AN EMPIRICAL ANALYSIS

OF ANTITRUST FINES IN

BRAZIL (2012–2020)

Uma Análise Empírica das Multas Antitruste no

Brasil (2012–2020)

Lucas Campio Pinha

Universidade Federal Rural do Rio de Janeiro (UFRRJ) – Três Rios/RJ, Brasil

Marcelo José Braga

Universidade Federal de Viçosa (UFV) – Viçosa/MG, Brasil

Luiz Alberto Esteves

Banco de Desarrollo de América Latina y el Caribe

STRUCTURED ABSTRACT

Background: the antitrust enforcement remains a primary concern for regulatory authorities globally.

A central issue within this domain is the determination of appropriate penalties for violators, as

antitrust ﬁnes serve both to deter future violations and to disrupt ongoing anticompetitive behaviors.

Objective: utilizing a comprehensive, hand-collected database of antitrust cases from 2012 to 2020,

this study aims to empirically examine the factors that are associated with the imposition of corporate

antitrust ﬁnes in Brazil.

Method: econometric estimates based on Ordinary Least Squares and Random-Intercept Model.

Conclusions: the number of subsections within Law No. 12.529/11, the duration of the infringement,

the number of corporations involved, and the type of corporation (with traditional companies

facing higher ﬁnes) are all positively associated with the severity of ﬁnes. Additionally, the scope

of the market at international and national levels also demonstrates a statistically signiﬁcant

positive relationship with the ﬁnes imposed. Cases involving the fuel sector are also associated

with higher penalties, suggesting the presence of sector-speciﬁc enforcement patterns. The ﬁndings

of this study provide valuable insights for both Brazilian policymakers and global stakeholders,

contributing to the enhancement of deterrence mechanisms by amplifying the perceived costs of

anticompetitive conduct.

Keywords: antitrust; antitrust ﬁnes; Cade; antitrust authority; random-intercept model.

1 Editora responsável: Profa. Dra. Camila Cabral Pires-Alves, Conselho Administrativo de Defesa Econômica (Cade),

Brasília/DF, Brasil. Lattes: https://lattes.cnpq.br/4687008805056384. ORCID: https://orcid.org/0000-0001-7888-3235.

Recebido em: 20/11/2025 Aceito em: 20/05/2026 Publicado em: 30/06/2026

2 Doutor em Economia Aplicada pela UFV. Professor Adjunto da UFRRJ – Campus Três Rios. E-mail: lucascpinha@ufrrj.

br Lattes: http://lattes.cnpq.br/6163773433756591. ORCID: https://orcid.org/0000-0003-1521-6747

3 Doutor em Economia Aplicada pela UFV. Professor Titular da UFV. E-mail: mjbraga@ufv.br. Lattes: http://lattes.cnpq.

br/0107443653772269. ORCID: https://orcid.org/0000-0002-8161-405X.

4 Doutor em Economia pela Università degli Studi di Siena, Itália. Consultor no Banco de Desarrollo de América Latina

y el Caribe. E-mail: estevesufpr@gmail.com. Lattes: http://lattes.cnpq.br/0106128017806941. ORCID: https://orcid.org/0000-

0002-4182-0983.

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

RESUMO ESTRUTURADO

Contexto: a defesa da concorrência constitui uma das principais preocupações das agências

reguladoras ao redor do mundo. Uma das questões centrais neste contexto é a determinação de

penalidades apropriadas para infratores, uma vez que as multas servem tanto para dissuadir futuras

infrações quanto para desencorajar comportamentos já existentes.

Objetivo: utilizando uma base de dados de casos coletados entre 2012 e 2020, o presente estudo visa

examinar empiricamente os fatores associados com a imposição de multas antitruste corporativas

aplicadas no Brasil pelo Conselho Administrativo de Defesa Econômica.

Método: estimativas econométricas baseadas em Mínimos Quadrados Ordinários e Modelo de

Intercepto Aleatório.

Conclusões: o número de incisos da Lei nº 12.529/11, a duração da infração, o número de corporações

envolvidas e o tipo de corporação (empresas tradicionais apresentando multas maiores) são fatores

positivamente associados com a severidade da multa aplicada. Em adição, infrações com escopo

internacional e nacional também demonstraram signiﬁcância estatística e sinal positivo em relação às

multas. Casos envolvendo o setor de combustíveis também estão associados positivamente a maiores

penalidades, sugerindo a presença de padrões de atuação em setores especíﬁcos. Os resultados

do estudo fornecem informações importantes tanto para formuladores de política pública no Brasil

quanto para interessados ao redor do mundo, o que contribui para ampliar o efeito dissuasivo por

elevar os custos percebidos da infração.

Palavras-chave: defesa da concorrência; multas antitruste; autoridade antitruste; modelo de

intercepto aleatório.

Classiﬁcação

JEL: L40; C40; K21.

Summary: 1. Introduction; 2. Corporate antitrust ﬁnes in

Brazil; 3. Methodology; 3.1. Data; 3.2. Empirical model; 4.

Results and discussion; 5. Conclusions; References.

1 INTRODUCTION

In certain occasions, ﬁrms may deliberately engage in practices that undermine competition

within markets. Addressing these anticompetitive behaviors constitutes a primary objective of

antitrust laws enacted by governments. Examples of such practices include collusive agreements,

wherein competitors coordinate their actions to restrict competition (Harrington, 2017); exclusionary

practices, where ﬁrms with dominant market positions take actions to stiﬂe competition; vertical

restraints, which involve agreements between parties operating at dierent levels of the supply

chain; and coordinated actions by unions or associations to inﬂuence market behavior (Cade,

2016a). Additionally, new forms of anticompetitive conduct have emerged in digital economies and

platforms. What unites these practices is their shared aim to restrict or eliminate present and future

competition, thereby facilitating the generation of supracompetitive proﬁts. These inﬂated proﬁts are

oten achieved to the detriment of consumers, who face higher prices, diminished product quality,

reduced variety, and other ineciencies inherent in markets plagued by antitrust violations.

In Brazil, the antitrust authority (Conselho Administrativo de Defesa Econômica – Cade, in

Portuguese) started to focus on antitrust issues ater the Law nº 8.884/1994, which was enacted

to regulate the Brazilian Competition Defense System. However, the major concern in the 90s was

mergers and acquisitions due the economic openness faced by Brazil in that period. The ﬁght

against anticompetitive conducts started eectively in the 2000s and has been improving over the

years, including the following actions: upgrade of the Brazilian Leniency Program adopted in the

year of 2000; public ﬁght against gas station cartels; public ﬁght against anticompetitive conducts

in healthcare systems; punishment of international hardcore cartels; improvement of the legal

and economic teams; enactment of Law nº 12.529/2011, which have been regulating the Brazilian

Competition Defense System since 2012.

