, which tax evasion allegations could severely damage.

Hypothesis H5 states that service sector companies have a higher incidence of tax evasion than those in the commercial sector. The regression results show a significant difference between the sectors, with service companies having a 51% increase in the odds of tax evasion with a p-value of less than 0.001. This supports the alternative hypothesis; therefore, H5 is accepted. These results corroborate Abdixhiku et al. (2017) and Pedroni et al. (2022). Abdixhiku et al. (2017) found that firms operating in business services are more evasive than wholesale and retail firms. Pedroni et al. (2022) also found that service companies have a greater propensity to under-report sales (a type of tax evasion) than commercial companies. Service sectors may have a higher tendency for tax evasion due to the nature of their transactions, which are often less tangible than commercial sectors dealing in physical goods. The invisibility of services can make revenues harder to track. Economically, the service sector often has more flexibility in accounting for income and expenses, which could be exploited for tax benefits.

Finally, Hypothesis H6 addresses the relationship between company size and tax evasion, suggesting that medium-sized companies would be more likely to evade tax than small companies. The odds ratio for SIZE (Medium vs. Small) was 2.349, with a p-value of less than 0.001, which strongly indicates that medium-sized companies are more likely to engage in tax evasion than small companies. Hence, H6 is accepted. The statistical significance indicates a robust size effect, with larger companies in the defined categories more prone to tax evasion. These results corroborate those of Ngah et al. (2020) and Mohamad et al. (2016) but are contrary to those of Azrina Mohd Yusof et al. (2014), Pedroni et al. (2022), and DeBacker et al. (2015). Ngah et al. (2020) also found that company size has a statistically significant positive relationship with tax evasion practices detected during tax audits. Similarly, Mohamad et al. (2016) found that SME taxpayers are more likely to become non-compliant with their tax obligations as their businesses expand. Oppositely, Azrina Mohd Yusof et al. (2014) found an adverse relationship between company size and SME tax non-compliance; as SME size grows, the SMEs become more compliant. In addition, Pedroni et al. (2022) found that the size factor is negatively associated with revenue under-reporting. Based on data from the IRS, DeBacker et al. (2015) found that corporations owned by individuals from countries with higher corruption norms tend to avoid more taxes in the United States. This trend is particularly noticeable among small corporations and becomes less significant as the corporation’s size increases.

6. Conclusions

We examined the relationship between the use of stablecoins and tax evasion behaviour among companies based on Brazilian Federal Revenue data through a logistic regression model that predicts tax evasion, considering various operational characteristics such as transaction intensity, frequency, environment (on-chain and off-chain), company age, sector, and size. The results suggest that INTENSITY, AGE, ON_OFFCHAIN (specifically IntraVaspandP2P), SECTOR, and SIZE are all significant predictors of tax evasion, as defined in this context. However, the regression analysis provided no support for the hypothesis that firms engaging in only P2P transactions have a higher propensity to evade tax than those engaging only in IntraVasp transactions; this hypothesis was rejected due to a lack of significant findings. In summary, the analysis of the results revealed that younger and medium-sized companies with higher stablecoin transaction intensity that engage in IntraVasp and P2P transactions and belong to the service sector are likelier to evade taxes.

The main limitation of this study is the sample selection criteria. The selection of companies that have used stablecoins and have bank deposits ranging between USD 100,000 and USD 20 million ensures a focus on entities with substantive economic activities and financial transactions. Considering this range as criteria for selection, the sample includes companies that are neither too small to be negligible nor so large that they are outliers. However, this excludes the extremes of the economic spectrum—microenterprises and large corporations—which could have distinct approaches to tax compliance and evasion. Furthermore, by including companies from the commerce and service sectors, this study recognises the sectors where stablecoin use is potentially more relevant. While this provides insight into the tax behaviours of companies in these sectors, it might also skew the findings towards sector-specific practices and norms, which might not apply to industries like manufacturing or agriculture. The sample, therefore, might miss out on evasion patterns unique to these excluded groups.

Further investigation and future research could examine the qualitative aspects of why companies with different transaction frequencies might exhibit varying tax behaviours, including the role of internal controls, the impact of regulatory awareness, and the influence of corporate culture on compliance attitudes. We intend to expand the sample from 2019 (the beginning of the collecting information about stablecoin transactions in Brazil) until 2023 and include other variables so that the analysis of the results may correspond as closely as possible to reality regarding the relationship between the use of stablecoin and tax evasion. Additionally, advanced machine learning techniques could be employed to identify and predict tax evasion propensity.