ABSTRACT Title of Dissertation: TYPOLOGIES OF FORCED LABOR EXPLOITATION IN BRAZIL Shelby Hickman, Doctor of Philosophy, 2021 Dissertation directed by: Distinguished University Professor, Dr. Sally Simpson, Department of Criminology and Criminal Justice Forced labor exploitation is the most common element of present-day institutional slavery. Despite the pervasive nature of this crime, little is known about the ways that perpetrators recruit workers and keep them in exploitive situations. Further, forced labor exploitation cases are rarely brought forward for prosecution and even more rarely receive a conviction. In this dissertation I examine the characteristics of forced labor exploitation in Brazil. Additionally, using a Focal Concerns framework, I examine the factors that influence the decision making of key investigative and court practitioners involved in processing forced labor exploitation cases. To do this, I analyzed administrative data from all (n=1,764) forced labor exploitation cases processed in the criminal and civil court systems in Brazil between 2008 and 2020. I also conducted 28 interviews with labor inspectors, federal police, and judges and prosecutors from the civil and criminal court systems. Using latent class analysis, I identified three typologies of forced labor exploitation: degrading conditions and debt servitude, degrading conditions, and degrading conditions and weapons and surveillance. I then examined the factors associated with different typologies of forced labor exploitation as well as the association between type of forced labor exploitation and sentencing outcomes. Respondents described several factors that increase uncertainty in forced labor exploitation cases, including: subjective interpretations of the criminal code, lack of formal training in handling forced labor exploitation cases, and uncertainty about who should be held accountable in larger organizational schemes. Interview participants further reported that cases that include physical violence, weapons, and ostensive surveillance are more likely to receive a conviction. In my analysis of the administrative data, I find that cases in the degrading conditions and weapons class are no more likely to receive a criminal conviction; however, cases in the degrading conditions and weapons class that received a conviction received more severe punishments. I discuss ways to improve investigation and prosecution of forced labor exploitation cases based on the study findings as well as potential alternatives to criminal court processing that may be more effective in reducing the burden of forced labor exploitation. TYPOLOGIES OF FORCED LABOR EXPLOITATION IN BRAZIL by Shelby Nichole Hickman Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2021 Advisory Committee: Professor Sally Simpson, Chair Professor James Lynch Professor Wendy Stickle Professor Sarah Tahamont Professor Elaine Anderson ? Copyright by Shelby Nichole Hickman 2021 ii Acknowledgements I am incredibly thankful for the tremendous support I received while preparing this dissertation. First and foremost, to my advisor, Dr. Sally Simpson, who guided me throughout my graduate school career. She has been generous with her time and support since my first semester of graduate school. Thank you. My committee members Dr. Wendy Stickle, Dr. Sarah Tahamont, Dr. James Lynch, and Dr. Elaine Anderson, whose feedback was critical for shaping this dissertation. Dr. Wendy Stickle and Christine White, who invited me to co-author a textbook on human trafficking with them which ultimately led me to make a connection with Dr. Carlos Haddad and make this dissertation possible. Dr. Carlos Haddad, who invited me to collaborate with him and use his clinic?s data for this dissertation. He also helped me to recruit interview participants and provided critical insight into my findings which helped to shape this dissertation. Thank you to my friends and colleagues who helped me navigate graduate school and become a stronger researcher. Finally, I could not have completed this dissertation without the support of my family--my parents Jennifer, Bruce, and Tracie, and Alisha who has always been like a parent, and my siblings Casey, Grace, Danielle, and Tanner. And thank you Grandpa Liner for immediately telling me to get a PhD when I graduated and told you I never wanted to go to school again. Thank you to Elenice, Mara, Sebastiao, and Deborah for all of your support especially while we were in Brazil. Finally, my husband Daniel, and my kids Riley and Sebastian. Thank you for understanding the early mornings, limited free time, stress, and for moving to another country and back with two cats and a suitcase full of Legos. I could not have done this without you. i Table of Contents Acknowledgements ........................................................................................................ i Table of Contents .......................................................................................................... ii List of Tables ............................................................................................................... iii List of Figures .............................................................................................................. iv Executive Summary ...................................................................................................... v Chapter 1: Introduction ................................................................................................. 1 Chapter 2: Literature Review ........................................................................................ 7 2.1 Definition of forced labor exploitation ............................................................... 8 2.1 Prevalence of forced labor exploitation ............................................................ 12 2.2 Economic Sectors.............................................................................................. 13 2.3 Vulnerabilities for forced labor exploitation ..................................................... 17 2.4 Causes of forced labor exploitation .................................................................. 23 2.5 Gaps in understanding forced labor exploitation .............................................. 26 Chapter 3: Forced labor exploitation and the Brazilian context ................................. 31 3.1 Introduction of forced labor exploitation legislation in Brazil ......................... 31 3.2 Impact of public policies targeting forced labor exploitation ........................... 39 3.3 Case processing of forced labor exploitation in the criminal and labor court systems .................................................................................................................... 41 3.4 Theoretical framework to understand court response to forced labor exploitation in Brazil............................................................................................... 50 Chapter 4: Research Questions, Data, and Methods ................................................... 60 4.1 Administrative data ........................................................................................... 61 4.2 Qualitative data ................................................................................................. 64 4.3 Analytic methods .............................................................................................. 66 Chapter 5: Results ....................................................................................................... 77 5.1 Qualitative data ................................................................................................. 77 5.2 Administrative data ........................................................................................... 78 5.3 Descriptive information .................................................................................... 