From 2014 onwards, Operation Car Wash (Operação Lava

Jato) also played a signiﬁcant role in strengthening antitrust enforcement in Brazil. The investigations

uncovered several bid-rigging cartels in public procurement, oten linked to corruption schemes, and

reinforced the use of Cade’s leniency program as a key detection tool. This context contributed to a

more intensive and coordinated ﬁght against cartels in the country (Munhoz; Oliveira, 2020).

An important element in this context is the imposition of penalties for anticompetitive

conduct. Antitrust sanctions have two primary eects: deterrence of future violations and cessation of

ongoing infringements. As noted by Spagnolo (2008), ex-ante deterrence, achieved through the threat

of suciently severe sanctions, is the most critical objective, as it has the potential to prevent a wide

range of potential violations at relatively lower social and individual costs. Additionally, penalizing

detected infringements plays a vital role in prompting oenders to cease their anticompetitive

behavior, thereby contributing to the overall ecacy of antitrust enforcement.

When discussing monetary penalties, the literature on optimal criminal ﬁnes emphasizes

some crucial considerations. Regarding the deterrence eect, Becker’s (1968) foundational work in

the economics of crime suggests that oenders weigh the marginal beneﬁts of committing a crime

against the marginal costs, with the probability of detection and the anticipated penalties forming key

components of those costs. In this context, to eectively deter infringements, optimal ﬁnes should

be calibrated to align with this decision-making process, ensuring that the costs of committing an

antitrust violation exceed the potential beneﬁts. On the other hand, when addressing the punishment

of actual infringements, Wils (2006) outlines several factors that antitrust authorities should take into

account: the ﬁnancial capacity of oenders to pay the ﬁnes, the broader social and economic costs

associated with the imposition of penalties (such as the impact on corporate structure, employees,

stock market performance, and investments), and the proportionality of ﬁnes relative to the illegal

gains derived from the infringement. It is important to note that the objective of imposing ﬁnes is

not to bankrupt the oending corporation but rather to encourage the cessation of anticompetitive

conduct. Bankrupting a ﬁrm could result in undesirable market concentration. For instance, in a

duopoly where two ﬁrms are engaged in collusion, if one ﬁrm goes bankrupt, the surviving ﬁrm may

become a monopolist, which would lead to a worse market concentration scenario in the future,

further exacerbating competitive harm.

In this context, it is essential to understand which factors are associated to the level of ﬁnes

imposed to antitrust oenders. Despite the guidelines provided by antitrust agencies, such as European

Comission (2011) and United States Sentencing Commission (2020), penalizing antitrust conducts is

5 Further details on the history of antitrust enforcement in Brazil can be found in Todorov and Torres Filho (2012).

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

not an exact science. In the United States, for example, the prevalence of plea agreements introduces

a subjective element into the penalty-setting process. According to Law nº 12.529/2011 (Brasil, 2011),

in Brazil ﬁnes are supposed to be proportional to the revenue obtained and the aggravating factors,

but the necessary data on revenue and aggravating circumstances is not always readily available (or

it is not available to Cade, or it is restricted to the public). Considering these challenges, conducting

an empirical analysis to identify the primary factors inﬂuencing antitrust ﬁnes becomes an essential

task. Such an analysis can provide a deeper understanding of the rationale behind penalty decisions

and enhance the transparency and eectiveness of antitrust enforcement.

The objective of this paper is to empirically analyze which factors are associated to corporate

antitrust ﬁnes in Brazil. Since the rules for individuals are distinct, the focus is on corporations that

participated in anticompetitive conducts. We rely on a hand-collected database that contains all

antitrust cases ater the implementation of Law nº 12.529/2011 where at least one corporation was

punished regularly with monetary penalties.

This paper makes a threefold contribution to the existing literature. First, from an international

perspective, it is among the ﬁrst studies to focus on an emerging economy – Brazil – where eective

antitrust enforcement only began in the last decade of the twentieth century. With the exception of

Jing, Gong and Yi. (2020), who focused on China, there is a limited body of work exploring antitrust

ﬁnes in emerging countries (and China and Brazil dier greatly regarding the economy, institutions,

politics, industrialization, among many other aspects). Second, in Brazil, the ﬁndings oer valuable

insights for Cade’s team, researchers, legal professionals, and other interested parties by informing

discussions on deterrence and the setting of antitrust ﬁnes, that is, which factors judges tend to

consider most relevant when deﬁning sentences. Third, from a methodological standpoint, while

much of the existing literature employs statistical techniques and regressions that do not account for

group similarities, this study applies a random-intercept model, a more robust method that controls

for group-level variations through random coecients, providing a more nuanced understanding of

the factors inﬂuencing antitrust penalties.

From the perspective of enforcement eciency, the identiﬁcation of statistically signiﬁcant

drivers of ﬁnes also contributes to improving Cade’s transparency and predictability. If economic

agents can anticipate how ﬁnes are likely to be calculated, deterrence is strengthened ex ante,

thereby increasing the authority’s eciency in reducing cartel activity (Spagnolo, 2008). Furthermore,

the divergence of results regarding cartel duration compared to studies of U.S. and EU ﬁnes (Connor

& Miller, 2009, 2013) highlights a contextual dierence in Brazil’s enforcement, which can inform

the comparative literature on whether emerging economies converge or diverge from the “optimal

ﬁnes” benchmark.

This paper contributes to the existing body of literature on the statistical and econometric

analysis of antitrust sanctions. Notable foundational works in this ﬁeld include Posner (1970) and

Gallo, Craycrat and Dutta (1986) for the United States, while more recent studies include Bolotova and

Connor (2008), Connor and Miller (2009), and Connor and Miller (2013), which examine international

cartels penalized in the United States and the European Union. Additionally, two signiﬁcant studies—

Allain et al. (2015) and Jing, Gong and Yi (2020)—investigate whether actual antitrust ﬁnes in the

European Union and China, respectively, align with the theoretical framework of optimal sanctions.

The remainder of the paper is organized as follows. Section 2 discusses the law governing

corporate antitrust ﬁnes in Brazil, while Section 3 contains the methodology used in the paper.

Section 4 presents the results and discussion. Conclusions are provided in Section 5, followed by

the references.