79 5.4 Variables used in analyses for objectives one through three .......................... 135 5.5 Objective One: Describe characteristics of forced labor exploitation cases forwarded for processing in criminal and civil courts .......................................... 138 5.6 Objective Two: Examine factors that predict membership in each Forced Labor Exploitation Case Profile ...................................................................................... 142 5.7 Objective Three: Examine the sentencing outcomes for different types of forced labor exploitation cases ......................................................................................... 144 Chapter 6: Discussion .............................................................................................. 166 6.1 Overview of findings ...................................................................................... 166 6.2 Limitations ...................................................................................................... 172 6.3 Implications and recommendations ................................................................ 175 Appendices ................................................................................................................ 194 References ................................................................................................................. 243 ii List of Tables Table 1. Labor inspections conducted by the Public Ministry of the Economy between 1995 and 2013 .................................................................................................................................. 39 Table 2. Research questions, data sources, and method. ........................................................ 60 Table 3. Summary of recruited participants ............................................................................ 64 Table 4. Respondent sex and role ........................................................................................... 77 Table 5. Respondent tenure ..................................................................................................... 78 Table 6. Respondent experience with urban and rural cases .................................................. 80 Table 7. Respondents' current geographic location ................................................................ 84 Table 8. Number of workers reached, rescued, and receiving unemployment ....................... 90 Table 9. Other crimes committed alongside forced labor exploitation, interview data .......... 91 Table 10. Other crimes identified alongside forced labor exploitation ................................... 96 Table 11. Characteristics of forced labor exploitation identified in inspection reports ........ 104 Table 12. Tools to combat forced labor exploitation mentioned by interview participants .. 106 Table 13. Amount of individual moral damages requested by Regional Labor Court ......... 117 Table 14. Amount of collective moral damages requested by Regional Labor Court .......... 119 Table 15. Number of public civil actions in which declaration of forced labor exploitation was requested and granted .................................................................................................... 121 Table 16. Characteristics of forced labor exploitation recognized in case outcome and in inspection report for public civil actions ending in an agreement or sentence ..................... 124 Table 17. Criminal action outcomes ..................................................................................... 128 Table 18. Characteristics of forced labor exploitation recognized at criminal sentencing ... 130 Table 19. Other crimes indicted and sentenced .................................................................... 131 Table 20. Summary of variables used in analyses ................................................................ 137 Table 21. Goodness of fit indicators for latent class analysis (N=1,075) ............................. 138 Table 22. Latent class marginal probabilities for characteristics of forced labor exploitation, 3-class solution ..................................................................................................................... 140 Table 23. Tests of categorical latent variable multinomial logistic regression using the 3-step procedure (N=944) ................................................................................................................ 143 Table 24. Cross tab of latent classes and inclusion in dirty list, column frequencies (N=672) .............................................................................................................................................. 157 Table 25. Cross tab of latent classes and public civil action outcomes, column frequencies 158 Table 26. Significant differences between class-specific means for individual moral damages and collective moral damages ............................................................................................... 159 Table 27. Significant differences between class-specific means for criminal case outcomes: Convicted, acquitted, sentence length, case length ............................................................... 163 iii List of Figures Figure 1 Map of Brazil .............................................................................................................. 8 Figure 2. Human trafficking and forced labor exploitation ...................................................... 9 Figure 3 Process for Inclusion in the Dirty List ...................................................................... 38 Figure 4. Timeline of forced labor exploitation policies in Brazil .......................................... 40 Figure 5. Page of a report from a labor inspector. .................................................................. 45 Figure 6. Summary of three-step approach ............................................................................. 71 Figure 7. Economic sector of forced labor exploitation cases, N=1,147 ................................ 83 Figure 8. Forced labor exploitation identified in rural sectors has declined over time ........... 84 Figure 9. Number of cases identified in each state and cumulative percentage of cases, N=1,638 .................................................................................................................................. 87 Figure 10. Proportion of force labor exploitation cases identified in each region, N=1,638 .. 88 Figure 11. Distribution of cases included in Dirty List by state (N=651) ............................. 115 Figure 12. Number of Cases by Regional Labor Court (N=432) .......................................... 116 Figure 13. Sentencing outcomes of forced labor exploitation cases processed in civil court system (N=432) ..................................................................................................................... 123 Figure 14. Stages of public civil action processing .............................................................. 125 Figure 15. Mean number of days case processing time for public civil actions, by case outcome ................................................................................................................................. 126 Figure 16. Initiation (N=431) and final adjudication (N=295) of public civil actions, by year .............................................................................................................................................. 126 Figure 17. Number and cumulative percent of criminal actions by federal court district ..... 127 Figure 18. Criminal action processing flow chart ................................................................. 133 Figure 19. Stages of criminal case processing ...................................................................... 134 Figure 20. Mean number of days case processing time for criminal actions, by case outcome .............................................................................................................................................. 134 Figure 21. Latent classes of forced labor exploitation .......................................................... 139 iv Executive Summary Introduction Forced labor exploitation is the most common element of present-day institutional slavery. Despite the pervasive nature of this crime, little is known about the ways that perpetrators recruit workers and keep them in exploitive situations. Further, forced labor exploitation cases are rarely brought forward for prosecution and even more rarely receive a conviction. In this dissertation I examine the characteristics of forced labor exploitation in Brazil. The main objectives are as follows: ? Describe characteristics of forced labor exploitation cases forwarded for processing in criminal and civil courts in Brazil. ? Analyze differences in typologies of forced labor exploitation by key characteristics. ? Examine the court outcomes for different types of forced labor exploitation cases. Research Design I analyzed administrative data from all (n=1,764) forced labor exploitation cases processed in the criminal and civil court systems in Brazil between 2008 and 2020. I also conducted 28 interviews with labor inspectors, federal police, and judges and prosecutors from the civil and criminal court systems. Using latent class analysis, I identified three typologies of forced labor exploitation: 1) degrading conditions and debt servitude, 2) degrading conditions, and 3) degrading conditions and weapons and surveillance. I then examined the factors associated with different typologies of forced labor exploitation as well as the association between type of forced labor exploitation and sentencing outcomes. Across all research objectives, I identified factors that influence practitioners? decision making using a focal concerns framework. Overview of Findings What are the characteristics of known forced labor exploitation cases in Brazil? ? I identified three typologies of forced labor exploitation cases in Brazil: degrading conditions and debt servitude, degrading conditions, and degrading conditions and weapons and surveillance. In line with findings from the interviews, the greatest share (83.8 percent) of cases is predicted to fall into the degrading conditions class, the second greatest share fall into the degrading conditions and debt servitude (13.21%) and just about three percent belong to the degrading conditions and weapons and surveillance class (2.98%). What factors influence the type of forced labor exploitation activities used? ? The presence of crimes including danger to life or health of others, relocating workers under false pretenses, and other crimes all were associated with a v lower odds of belonging to the degrading conditions class relative to the degrading conditions and weapons and surveillance class. ? Cases in rural sectors had higher odds of belonging to the degrading conditions class relative to degrading conditions and weapons and surveillance. Findings from the interviews shed some light on this finding: o First, it is possible that other forms of coercion that could not be measured in the data were present, such as social and linguistic isolation. o Second, it is possible that the workers did not feel they had any other prospects for earning money and thus did not attempt to leave the situation. The latter explanation is one that judges, during the interviews, described mixed understandings of: some would not convict in a situation where a worker voluntary submitted themselves to degrading conditions and others said the worker?s opinion did not matter. What are the outcomes of cases prosecuted with forced labor exploitation charges, and how are different typologies of forced labor exploitation cases associated with criminal sentences? Labor Court Outcomes ? For collective moral damages, there is a stepwise pattern as would be expected where the degrading conditions and debt servitude class is associated with higher collective moral damages than the degrading conditions class, and the degrading conditions and weapons and surveillance class is associated with higher collective moral damages than the debt servitude class. ? For individual moral damages, the degrading conditions and debt servitude class has the highest individual moral damages, followed by the degrading conditions and weapons and surveillance class, and then the degrading conditions class. Criminal Court Outcomes ? Interview respondents said that many judges and even prosecutors have a hard time convicting or bringing a case forward in the absence of physical control. o I did not, however, observe an increased likelihood of conviction for the weapons and surveillance class relative to the other classes in the administrative data. ? Degrading conditions and weapons and surveillance cases had significantly longer case lengths relative to degrading conditions cases, however. ? Degrading conditions and weapons and surveillance cases are associated with a longer sentence length than cases in the other two classes which suggests that when the cases are convicted, they are more severely punished. ? Degrading conditions and debt servitude cases are more likely to receive a conviction relative to degrading conditions cases. o Interview respondents said that debt servitude is perhaps the easiest element of forced labor exploitation to prove because employers often keep physical workbooks tracking the debts. The availability of physical evidence that is not reliant on victim testimony is likely a contributing factor to the elevated rates of conviction in debt servitude cases. vi What factors influence practitioner decision making in forced labor exploitation cases (informed by Focal Concerns Theory)? ? In line with other research on decision making among court actors and investigators, respondents in this study were focused on reducing uncertainty in their decision making. Respondents identified several challenges to reducing uncertainty in decision making during forced labor exploitation cases, including: o Subjectivity of the law o Lack of training o Barriers to obtaining victim testimony o Quality of investigation ? Federal police and prosecutors based their decisions about forced labor exploitation cases, in part, on their calculations about the focal concerns of practitioners who would make subsequent decisions about the case. Sometimes these calculations were inaccurate. o For example, prosecutors think that judges are more likely to convict in the presence of physical violence, restriction of freedom, or ostensive surveillance, but the analysis of the administrative data finds that cases with those elements are actually no more likely to receive a conviction. when multiple people are involved in making a decision during a case, such as police, then prosecutors, then judges, focal concerns theory should take into account an individual?s assessment of the focal concerns of the practitioner down the line. ? Respondents disagreed about blameworthiness of offenders. For example, respondents had differing opinion on whether a local farm manager, or the recruiter (gato) should be held accountable for forced labor exploitation if they were only a middleman in a larger organizational scheme. o Participants simultaneously condemned inhumane working conditions and expressed empathy for potential perpetrators of this crime whom they referred to not as criminals but as farmers or bosses. All respondent types overwhelmingly agreed that the dirty list was an effective and appropriate punishment for forced labor exploitation, but there was more disagreement around criminal convictions. This suggests that respondents considered the business or the company as a whole to be worthy of sanctions but they did not necessarily view the employees of that business, operating for the company?s benefit, to be guilty. Key takeaways for practitioners 1. All practitioners, including inspectors, police, prosecutors, and judges in the labor and criminal systems, would benefit from regular training around forced labor exploitation. vii 2. Different practitioner types should debrief after a forced labor exploitation is closed, and at regular intervals each year to discuss the case outcome and contributing factors to the case outcome. 