2 CORPORATE ANTITRUST FINES IN BRAZIL

As previously mentioned, despite former antitrust laws, the Law nº 8.884/94 in 1994 was

the ﬁrst one to establish the Brazilian System of Competition Policy, which was replaced by Law nº

12.529/11 that is in force from 2012 until now

. Article 36 and the subsequent subsections of Law nº

12.529/11 establish the following conducts as subject to punishment:

Art. 36. The acts under any circumstance, which have as object or may have the

following eects shall be considered violations to the economic order, regardless of

fault, even if not achieved:

I – to limit, restrain or in any way injure free competition or free initiative;

II – to control the relevant market of goods or services;

III – to arbitrarily increase proﬁts, and

IV – to abusively exercise a dominant position.

§ 1 The conquest of the market resulting from the natural process of the most ecient

economic agent in relation to its competitors does not characterize the tort set forth

in item II of the caput of this article.

§ 2 A dominance position is assumed when a company or group of companies is able

to unilaterally or jointly change market conditions or when it controls 20% (twenty

percent) or more of the relevant market, provided that such percentage may be

modiﬁed by Cade for speciﬁc sectors of the economy.

§ 3 The following acts, among others, to the extent in which they conﬁgure the

hypothesis set forth in the caput of this article and items thereof, shall characterize

violation of the economic order:

I – to agree, join, manipulate or adjust with competitors, in any way:

a) the prices of goods or services individually oered;

b) the production or sale of a restricted or limited amount of goods or the provision

of a limited or restricted number, volume or frequency of services;

c) the division of parts or segments of a potential or current market of goods or

services by means of, among others, the distribution of customers, suppliers, regions

or time periods;

d) prices, conditions, privileges or refusal to participate in public bidding;

II – to promote, obtain or inﬂuence the adoption of uniform or agreed business

practices among competitors;

III – to limit or prevent the access of new companies to the market;

6 The changes from Law nº 8.884/94 to Law nº 12.529/11 related to antitrust infringements are in the sense of reor-

ganizing subsections, changing penalties ranges and establishing self-report policies that were already adopted (Brazilian

Leniency Program and Cease and Desist Agreements). For this reason, we focus this Section on the current Law in Brazil.

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

IV – to create diculties for the establishment, operation or development of a

competitor company or supplier, acquirer or ﬁnancier of goods or services;

V – to prevent the access of competitors to sources of input, raw material, equipment

or technology, and distribution channels;

VI – to require or grant exclusivity for the dissemination of advertisement in mass

media;

VII – to use deceitful means to cause oscillation of the prices practiced by third parties;

VIII – to regulate markets of goods or services by establishing agreements to limit

or control the research and technological development, the production of goods or

services, or to impair investments for the production of goods or services or their

distribution;

IX – to impose, on the trade of goods or services, to distributors, retailers and

representatives, resale prices, discounts, payment terms, minimum or maximum

quantities, proﬁt margin or any other market conditions related to their business

with third parties;

X – to discriminate against purchasers or suppliers of goods or services by establishing

price dierentials, or operating conditions of sale or provision of services;

XI – to refuse the sale of goods or provision of services, within regular payment

conditions to the business practices and customs;

XII – to hinder or disrupt the continuity or development of business relationships

of undetermined term, because the other party refuses to abide by unjustiﬁable or

anticompetitive terms and conditions;

XIII – to destroy, render useless or monopolize the raw materials, intermediate or

ﬁnished products, as well as to destroy, disable or impair the operation of equipment

to produce, distribute or transport them;

XIV – to monopolize or prevent the exploitation of industrial or intellectual property

rights or technology;

XV – to sell goods or services unreasonably below the cost price;

XVI – to retain production or consumption goods, except for ensuring recovery of

production costs;

XVII – to partially or totally cease the activities of the company without proven just

cause;

XVIII – to condition the sale of goods to the acquisition of another or use of a service,

or to condition the provision of a service to another or to the acquisition of goods.

XIX – to abusively exercise or exploit intellectual or industrial property rights,

technology or trademark (Brasil, 2011).

As will be detailed later, based on the subsections above the antitrust cases contained

in the sample can be divided in some big groups: collusive agreements between competitors in

ordinary markets and/or procurements; inﬂuence of uniform conducts by unions, associations and/

or cooperatives; collusive agreements between competitors with inﬂuence of uniform conducts; and

abuse of dominant position.

Article 37 of Law nº 12.529/11 establishes the following penalties for antitrust offenders

in Brazil:

Art. 37. A violation of the economic order subjects the ones responsible to the following

penalties:

I – in the case of a company, a ﬁne of one tenth percent (0.1%) to twenty percent

(20%) of the gross sales of the company, group or conglomerate, in the last ﬁscal

year before the establishment of the administrative proceeding, in the ﬁeld of the

business activity in which the violation occurred, which will never be less than the

advantage obtained, when possible the estimation thereof;

II – in the case of other individuals or public or private legal entities, as well as

any association of persons or de facto or de jure legal entities, even if temporary,

incorporated or unincorporated, which do not perform business activity, not being

possible to use the gross sales criteria, the ﬁne will be between ﬁty thousand reais

(R$ 50,000.00) to two billion reais (R$ 2,000,000,000.00);

III – if the administrator is directly or indirectly responsible for the violation, when

negligence or willful misconduct is proven, a ﬁne of one percent (1%) to twenty percent

(20%) of that applied to the company, in the case set forth in Item I of the caput of this

article, or to legal entities, in the cases set forth in item II of the caput of this article.

§ 1 – In case of recurrence, the ﬁnes shall be doubled;

§ 2 – In the calculation of the value of the ﬁne referred to in item I of the caput of this

article, Cade may consider the total turnover of the company or group of companies,

when the value of sales in the ﬁeld of business activity in which the violation occurred

is not available, deﬁned by Cade, or when it is incompletely presented and/or not

clearly and credibly demonstrated (Brasil, 2011).

Observe that Art. 37 establishes the range of penalties, but the proportion and the ﬁne depend

on the following criteria contained in Art. 45:

Art. 45. In the application of the penalties set forth in this Law, the following shall be

taken into consideration:

I – the seriousness of the violation;

II – the good faith of the transgressor;

III – the advantage obtained or envisaged by the violator;

IV – whether the violation was consummated or not;

V – the degree of injury or threatened injury to free competition, the national economy,

consumers, or third parties;

VI – the negative economic eects produced in the market;

VII – the economic status of the transgressor;

VIII – any recurrence (Brasil, 2011).