3. Labor inspectors and federal police should work with community partners to develop a greater understanding of urban forced labor exploitation and consider whether it is worthwhile to devote personnel to investigating urban forced labor. 4. Develop an infrastructure to maintain contact with victims of forced labor exploitation after they are rescued from the exploitive situation. This may need to include incentives for victims to maintain contact with courts. 5. Forced labor exploitation cases have several elements of white-collar crime, and investigations into the white-collar crimes can help in identifying the supply chain responsible for worker exploitation. Key takeaways for policymakers 1. Fund and implement more state-level labor inspection teams that follow the model in place in Sao Paolo that allows inspectors to investigate cases after the two-week initial inspection. This will allow inspectors to investigate the full supply chain of forced labor exploitation cases. 2. Work with community partners and survivors of forced labor exploitation to identify and implement alternatives to criminal justice processing, when appropriate. 3. Policy makers outside of Brazil should consider the feasibility of implementing an approach similar to the dirty list. 4. More resources should be dedicated to investigating forced labor exploitation, and the number of labor inspectors should remain constant or increase each year. 5. Strengthening labor protections for domestic workers is essential to combat forced labor exploitation. Key takeaways for researchers 1. In studies of decision making that use a focal concerns approach, there may be a discrepancy between an individual?s focal concerns and what others think are the focal concerns of that individual. 2. Explore to what extent typologies of forced labor exploitation in other countries parallel those identified in Brazil. 3. Evaluate alternatives to criminal justice system processing for forced labor exploitation cases. viii Chapter 1: Introduction Forced labor exploitation is a pervasive crime; it has been identified in every nation and in nearly all economic sectors from domestic work to charcoal production (International Labor Office, 2017b). The International Labor Organization and United Nations describe forced labor exploitation as the most common element of modern- day slavery; it is any work or service that someone is forced to do against their will (International Labor Office, 2012; United Nations Office on Drugs and Crime, 2018). In this dissertation, I describe the different types of forced labor exploitation in Brazil, analyze the differences in typologies of forced labor exploitation by key characteristics, and examine how sentencing outcomes differ for different types of forced labor exploitation cases. Over 369,000 people are estimated to be in a situation of forced labor exploitation at any given time in Brazil. Article 149 of the Brazilian penal code categorizes forced labor exploitation as submitting an individual to forced labor, an exhausting workday, or degrading conditions through any means, including debt servitude or restricting their ability to move freely including through ostensive surveillance or withholding personal documents. Since legally defining forced labor exploitation in 1995, Brazil has made significant progress in combating this crime through the implementation of public policies as well as revising the definition of forced labor exploitation in 2003. The Central Division of Human Rights within the federal police system in Brazil is charged with investigating forced labor exploitation cases and other human rights violations. Data from the federal police show that there have been substantial 1 increases in arrests for forced labor exploitation since the revision of article 149 in 2003. In 2003, the federal police investigated 52 cases of slave labor and this number more than doubled in 2004 when 130 cases were investigated and continued to increase with 200 cases investigated in 2007 (United Nations Office on Drugs and Crime, 2009). Despite this progress, most cases of forced labor exploitation in Brazil, as in the rest of the world, remain undetected. Further, identified cases are unlikely to be prosecuted and even less likely to receive a criminal sentence (Haddad & Miraglia, 2018a; Kappelhoff, 2008). Between 2003-2007, only 11 cases of slave labor were prosecuted and none of these cases were convicted (United Nations Office on Drugs and Crime, 2009). This low number of prosecutions is more jarring when contrasted with the 2,000-5,000 individuals rescued from conditions of forced labor exploitation each year during the same time period (United Nations Office on Drugs and Crime, 2009). The low rates of prosecution for forced labor exploitation mean there is significant impunity for the crime. Researchers, advocates, and policymakers have prioritized development of information that can help to improve our understanding of forced labor exploitation, curb existing human rights violations, and prevent future victimization (e.g., Bouch?, 2017; Brewer, 2009; Inglis, 2001; Issa, 2017; Shively, Smith, Jalbert, & Drucker, 2017). As a result, there have been substantial strides in estimating the prevalence of forced labor exploitation, describing the vulnerabilities of victims, and implementing policies that aim to reduce the burden of forced labor exploitation. There remain several gaps, however, in understanding this crime, in Brazil and globally. 2 First, there is limited information on the characteristics of forced labor exploitation in Brazil, and worldwide (Costa, 2009; DW, 2014; International Labor Office, 2017a; Owens, 2015). The bulk of information on the characteristics of forced labor exploitation comes from small, purposive samples, anecdotes, or stories documented by journalists. These small samples do not allow for rigorous quantitative analysis, and their findings do not elicit generalizable recommendations for policy and practice (Owens, 2015). In this study, I use a large administrative dataset that contains detailed information about each case to generate typologies of forced labor exploitation. I use data from interviews with labor inspectors, federal police, judges, and prosecutors to guide my analysis and to provide additional context to quantitative findings. There is also limited information on how different characteristics of forced labor exploitation are associated with sentencing outcomes (Kappelhoff, 2008; OSCE et al., 2008). This gap in understanding is due in part to overall low levels of prosecution for forced labor exploitation in Brazil and worldwide. The United Nations Office on Drugs and Crime (UNODC) finds that of the 155 countries included in their biannual trafficking in persons report, 63 (41 percent) reported no convictions for forced labor exploitation in 2018 (United Nations Office on Drugs and Crime, 2018). While this demonstrates that there are overall low levels of prosecution, it does not explain why. In one study that examined 140 human trafficking cases in the U.S., the authors found that most cases met legal requirements for prosecution, but in general, suspects were never charged. The authors cited several barriers to prosecution and conviction in these cases including subjective legal 3 definitions and interpretation of case law, lack of specialized training for court actors related to human trafficking, and perceived credibility of victims (Farrell et al., 2014). I use a mixed method approach