In this context, we can conclude that penalties setting vary case-by-case and oender-by-

oender. First, because the outcome that the ﬁne proportion will aect (deﬁned in Art. 37, paragraph

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

I) can be distinct, and in some cases companies do not disclose outcomes, which require another

type of criteria used by Cade, such as a similar company, a similar case, or a penalty based on the

reasonability. Second, because the aggravating factors listed in Art. 45 are subjective

. As better

explained later in the paper, our dependent variable is the value of the penalty set for each ﬁrm,

and covariates are related to this penalty or by inﬂuencing the ﬁrm revenue or by interfering on the

aggravating factors, or both

Despite the criteria above, the subjectivity of antitrust ﬁnes in Brazil reinforces the importance

of an econometric analysis to understand which factors are associated with penalties, which may

contribute to the literature and may increase the deterrence of anticompetitive conducts. Regarding

other jurisdictions, Connor and Miller (2009) found that ﬁnes imposed on international cartels by

the Department of Justice (DOJ) in United States are positively related to the economic injuries of

collusion, at the same time that ﬁnes complement other antitrust penalties: prison sentences and

private damages paid. The rest of coecients were non-signiﬁcant and/or showed a sign contrary

to the expected, as the ones related to leadership, bid-rigging cartels, and the duration of the

infringement (which may indicate that long-lasting cartels receive higher ﬁnes through the harm

caused). Similarly, Connor and Miller (2013) analyzed the European Commission ﬁnes for global price-

ﬁxing and found that monetary penalties are directly related to economic injuries, recidivism, and

having a whistleblower with immunity in the case, while inversely related factors that increase the

probability of detection and conviction. Even though many of our variables dier from the mentioned

papers due to available information, our results increase the comprehension of the subjective criteria

of Cade in Brazil and allow the comparison with other jurisdictions in which is comparable.

3 METHODOLOGY

3.1 Data

As mentioned in the previous Section, antitrust ﬁnes for companies in Brazil are primarily set

based on the company’s sales and the aggravating factors

. For other types of corporations (public and

private legal entities) and when company’s sales are not available, ﬁnes must respect the maximum

of ﬁty thousand Reais up to two billion Reais, considering the aggravating factors. A challenge in our

paper is that Cade does not disclosure market information when lawyers require for secrecy, which

prevents access to important control variables of corporations. It occurred many times in the sample,

thus we needed to ﬁnd other ways to control for factors that inﬂuence monetary penalties imposed

by Cade.

The dependent variable in our paper is the ﬁne applied by Cade to each corporation. Control

variables comprise corporation-level covariates and case-level covariates, all collected in the antitrust

cases judge by Cade through the electronic system of information (in Portuguese, Sistema Eletrônico

7 Cade recently released a guideline for cartel ﬁnes setting that suggests objective criteria for calculating penalties

based on Cade previous decisions. However, this document is focused on cartels and has a suggestive character. Even with the

goal of turning the ﬁne setting more objective, possibly this guideline will not remove completely the subjectivity of decisions.

The document in Portuguese can be accessed in Cade (2023).

8 Unfortunately, aggravating factors are not always listed, and we could not access ﬁrms’ revenues due legal aspects.

9 When gross sales are not available, Cade may use another criterium, such as the percentage of ﬁne for a similar

company in the same case.

de Informações – SEI)

. As aforementioned, our sample consists of a hand-collected database that

contains all antitrust cases judged ater the implementation of Law nº 12.529/2011 where at least one

corporation was punished regularly with monetary penalties, totalizing 597 corporations judged in 146

antitrust cases. The ﬁnal date considered in the paper is December 2020. The Table 1 below presents

the variables information:

Table 1 - Variables names and description

Name Description

Fines

(corporation-level)

Fines set by Cade to each corporation, in Reais and in values of December 2020*.

Subsections

(corporation-level)

Number of subsections of Law n° 12.529/2011, or equivalent if the corporation was

judged by Law n° 8.884/1994**.

Company

(corporation-level)

Dummy variable with value one if the corporation is an ordinary company, zero

otherwise (union, association, or another legal entity).

Corporations

(case-level)

Number of corporations involved in the infringement.

Duration

(case-level)

Duration of the infringement, in years.

Leniency

(case-level)

Dummy variable with value one if there was at least one leniency agreement in the

case, zero otherwise

TCC

(case-level)

Dummy variable with value one if there was at least one Cease and Desist

Agreements (in Portuguese, Termo de Compromisso de Cessação – TCC) in the case,

zero otherwise.

International

(case-level)

Dummy variable with value one if the market aected by the infringement is

international, zero otherwise***.

National

(case-level)

Dummy variable with value one if the market aected by the infringement is

national, zero otherwise. Zero value means an infringement with a scope of

operation smaller than the state level***.

State

(case-level)

Dummy variable with value one if the market aected by the infringement is state-

level, zero otherwise***.

Health

(case-level)

Dummy variable with value one for cases in the health sector, zero otherwise****.

Fuel

(case-level)

Dummy variable with value one for cases in the fuel sector, zero otherwise****.

Year dummies

(case-level)

Dummy variable with value one if the case was judged in the respective year, zero

otherwise.

Source: data collected by the authors (2021).

* Note: Values are adjusted by the Brazilian basic interest rate, SELIC. This is the same procedure

done by Cade to adjust monetary values.

** Note: Corporations may be judged under the previous law if the case was initiated prior to the

enactment of the new law, and if Cade deems it more advantageous to the defendant.

*** The reference category (zero value) means an infringement with a scope of operation smaller

than the state level, for example, at the municipal level or lower.

**** Both categories were selected given their importance in the sample, as will be discussed later.

10 The system can be accessed in the following link: https://sei.cade.gov.br/sei//controlador_externo.php?acao=usuario_

externo_logar&id_orgao_acesso_externo=0

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

Some further details on the data are worth noting. Regarding “subsections”, Article 36 and

the subsequent subsections of Law No. 12.529/11 deﬁne the conducts that are subject to penalties,

including practices associated to collusive practices, inﬂuence of uniform conduct and abuse of

dominant position. Corporations that applied for leniency or TCC agreements were excluded from

the sample, since the logic of ﬁnes are distinct

, but they are included in the number of the variable

“corporations”. Regarding the variable “duration”, the number of years was deﬁned based on Cade

deﬁnition and/or case information. Lastly, our database consists of a cross-section of 597 corporations

judged in distinct periods, where speciﬁc factors of time are controlled by year dummy variables. The

empirical model described below allows for the identiﬁcation of correlations between covariates and

the dependent variable

3.2 Empirical model

In this paper, each convicted corporation represents a unit of observation, with these

convictions being associated with distinct antitrust cases. It is plausible that corporations convicted

within the same case share common characteristics, suggesting the presence of factors within each

group that may inﬂuence the determination of ﬁnes and, consequently, should be accounted for in

the estimation process. While one might consider the inclusion of group-speciﬁc dummies to capture

these eects, the sample consists of 146 antitrust cases, and such an approach could compromise

the quality of the estimates. To address this concern, we employ a multilevel model in our analysis.

Multilevel models (also referred to as hierarchical models) are particularly appropriate when

data is structured hierarchically, meaning that observations represent units clustered at dierent

levels (Goldstein, 2011). According to Steenbergen and Jones (2002), the primary advantage of multilevel

models over traditional regression methods lies in their ability to account for the natural nesting of

data. In these models, the variability among individual units is explained not only by factors at the

individual level, as in standard regression, but also by characteristics at the group level. A classic

example of this is the analysis of student grades: while individual factors such as parents’ education,

family wealth, and personal dedication inﬂuence grades, school-level characteristics such as teacher

quality and institutional resources also contribute signiﬁcantly. In the context of our study, each

convicted corporation is treated as a unit nested within its respective antitrust case.

Multilevel models account for both unit-level and group-level factors through random

coecients (intercepts and/or slopes). In a random-intercept model, a random component is added to

the unit-level error term to capture the variation in the intercept across groups. This means that while

the overall regression intercept is ﬁxed, the intercept for each group can vary according to a random

component. In contrast, random slopes allow for random components at the group level that are

associated with the coecients of the predictor variables. The decision regarding which coecients

should be treated as random depends on the research question, the statistical signiﬁcance of the

random components, the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC),

and other case-speciﬁc considerations.

11 The Brazilian Leniency Program may provide full or partial amnesty of ﬁnes according to certain criteria, while TCC

may provide partial amnesty. Further details on these programs can be accessed in Cade (2016a) and Cade (2016b), respectively.

12 In simple terms, correlation (or association) measures the extent to which two variables move together. This type of

study does not account for causality.

In line with the previous discussion, we estimate two models: an Ordinary Least Squares

(OLS) model and a random-intercept model. We chose to include both models to highlight their

dierences and to provide robustness to our results; however, the primary focus of our analysis is

on the random-intercept model. Moreover, models are estimated in a log-linear form, meaning that

the variable Fines is in its natural logarithm and all other variables are in their natural form. This

functional form was chosen because it provides percentage variations, which is more interesting

than analyzing absolute variations in Brazilian currency. Coecients in log-linear models express an

approximated percentage variation on the dependent variable in its natural form in consequence of

a unit variation in regressors, but the exact percentage variation is expressed by 100*(e

βi

-1),i=1,2,…

. We provide both information on the results and discussion section. The OLS speciﬁcation can be

expressed as the following:

ln(Fines)

=β

+β

Subsections

+β

Company

+β

Corporations

+β

Duration

+β

Leniency

+β

TCC

+β

International

+β

National

+β

State

+β

Health

+β

Fuel

+β

Year2020

+β

Year2019

+β

Year2018

+β

Year2017

+β

Year16

+β

Year2015

+β

Year2014

+ε

(1)

Where β

is the intercept, β

,β

,… are variables coecients and ε

is the error term for each

condemned corporation of the sample, represented by i.

For the random intercept model, suppose that the intercept in (1) is speciﬁed as β

=γ

+μ

where γ

is the ﬁxed overall intercept and μ

is the random component at the antitrust case level

j , assumed independent, normally distributed with zero mean and constant variance. The random

intercept model can be expressed as the following:

ln(Fines)

=γ

+β

Subsections

+β

Company

+β

Corporations

+β

Duration

+β

Leniency

+β

TCC

+β

International

+β

National

+β

State

+β

Health

+β

Fuel

+β

Year2020

+β

Year2019

+β

Year2018

+β

Year2017

+β

Year16

+β

Year2015

+β

Year2014

+μ

+ε

(2)

In this model, intercepts are the same for corporations in the same antitrust cases but are

distinct between dierent cases according to μ

. Deﬁning the variance of μ

by τ

(between-group

variance) and the variance of ε

by σ

(within-group variance), the proportion of total variance that is

between-group can be expressed by the intraclass correlation coecient (ρ), deﬁned by the following:

(3)

This index is an important argument in favor (or not) of estimating a random-intercept model,

since a high value of ρ indicates that a great amount of the variance of corporations’ ﬁnes is related

to antitrust cases characteristics, which is neglected in the OLS regression.

We estimate the random intercept model using maximum likelihood, which Snijders and

Bosker (2012) identify as a primary method for estimating statistical parameters in random coecients

models. Following a Breusch-Pagan test for heteroscedasticity on the OLS regression, we found strong

evidence of heteroscedastic residuals. Consequently, both the OLS and random intercept models

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

are estimated with robust variance-covariance matrix estimations.

The estimates presented in the

following section indicate that the selection of variables is appropriate in terms of model ﬁt.

4 RESULTS AND DISCUSSION

Starting with a description of the sample, approximately 36% of the cases (53 antitrust cases)

involve collusive agreements between competitors in traditional markets and/or procurement sectors.

Around 39% of the cases (57 antitrust cases) pertain to the inﬂuence of uniform conduct by unions,

associations, and/or cooperatives. Approximately 12% of the cases (17 antitrust cases) represent a

combination of collusive agreements and the inﬂuence of uniform conduct, while the remaining 13%

(19 antitrust cases) involve abuse of dominant position.

Figure 1 below illustrates the number of cases and the number of convicted corporations for

each year within the period covered by this paper.

Figure 1 - Number of cases and number of convictions of antitrust cases in Brazil, by year

Source: own elaboration (2021).

Law No. 12.529/2011 has been in force in Brazil since July 2012. Notably, no antitrust cases were

condemned in 2012, but the number of condemnations increased signiﬁcantly in the subsequent three

years, before stabilizing at a lower level between 2016 and 2020. This suggests that the enforcement

of antitrust violations became more rigorous and eective following the implementation of the law.

Regarding the observed stability since 2016, Harrington and Chang (2009) point out a key challenge in

addressing cartels (and, more broadly, antitrust infringements): only discovered cases are observable,

meaning that the true extent of violations remains unknown. Therefore, a low rate of condemnations

could indicate either ineective enforcement or a reduced number of infringements. The latter

13 Further details are available in the Stata User’s Guide (Stata Corp, 2021).

interpretation appears more likely, implying that Law No. 12.529/2011 has strengthened enforcement,

leading to fewer violations over time. This deterrence eect is further corroborated by Miller (2009),

who demonstrates that eective anti-cartel measures can reduce the incidence of detected and

condemned cases in the long run.