to explore how different typologies of forced labor exploitation are associated with outcomes in the criminal and civil court systems at the national level in Brazil. This is the first study to comprehensively examine forced labor exploitation for the entire country. I also interview labor inspectors, federal police, prosecutors, and judges to learn how cases are prioritized for investigation and prosecution. What constitutes forced labor exploitation is subject to discretion at several stages in the criminal justice system. First, an investigator must determine that a crime meets the requirements; then, the criminal prosecutor determines whether the case is viable for prosecution; and finally, the judge must decide whether the weight of evidence constitutes a criminal activity. There is no research to date, however, on how decisions are made at each of these stages. My study draws from interviews with actors at every stage of the criminal justice system to determine how decisions are made, how different elements of the crime are defined, and what kinds of evidence would need to be present to support claims that those elements are present. I then compare the subjective interpretation of the law across practitioners to identify areas of concordance and disagreement. There are a handful of studies, based in the United States, that examine the decision making of investigators, prosecutors, or judges in forced labor exploitation cases (Farrell et al., 2008, 2014; Kappelhoff, 2008). In this study, I use focal concerns theory to guide these questions so that I can establish which factors are most relevant to facilitating or inhibiting case processing and guilty 4 convictions. This theory has been applied in U.S.-based studies, and I update the framework based on the local context in Brazil (Farrell et al., 2014; Spohn, 2014). Examining the factors that influence decision making is particularly important in Brazil because the forced labor exploitation law leaves a large margin for subjective interpretation. This study is one of only two studies to examine different case outcomes for forced labor exploitation cases in Brazil, and the only study to do so at a national level (Haddad & Miraglia, 2018a). This is also the first study in Brazil, and one of a few globally, that examines how court actors make decisions about whether to bring a forced labor exploitation case forward for prosecution, and whether to find the defendant(s) guilty (Farrell et al., 2014; Go?dziak & Bump, 2008; Kappelhoff, 2008; Laan et al., 2011). Dissertation overview This dissertation has three objectives: 1) describe characteristics of forced labor exploitation cases forwarded for processing in criminal and civil courts in Brazil, 2) analyze differences in typologies of forced labor exploitation by key characteristics, and 3) examine the sentencing outcomes for different types of forced labor exploitation cases. To achieve these objectives I carry out a mixed-methods study that incorporates interviews with key stakeholders from the criminal and labor court systems as well as administrative data for all cases nationwide. Using this data I answer four key research questions: 1) What are the characteristics of known forced labor exploitation cases in Brazil?; 2) What factors influence the type of forced labor exploitation activities used (e.g. geography, economic sector, co-occurring criminal 5 activities, size of operation)?; 3 What are the outcomes of cases prosecuted with forced labor exploitation charges?, and; 4) How are different typologies of forced labor exploitation cases associated with criminal sentences? In chapter two I provide an overview of the existing literature on forced labor exploitation. In chapter three, I describe the landscape of efforts to combat forced labor exploitation in Brazil, including describing the criminal and labor court processing systems. I also introduce my theoretical framework, focal concerns theory. In chapter four I present an overview of my data sources and analytic methods. In chapter five I present findings from the analyses. Finally, in chapter six, I describe the limitations of the project and a discussion of the results. 6 Chapter 2: Literature Review In this chapter, I first define forced labor exploitation and describe the differences between forced labor exploitation and other related crimes such as sex and labor trafficking. I then describe what is known about forced labor exploitation, including the estimated prevalence, economic sectors where it takes place, the vulnerabilities for perpetration and victimization, and root causes. Given the hidden nature of forced labor exploitation, there is still limited research available in many of these areas; thus, in each section I first provide statistics pertaining to the international context and then provide information specific to Brazil when it is available. Finally, I discuss gaps in our understanding of forced labor exploitation and describe how this study fills those gaps. A map of Brazil is available below to reference during discussions about geographic differences in forced labor exploitation. 7 Figure 1 Map of Brazil Source: Wikipedia 2.1 Definition of forced labor exploitation Human rights organizations typically describe forced labor exploitation as the most common element of modern-day slavery (International Labor Office, 2012; United Nations Office on Drugs and Crime, 2018). Forced labor exploitation is any work or service that someone is forced to do against their will; sometimes forced labor exploitation is used interchangeably with the term human trafficking. Figure 1 8 describes the overlap in the definitions of human trafficking and forced labor exploitation. Figure 2. Human trafficking and forced labor exploitation Human trafficking F orced labor Forced labor ? Trafficking into exploitation ? State-imposed forced marriage ? Trafficking forced labor into labor ? Trafficking into exploitation adoption ? Trafficking ? Trafficking for into forced organ removal sex work Forced labor exploitation overlaps conceptually with human trafficking into labor exploitation and human trafficking into forced sex work. Most countries, including the United States for example, define human trafficking as one crime with multiple types, including sex and labor trafficking. The International Labor Office (ILO) similarly defines forced labor exploitation as an umbrella term with subtypes for forced labor and forced sexual labor. There are important reasons for the distinction of sex work from other types of work in this conceptual framework. While there are some commonalities for vulnerability for either type of forced labor exploitation, the victim and perpetrator profiles for forced sexual labor versus other forms of forced labor are distinct and thus efforts to research, investigate, and intervene to combat these crimes must be 9 uniquely tailored (United Nations Office on Drugs and Crime, 2018). I focus on forced labor exploitation for non-sex work which I refer to simply as forced labor exploitation. Under international criminal frameworks, forced labor exploitation must have three elements (International Labor Office, 2005): ? Provision of work or service in any economic sector or industry; ? Use of threat or force to compel work or service; ? And, involuntary nature of the work or service performed. Importantly, voluntarily submitting to work requires free and informed consent to take the job and freedom to leave at any time. A primary challenge to understanding and combatting forced labor exploitation is the subjectivity of the second and third elements, threat or use of force and involuntariness. For example, if a recruiter convinces a worker to accept a job by providing false information about compensation or work conditions and the worker accepts, this is considered involuntary work. If the worker is forced to stay and work because the employer holds a gun to his head, then this is forced labor exploitation. However, if the worker traveled 200 miles from home and arrives at the job site to discover the job is not what the worker described, but no one is stopping him from leaving, it becomes less clear whether this is forced labor exploitation (legally speaking). The definition of forced labor exploitation can also vary significantly by country (Patricia Trindade Maranhao Costa, 2009; United Nations Office on Drugs and Crime, 2018). Article 149 of the Brazilian penal code categorizes forced labor 10 exploitation as submitting an individual to forced labor, an exhausting workday, or degrading conditions through any means, including debt servitude or restricting their ability to move freely including through ostensive surveillance or withholding personal documents. Directly translated, Article 149 defines ?slave labor? rather than forced labor exploitation; however, forced labor exploitation is more aligned with international standards (Haddad, 2017). Only one element of the definition from Article 149 is needed to constitute forced labor exploitation in Brazil, unlike the international definition which requires all three elements listed above. This means that Brazil?s definition does not require the presence of the element of the threat or use of force, making this definition broader than that used by the ILO and other international organizations (Patricia Trindade Maranhao Costa, 2009; Haddad, 2017). In theory, Brazil?s broad definition in Article 149 should make it easier to identify and convict perpetrators of forced labor exploitation; however, research conducted over the past 25 years since this law was implemented finds that this is not the case (Haddad & Miraglia, 2018b). As is the pattern globally, prosecution for forced labor exploitation in Brazil is rare, and convictions are even rarer. In this chapter, I review the known characteristics of forced labor exploitation, including the estimated prevalence, economic sectors where it is found, the vulnerabilities for victimization, and the root causes of this crime, internationally and in Brazil. Then, I describe the gaps in our understanding of forced labor exploitation and explain how this dissertation builds on what is known to fill the gaps. 11 2.1 Prevalence of forced labor exploitation Forced labor exploitation exists in a broad range of industries from domestic servitude to charcoal production (International Labor Office, 2017b). It is difficult to measure this crime because it often goes undetected. Global and country-specific estimates are based on a survey developed by the ILO and the Walk Free Foundation. The survey uses a random sample and is implemented in over 50 countries and 53 languages. Prevalence estimates are then extrapolated based on the survey (International Labor Office, 2017b). There are an estimated 16.4 million victims of forced labor exploitation worldwide, including 14.2 million victims of privately imposed forced labor and 2.2 million victims of government or state-imposed forced labor (International Labor Office, 2017b). The proportion of victims subject to forced labor exploitation has steadily increased each year, from 32 percent in 2007 to 34 percent in 2016, the most recent year of available data (United Nations Office on Drugs and Crime, 2018b). The increase in the proportion of victims of forced labor exploitation is due, in part, to increased knowledge surrounding this crime and better reporting (United Nations Office on Drugs and Crime, 2018b). Over 369,000 people are estimated to be in a situation of forced labor exploitation at any given time in Brazil. Most of these individuals are men, Brazilian citizens, and work in rural industries. Data collected by the ILO in Brazil reports that more than 35,000 individuals have been ?rescued? from situations of forced labor exploitation since 2003 (International Labor Office, 2012). Most of these individuals were located in areas undergoing rapid economic growth, largely in the agricultural industry (International Labor Office, 2012). In 2016, according to the Ministry of 12 Economy in Brazil, 31 percent of workers rescued were in the mining sector, 19 percent in the construction industry, and 15 percent in agriculture (Guedes et al., 2019). Globally, forced labor exploitation is found in three primary industry categories: 1) domestic work; 2) agriculture, forestry, and fishing; and 3) construction, manufacturing, mining, and utilities (International Labor Office, 2017b). Below, I review the prevalence and characteristics of forced labor exploitation in each of these three categories. 2.2 Economic Sectors 2.2.1 Domestic Work Domestic workers provide a range of services related to managing households including cleaning, cooking, and taking care of children. Domestic workers commonly live in the house with their employer. Approximately 3.4 million domestic workers are victims of forced labor exploitation, representing 24 percent of all forced labor victims and 6.5 percent of all domestic workers globally (International Labor Organization, 2014). Victims of forced labor exploitation in the domestic sector generate $7.9 billion in stolen wages or $2,300 per victim, on average per year (International Labor Organization, 2014). Brazil employs nearly seven million domestic workers, more than any other country worldwide (International Labor Organization, 2018). In Brazil, domestic work has historical origins in slavery; today most domestic workers are women and are disproportionately Indigenous or of Afro-Brazilian descent. Until 2015, domestic workers were excluded from national labor protections in Brazil (Walk Free Foundation, 2018). Consequently, domestic laborers worked long hours with low or 13 no pay, often were subjected to physical and sexual abuse, and their movement was monitored and restricted. Forced child labor is also prevalent in the domestic work industry in Brazil, particularly among Paraguayan and Indigenous Brazilian children (Bureau of International Labor Affairs, 2018). In 2015, Brazil ratified ILO convention 189, which imposes a 44-hour maximum work week; legally prohibits domestic work for children; and ensures other labor protections such as paid vacation time, social security, and compensation for unfair dismissal (International Labor Organization, 2018). 2.2.2 Agricultural, forestry, and fishery workers An estimated 1.7 million individuals are victims of forced labor exploitation in the agricultural, forestry, and fishery sectors, accounting for 11 percent of forced labor exploitation victims globally. These victims generate an estimated $9 billion in profits for their employers. Several factors facilitate force labor exploitation in the agricultural sector, one being that there are weak labor protections for agricultural workers in many countries, including Brazil (International Labor Organization, 2014). Historically, forced labor exploitation in Brazil has been focused in isolated, rural areas, where the demand for cheap labor is driven by industries which require intensive labor such as coffee production, forestry, and cattle ranching. A research study in 2007 found that 62 percent of all forced laborers in Brazil work in livestock farming, 19 percent in agricultural production (e.g., soy, corn, rice, beans, coffee), and 12 percent in charcoal production (Costa, 2009). In a typical scenario of forced labor exploitation in the agricultural industry in Brazil, employment brokers bring workers, often young men, to rural areas for work. Once workers arrive, they realize 14 the work and living conditions are poor, but they are unable to leave either because they in a situation where they owe debts to their employer or because they have no means to return to their home community. Agricultural industries, which have historically had the highest concentration of forced labor exploitation, are also industries critical to the Brazilian economy (Campbell, 2008). Often, forced