Table 2 below presents descriptive statistics for the non-dummy variables.

Table 2 - Descriptive statistics of ordinary (non-dummy) variables

Variables Mean Standard deviation Min. Max.

Corporation

- level

Fines

(in Reais)

14.327.437,48 122.063.099,03 5.866,28 2.745.929.130,14

Subsections 3,94 1,75 2 10

Case-level

Duration

(in years)

5,24 5,69 1 33

Corporations 5,10 5,82 1 46

Source: own elaboration (2021).

For the corporation-level variables (597 condemnations), the table above indicates that, on

average, corporations’ activities were classiﬁed under approximately four subsections of Law No.

12.529/11, with a relatively modest standard deviation. The ﬁnes, however, exhibit a high standard

deviation, reﬂecting the considerable variability across both corporations and cases. These values

should be interpreted with caution, as ﬁnes were determined at dierent points in time. Despite

adjustments to December 2020, unobserved factors not captured by simple descriptive statistics

may inﬂuence these values, though they are addressed in the main model. Regarding the case-level

variables (146 antitrust cases), the average duration of cases was ﬁve years, and each case involved,

on average, ﬁve corporations.

For the corporation-level dummy variable “Company,” nearly 75% represent traditional

companies (value one), while the remaining 25% are unions, associations, and other legal entities

(value zero). For the case-level dummy variables, about 20% of the cases involved at least one Cease

and Desist Agreement (coded as one in “TCC”), while nearly 12% of cases involved at least one leniency

agreement (coded as one in “Leniency”). Concerning the aected market, approximately 8% of the

cases were international (coded as one in “International”), around 20% impacted the national market

(coded as one in “National”), and about 31% aected the state level (coded as one in “State”), with

the remainder of cases aecting the municipal level. Approximately 19% of the cases occurred in the

health sector (28 cases and 157 condemned corporations), while 11% occurred in the fuel sector (16

cases and 145 corporations). Lastly, with respect to the year of condemnation, as shown in Figure 1 in

absolute terms, approximately 16% of corporations were condemned in 2013, 24% in 2014, 22% in 2015,

9% in 2016, 6% in 2017, 8% in 2018, 10% in 2019, and 5% in 2020.

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

Estimates are presented in Table 3 below.

Table 3 - OLS and random-intercept model estimates

Dependent variable: ln(Fines) OLS Random-intercept model

Coecient % change Coecient % change

Intercept

10.62***

(0,30)

11,39***

(0,65)

Subsections

0.25***

(0.05)

28,39%

0,27**

(0,11)

31,31%

Duration

0.06***

(0.01)

6,71%

0,06**

(0,02)

6,67%

Corporations

-0,04***

(0,01)

-3,53%

-0,05**

(0,02)

-4,94%

Company

0,80***

(0,20)

123,10%

0,76***

(0,24)

113,75%

Leniency

0,32

(0,47)

37,92%

0,31

(0,69)

36,75%

TCC

0,47*

(0,25)

60,43%

0,53

(0,44)

69,48%

International

2,14***

(0,64)

746,08%

1,73**

(0,88)

462,97%

National

2,16***

(0,25)

770,31%

1,55***

(0,42)

372,42%

State

0,09

(0,17)

9,64%

0,13

(0,25)

13,98%

Health

0,41

(0,20)

50,01%

0,13

(0,26)

13,36%

Fuel

2,53***

(0,21)

1158,17%

2,16***

(0,47)

768,24%

Year2020

0,63

(0,40)

86,92%

0,17

(0,60)

19,10%

Year2019

0,36

(0,38)

43,28%

-0,04

(0,79)

3,69%

Year2018

0,55

(0,38)

72,84%

-0,11

(0,75)

-10,30%

Year2017

-0,43*

(0,35)

-34,98%

-0,84

(0,55)

-56,68%

Year2016

-0,62*

(0,34)

-46,46%

-0,55

(0,66)

-42,09%

Year2015

0,21

(0,23)

22,85%

-0,34

(0,50)

-28,95%

Year2014

0,27

(0,23)

30,69%

-0,15

(0,47)

-13,72%

Random intercept variance - -

1,29***

(0,24)

R 51% - - -

F-test 34.75*** - - -

Likelihood ratio test - - 196,37*** -

Intraclass correlation - - 0,54 -

Source: own elaboration (2021).

Note: Standard errors were clustered at the case level to account for potential correlation of

residuals among ﬁrms judged within the same antitrust case. Number of observations - 597

corporations condemned in 146 antitrust cases.

*** Statistically signiﬁcant at 1%; ** Statistically signiﬁcant at 5%; * Statistically signiﬁcant at 10%.

First, for the OLS model, nearly 51% of the variance in is explained by the regressors, and

the F-test rejects the null hypothesis that all coecients are jointly equal to zero. For the random-

intercept model, the Likelihood-Ratio test rejects the null hypothesis that both models provide

similar ﬁts, suggesting that a random coecient model is more appropriate. The random-intercept

variance is statistically signiﬁcant at the 1% level, further supporting the use of a random-intercept

model. Additionally, the intra-class correlation indicates that 54% of the variance in ln(Fines) can be

explained by group-level factors (i.e., antitrust cases in this study).

Turning to the random-intercept model estimates, the coecient for the subsection variable is

statistically signiﬁcant at the 5% level and has the expected sign, indicating that being classiﬁed under

an additional subsection of Law No. 12.529/11 is associated with a 31,31% increase in the average level

of ﬁnes. This ﬁnding is particularly relevant for legal teams representing defendants, as it suggests

that reducing the number of subsections under which a corporation is framed could potentially lower

the ﬁnes imposed by Cade. Regarding the duration variable, the coecient is signiﬁcant at the 5%

level with a positive sign, suggesting that an additional year of infringement is associated with a 6,67%

increase in ﬁnes. This relationship aligns with the notion that longer durations are associated with

greater harm, which in turn inﬂuences the severity of penalties based on aggravating factors. This

positive association is valuable in reinforcing the deterrence eect, as it underscores the ﬁnancial

risks of prolonged violations.

Regarding “Corporations”, the negative and statistically signiﬁcant coecient at 5% indicates

that cases involving a larger number of ﬁrms are associated with lower ﬁnes per corporation, suggesting

a possible dilution eect in penalty setting across participants within the same infringement. For

the “Company” dummy variable, being classiﬁed as a traditional company is associated with a

113.75% increase in ﬁnes, statistically signiﬁcant at the 1% level. As outlined in Section 2, traditional

companies are legally required to provide gross sales data to inform penalty calculations, making it

easier for Cade to set ﬁnes, particularly when considering aggravating factors. Furthermore, traditional

companies tend to be more proﬁtable than unions, associations, and other legal entities, which may

explain the higher ﬁnes they receive. These companies are also more likely to engage in high-proﬁle

infringements, such as horizontal collusion, while other entities are typically involved in inﬂuencing

uniform conduct or abusing dominant positions. The positive and statistically signiﬁcant coecient

of “Fuel” indicates that infringements in the fuel sector are associated with substantially higher ﬁnes,

suggesting that Cade treats violations in this sector as particularly harmful, possibly due to their

broad consumer impact and recurrent cartelization patterns.

Regarding market scope, the coecient for “International” and “National” dummy variables

are positive and statistically signiﬁcant at 5% and 1%, respectively. Antitrust violations with an

international (national) scope is associated with a 462.97% (372,42%) increase in ﬁnes, which is

consistent with expectations: the broader the market aected, the greater the damage caused, and

thus, the higher the aggravating factors considered by Cade.

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

The year dummies are included to control for various factors, such as changes in the

political and economic environment, as well as administrative and legal shits within Cade. Given the

complexity of these factors, it is dicult to derive direct interpretations from signiﬁcant coecients.

As the primary focus of this paper is on the determinants of antitrust ﬁnes, detailed analysis of these

year-speciﬁc factors is omitted here.

Turning to the non-signiﬁcant coecients, the “State” market scope variables are not

statistically signiﬁcant, although they exhibit the expected signs. Although the coecients for

“Leniency” and “TCC” (Cease and Desist Agreements) have the expected signs, they are also statistically

insigniﬁcant. These policies, which encourage cooperation with investigations in exchange for

reduced penalties, were anticipated to correlate positively with higher ﬁnes for other members of the

violation, but in our analysis, neither the number of corporations nor the presence of leniency or TCC

agreements were signiﬁcant determinants of the ﬁnes imposed by Cade. The coecient for “Health”

is positive but not statistically signiﬁcant in the random-intercept model.

The ﬁndings for Brazil reveal both similarities and dierences relative to the empirical

literature on antitrust ﬁnes in the United States and the European Union. Consistent with the evidence

for both jurisdictions (Connor and Miller, 2009; Connor and Miller, 2013), ﬁnes imposed by Cade are

positively associated with indicators of infringement seriousness, including the scope of the aected

market and the legal classiﬁcation of the conduct, thereby reinforcing the role of monetary sanctions

as a central instrument of deterrence. With respect to infringement duration, the Brazilian results are

aligned with those documented for the European Union, where duration is statistically signiﬁcant,

indicating that longer-lasting infringements are understood as generating greater cumulative harm

and thus warrant higher penalties. By contrast, in the United States, cartel duration is not statistically

signiﬁcant at the 10% level, suggesting that the prevalence of plea bargaining may attenuate the

direct relationship between observable characteristics of the infringement and ﬁnal ﬁne levels. From

a policy perspective, these ﬁndings support the continued treatment of duration as an aggravating

factor in ﬁne-setting, as it enhances deterrence while contributing to the coherence and consistency

of enforcement decisions.

5 CONCLUSIONS

Given the factors that Cade considers when imposing monetary penalties on oenders,

understanding which ones inﬂuence the level of ﬁnes has become a crucial task. This paper sought to

analyze the determinants of corporate antitrust ﬁnes in Brazil using a random-intercept model.

Our ﬁndings suggest that the number of subsections under Law No. 12.529/11, the duration

of the infringement, the number of corporations involved, and the classiﬁcation as a traditional

company (compared to other types of corporations) are positively correlated with the level of ﬁnes.

Additionally, the scope of the market plays a signiﬁcant role in explaining the penalties, particularly at

the international and national levels. We also ﬁnd that cases involving the fuel sector are associated

with higher ﬁnes, indicating sector-speciﬁc enforcement patterns.

Our ﬁndings also contribute to the literature on optimal ﬁnes by highlighting how Brazilian

enforcement practice aligns with core theoretical principles. The positive association between

infringement duration and ﬁne severity reinforces the deterrence rationale discussed by Becker (1968),

while the higher penalties applied to traditional ﬁrms reﬂect Wils’ (2006) concern with proportionality

and capacity to pay. By clarifying these mechanisms, the paper shows how Cade’s sanctioning patterns

can strengthen deterrence and enhance enforcement eciency, thus connecting the empirical results

to broader debates on the optimal design of antitrust ﬁnes.

These results oer valuable insights for Cade, legal professionals, researchers, and other

stakeholders. Moreover, they highlight the importance of this information for corporations, as it

enhances the deterrence eect by making the consequences of misconduct clearer. In other words,

our ﬁndings contribute to the strengthening of the deterrent eect of antitrust laws, which is a key

goal of law enforcement.

For future research, we recommend exploring additional potential determinants, including

institutional factors. Investigating the underlying causes behind the year dummies in our model

would also be an interesting avenue for further study.

REFERENCES

ALLAIN, Marie-Laure; BOYER, Marcel; KOTCHONI, Rachidi; PONSSARD, Jean-Pierre. Are cartel

ﬁnes optimal? Theory and evidence from the European Union. International Review of Law and

Economics, v. 42, p. 38-47, 2015. DOI: https://doi.org/10.1016/j.irle.2014.12.004. Available at: https://

www.sciencedirect.com/science/article/abs/pii/S0144818814000878. Accessed on: Feb. 20, 2026.

BECKER, Gary. Crime and punishment: an economic approach. In: BECKER, Gary; LANDES, W. M. (ed.).

The economic dimensions of crime. London: Palgrave Macmillan, 1968. p. 13-68.

BOLOTOVA, Y.; CONNOR, J. M. Cartel sanctions: an empirical analysis, SSRN, 2008. Available at:

https://ssrn.com/abstract=1116421. Accessed on: Feb. 20, 2026.

BRASIL. Lei nº 12.529, de 30 de novembro de 2011. Estrutura o Sistema Brasileiro de Defesa da

Concorrência; [...] e dá outras providências. Brasília, DF: Presidência da República, 2011. Available at:

https://www.planalto.gov.br/ccivil_03/_ato2011-2014/2011/lei/l12529.htm. Accessed on: Jul. 12, 2024.