labor exploitation in Brazil co-occurs with environmental crimes (Patric?a Trindade Maranh?o Costa, 2009). Workers are contracted to clear dense vegetation that grows in areas that have already been deforested and converted to pasture, a process referred to as ?juquira.? The deforestation arch of Brazil corresponds to 123,552,691 acres of land extending from the eastern and southern border of the state of Para through the states of Mato Grosso, Rond?nia, and Acre. This area has one of the highest incidences of forced labor exploitation in the country. Trees are cleared to make room for production of a variety of goods including meat, cotton, soy, sugar cane, coffee, black pepper, and charcoal. Forced labor exploitation has been identified in the production of all of these goods (Costa, 2009). Forced labor exploitation is also prevalent in the illegal logging industry. Since 2003, 931 individuals, predominantly men aged 15-20, have been identified to be in forced labor exploitation conditions by harvesting trees in the Amazon (Lazzeri, 2017). Forced labor exploitation is also prevalent in the cattle industry in Brazil. Workers may be involved in a variety of activities to support cattle ranching, including clearing land, monitoring livestock, and producing goods. The United States Department of Labor reports that cattle ranching and herding as well as beef 15 production are conducted with forced child labor in Brazil (Bureau of International Labor Affairs, 2018; U.S. Department of Labor, 2019). Brazil is the world export leader in beef production; Brazilian beef is commonly exported to the United States, Saudi Arabia, and Russia (Sharma, 2017). In many Brazilian states, cattle ranching is one of the only available stable employment options (Sharma, 2017). However, workers on cattle ranches are vulnerable to forced labor exploitation because they are geographically isolated (Costa, 2009). It is estimated that one-third of victims of forced labor exploitation in Brazil were exploited working on cattle ranches (Plassat, 2015; The Guardian, 2017). Federal police conducting a series of raids on cattle ranches in 2016 found that workers were kept on ranches through debt bondage; lived in degrading conditions without shelter, bathrooms, or water; and often received just one meal per day (Walk Free Foundation, 2018). Research on identified perpetrators of forced labor in Brazil finds that employers at cattle ranches are not typically individual estate owners, but businessmen with access to the most current technology, including computerized vaccination systems and artificial insemination for livestock. Workers, on the other hand, are denied access to clean water, food, housing and are subjected to abuse (Sakamoto, 2006; Walk Free Foundation, 2018). 2.2.3 Construction, manufacturing, mining, utilities, and hospitality Nearly half (47 percent) of forced labor exploitation victims worldwide work in urban industries like construction, manufacturing, mining, utilities, or hospitality. Victims in these industries generate over $34 billion dollars per year, accounting for roughly one third of profits from forced labor exploitation globally (ILO, 2014). 16 Although forced labor exploitation in Brazil was traditionally a problem that pervaded industries in remote, isolated areas, such as forestry and agriculture, increased economic growth and urbanization in Brazil has led to increased prevalence of forced labor exploitation in a wider range of industries (Costa, 2009; R. R. Figueira, 2012; Haddad & Miraglia, 2018). In 2013, for example, most identified victims of forced labor exploitation were found in construction or textile industries (Mello, 2016). Clothing produced in these factories is often part of the supply chain for large, multinational corporations who use exploited labor to keep overall costs low (Gagne, 2014; Kaos, 2016). 2.3 Vulnerabilities for forced labor exploitation Research finds that individuals are more vulnerable to forced labor exploitation if they express the following characteristics: they are migrant workers, they are younger, they are male, they have less than a primary school education, and/or they are poor (United Nations Office on Drugs and Crime, 2018a). Further, the Unites States Department of State and the United Nations report that historically marginalized groups such as Indigenous populations, religious minorities, and LGBTQ individuals are at greater risk for forced labor exploitation (United Nations Office on Drugs and Crime, 2017; United States Department of State, 2016). Children are also a vulnerable population, but specific groups of youth are at greater risk, including runaway youth, youth involved in child welfare systems, and Indigenous youth (United Nations Office on Drugs and Crime, 2018a; United States Department of State, 2019). Less research is available on perpetrators, but in general, they have similar characteristics as their victims: they are typically the same gender, the same 17 ethnicity, and live in the same area. Gender and ethnic match facilitate trust which makes easier for employers to recruit workers (United Nations Office on Drugs and Crime, 2018a). 2.3.1 Migrants and immigrants. Migration is a significant risk factor for forced labor exploitations. Individuals who leave their primary area of residence to find work in another state or country are at risk for a variety of coercive practices, including taking loans to cover their transportation costs, having passports withheld, and being isolated from their home with nowhere to go. When workers travel to communities with new cultures, they may be unaware of local laws and regulations and unable to speak the language, making them more vulnerable to dishonest and coercive practices from employers (International Labor Organization, 2014; United Nations Office on Drugs and Crime, 2018a; United States Department of State, 2019). Brazil?s growing economy brings migrants from all around the world in search of employment, including individuals from South and Central America as well as Bangladesh, Nigeria, Pakistan, Senegal, and Sri Lanka (Deutsche Welle, 2013; Walk Free Foundation, 2018). In a study based on 121 interviews with survivors of forced labor exploitation in Brazil, researchers found that migration was the most common characteristic of forced labor exploitation. Three quarters of study participants were rescued outside of their home community and 40 percent outside of their home state (International Labor Organization, 2011). Most victims did not have a spouse, but more than half reported having children and being the only working family member (International Labor Organization, 2011). In another case in 2013, an organized Brazilian gang smuggled 18 80 Bangladeshi workers into the country to work at warehouses producing clothing, in construction, and in car washing. The workers reported they owed the perpetrators $10,000 for transit to Brazil (Theuws & van Huijstee, 2015). Immigrant workers from other countries are highly vulnerable because they are undocumented, are isolated from their support systems, and may be linguistically isolated, leaving them unable to communicate with employers or the police (Shahinian, 2010). This causes migrant workers to depend on their employer for food, shelter, and medical care and because they are undocumented, they have no access to public services or legal employment in better conditions (Shahinian, 2010). 