ADMINISTRATIVE COUNCIL FOR ECONOMIC DEFENSE (Cade). Guidelines: Cade’s antitrust leniency

program. Brasília, DF: Cade, 2016a. Available at: https://cdn.cade.gov.br/Portal/centrais-de-

conteudo/publicacoes/guias-do-cade/guidelines-cades-antitrust-leniency-program-2020.pdf.

Accessed on: Jul. 12, 2024.

ADMINISTRATIVE COUNCIL FOR ECONOMIC DEFENSE (Cade). Guidelines: cease and desist agreement

for cartel cases (“TCC”). Brasília, DF: 2016b. Available at: https://cdn.cade.gov.br/Portal/centrais-de-

conteudo/publicacoes/guias-do-cade/guidelines_tcc-1.pdf. Accessed on: Jul. 12, 2024.

CONSELHO ADMINISTRATIVO DE DEFESA ECONÔMICA (Cade). Guia dosimetria de multas de cartel.

Brasília, DF: Cade, 2023. Disponível em: https://cdn.cade.gov.br/Portal/centrais-de-conteudo/

publicacoes/guias-do-cade/Guia-dosimetria-de-multas-de-cartel.pdf. Acesso em: Nov. 11, 2025.

CONNOR, John; MILLER, Douglas. Determinants of EC antitrust ﬁnes for members of global cartels.

SSRN, 2013. Available at: https://ssrn.com/abstract=2229358. Accessed on: Feb. 20, 2026.

CONNOR, John; MILLER, Douglas. Determinants of US antitrust ﬁnes of corporate participants

of global cartels. SSRN, 2009. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_

PINHA, Lucas Campio; BRAGA, Marcelo José; ESTEVES, Luiz Alberto. An empirical analysis of

antitrust ﬁnes in Brazil (2012-2020). Revista de Defesa da Concorrência, Brasília, v. 14, n. 1, p. 71-

89, 2026.

https://doi.org/10.52896/rdc.v14i1.1977

id=2229300. Accessed on 19 Feb, 2026.

EUROPEAN COMMISSION. Fines for breaking EU competition law. European Commission, 2011.

Available at: https://ec.europa.eu/competition/cartels/overview/factsheet_ﬁnes_en.pdf. Accessed

on: Jul. 07, 2021.

GALLO, Joseph; CRAYCRAFT, Jos; DUTTA, Shantanu. Incarceration and ﬁnes: an empirical study of

antitrust sanctions. Review of Industrial Organization, v. 3, n. 2, p. 38-66, 1986. DOI: https://doi.

org/10.1007/BF02230837. Available at: https://link.springer.com/article/10.1007/BF02230837. Accessed

on: May 14, 2026.

GOLDSTEIN, H. Multilevel statistical models. Chichester: John Wiley & Sons, 2011.

HARRINGTON, Joseph.; CHANG, Myong-Hun. Modeling the birth and death of cartels with an

application to evaluating competition policy. Journal of the European Economic Association, v. 7, n.

6, p. 1400-1435, 2009. DOI: https://doi.org/10.1162/JEEA.2009.7.6.1400. Available at: https://x.gd/deXH3.

Accessed on: Nov. 11, 2025.

HARRINGTON, Joseph. The theory of collusion and competition policy. Cambridge: MIT Press, 2017.

JING, Ran; GONG, Jiong; YI, Fang. Antitrust ﬁnes: experiences from China. Review of Industrial

Organization, v. 57, n. 1, p. 167-187, 2020. DOI: https://doi.org/10.1007/s11151-019-09743-0. Available at:

https://link.springer.com/article/10.1007/s11151-019-09743-0. Accessed on: May 14, 2026

MILLER, Nathan. Strategic leniency and cartel enforcement. American Economic Review, v. 99, n. 3,

p. 750-768, 2009. DOI: 10.1257/aer.99.3.750. Available at: https://www.aeaweb.org/articles?id=10.1257/

aer.99.3.750. Accessed on: May 14, 2026.

MUNHOZ, João Ricardo Oliveira; OLIVEIRA, Renan Cruvinel. A participação de agentes públicos em

cartéis investigados pelo Cade no âmbito da Operação Lava-Jato. Revista do IBRAC, n. 1, p. 318-339,

2020. Available at: https://revista.ibrac.org.br/revista/article/view/98/91. Accessed on: Nov. 3, 2025.

POSNER, Richard. A statistical study of antitrust enforcement. The Journal of Law and Economics,

v. 13, n. 2, p. 365-419, 1970. DOI: https://doi.org/10.1086/466698. Available at: https://www.journals.

uchicago.edu/doi/abs/10.1086/466698. Accessed on: May 14, 2026.

SNIJDERS, Tom; BOSKER, Roel. Multilevel analysis: an introduction to basic and advanced multilevel

modeling. London: Sage, 2011.

SPAGNOLO, Giancarlo. Leniency and whistleblowers in antitrust. In: BUCCIROSSI, P. (ed.). Handbook

of antitrust economics. Cambridge: MIT Press, 2008.

STATA CORP. Stata user’s guide release 17. College Station, Texas: StataCorp LLC, 2021. Available at:

https://www.stata.com/manuals17/u.pdf. Accessed on: Jul. 12, 2021.

STEENBERGEN, Marco.; JONES, Bradford. Modeling multilevel data structures. American Journal

of Political Science, v. 46, n. 1, p. 218-237, 2002. DOI: https://doi.org/10.2307/3088424. Available at:

https://www.jstor.org/stable/pdf/3088424.pdf. Accessed on: May 14, 2026.

TODOROV, Francisco Ribeiro; TORRES FILHO, Marcelo Maciel. History of competition policy in Brazil:

1930–2010. The Antitrust Bulletin, v. 57, n. 2, p. 207-257, 2012. Available at: https://papers.ssrn.com/

sol3/papers.cfm?abstract_id=2327925. Accessed on: Nov. 3, 2025.

UNITED STATES SENTENCING COMMISSION. Antitrust oenses. Washington, DC: United States

Sentencing Commission, 2020. Disponível em: https://www.ussc.gov/guidelines/primers/antitrust.

Acesso em: Jul. 01, 2021.

WILS, Wouter. Optimal antitrust ﬁnes: theory and practice. World Competition, v. 29, n. 2, p. 183-

208, 2006. DOI: https://doi.org/10.54648/woco2006014. Available at: https://kluwerlawonline.com/

journalarticle/World+Competition/29.2/WOCO2006014. Accessed on: Nov. 3, 2025.