2.3.2 Young men Adult victims of forced labor exploitation tend to be younger than individuals working freely. Children comprise a greater proportion of forced labor exploitation victims in low-income countries relative to high-income countries (United Nations Office on Drugs and Crime, 2018a). Among children, those in forced labor and those in free labor tend to be about the same age (International Labor Organization, 2014; United Nations Office on Drugs and Crime, 2018a). Males are disproportionately likely to be victims of forced labor exploitation, accounting for 63 percent of victims globally (United Nations Office on Drugs and Crime, 2018a). Males are also less likely to report their victimization than are females, so the share of male victims may be higher than current estimates suggest (International Labor Organization, 2014; United Nations Office on Drugs and Crime, 2018a; United States Department of State, 2019). Women who are subjected to forced labor exploitation are more likely than their male counterparts to also be subjected to sex trafficking. There is more 19 accurate reporting of sex trafficking than forced labor exploitation, which may facilitate the identification of female victims (United Nations Office on Drugs and Crime, 2018a). In Brazil, trends in forced labor exploitation by gender parallel these global trends. Men aged 15-40 are disproportionately likely to be victimized (Issa, 2017). Men living in poorer states in the Northeast, such as Maranhao, Piaui, or Tocantins, typically migrate to states that have a higher demand for workers, like Para (Parente et al., 2017). Employers commonly confiscate personal documents which prevents these workers from changing jobs or returning home. Debt bondage may also keep workers stuck in a cycle of poverty and exploitation (Parente et al., 2017). 2.3.3 Low levels of education Victims of forced labor exploitation have lower levels of education than workers in free labor. Further, the parents of child victims of forced labor exploitation have lower levels of education than parents of children in free labor. Victims of forced labor exploitation are also less likely to have a literate head of household (International Labor Organization, 2014; United Nations Office on Drugs and Crime, 2018a). Again, these global patterns are paralleled within Brazil. Between 2013 and 2017, 71 percent of the identified victims of forced labor exploitation in Brazil were illiterate or had less than four years of education (Walk Free Foundation, 2018). 2.3.4 Poverty Victims of forced labor exploitation are disproportionately likely to live in poverty (International Labor Organization, 2014; United Nations Office on Drugs and Crime, 2009, 2018a). Families who experience declines in household income are 20 more likely to have members that become victims of forced labor exploitation (International Labor Organization, 2014). After the 2015 financial crisis in Brazil, younger and more skilled workers found themselves living in poverty, but rural populations were hit the hardest. Many individuals that were surviving off an agricultural career had to take on additional jobs to make ends meet. Rural workers are particularly vulnerable to forced labor exploitation because of informal working conditions in rural industries. For example, 40-50 percent of coffee harvesters work without a contract. Coffee plantation owners are also known to offer higher wages for individuals willing to work without a contract (Danwatch, 2016). 2.3.5 Marginalization LGBTQ and gender non-conforming individuals are vulnerable to forced labor exploitation because in many countries worldwide, there are no laws to prevent discrimination against these populations (United Nations Office on Drugs and Crime, 2017). The absence of anti-discrimination laws is only one indicator of a broader problem of engrained societal and cultural stigma and discrimination, which contributes to increased vulnerability. LGBTQ and gender non-conforming individuals may have weaker and fewer ties to family and other support groups. They are also more likely to be abused as youth, to run away as youth, or to be homeless (United States Department of State, 2019). Further, discrimination limits employment prospects and may therefore increase their vulnerability to coercive labor practices. Crimes are also less likely to be prosecuted when individuals are LGBTQ and/or gender non-conforming. Further, these individuals are more likely to injured, abused, 21 or even prosecuted for behaviors related to their own victimization (United States Department of State, 2019). Indigenous populations are also particularly vulnerable to forced labor exploitation. Relative to non-Indigenous persons, Indigenous people: are more likely to live in poverty, less likely to have official citizenship or documentation in their home country, have more limited access to education, and experience lower financial returns of education due to discrimination and linguistic and geographic isolation (Cruz-Saco, 2018; International Labor Office, 2017b; United Nations, International Labor Organization, & UNICEF, 2013; United States Department of State, 2016). Though research is limited on forced labor exploitation among Indigenous people, the ILO conducted an in-depth study in Bolivia, Paraguay, and Peru and found that Indigenous people were at an increased risk of forced labor exploitation, particularly bonded labor, and initiated a new program to combat bonded labor among Indigenous people in these areas (International Labor Organization, 2009b). A study conducted in Guatemala found that at least one in four Indigenous households had a member that was a victim of forced labor exploitation and that 65 percent of domestic workers in Guatemala are Indigenous girls and adolescents (International Labor Organization & United States Department of State, 2014; United Nations et al., 2013). Marginalized groups in Brazil are also at increased risk of forced labor exploitation, though research is limited. Indigenous populations and Afro-Brazilians account for roughly 51 percent of the population, yet these groups have higher rates of poverty (Walk Free Foundation, 2018). These groups also have lower rates of employment, which is exacerbated by their lower socioeconomic status within the 22 country (Lustig, 2015). High rates of poverty coupled with few job opportunities leaves these populations at increased risk for exploitation (McGrath, 2013). LGBTQ individuals are also marginalized and discriminated against in Brazil, leaving them more vulnerable to forced labor exploitation (Walk Free Foundation, 2018). 2.4 Causes of forced labor exploitation Forced labor exploitation in Brazil is deeply rooted in complex, historical issues including colonial slavery, poverty, deforestation, and unequal land distribution. These structural characteristics are the primary causes of forced labor exploitation and work to sustain the cycle of labor exploitation in the country. 2.4.1 Colonial slavery An estimated 4.9 million enslaved African people were brought to Brazil during the transatlantic slave trade: this represents roughly 40 percent of all enslaved people brought to North and South America (Costa, 2009; Issa, 2017). Brazil was also the last western country to abolish slavery and did so slowly in three phases. First, in 1871, the Free Womb Act freed all children born to enslaved mothers. Then, the Sexagenarians Act in 1885 freed enslaved people 60 years and older. Finally, the Golden Act abolished slavery outright in 1888 (R. R. Figueira, 2012; Issa, 2017). After the prohibition of slavery in Brazil, formerly enslaved people were free; however, these individuals had no access to land. The Brazilian government limited land access to a small number of individuals out of fear that abolition would crash the rural economy because employers would have to compensate laborers (Sakamoto, 2008). Before abolition, land was made freely available to be occupied. However, just 23 before the Golden Law was signed, ending slavery, the government introduced a new law where the state would sell the land rather than give it away for free. Since formerly enslaved people had no access to wealth, they were largely required to continue working for wealthy landowners (Sakamoto, 2008). Under the new land purchasing system and with the end of slavery, new forms of exploitation emerged. Two new forms of exploitation became particularly prominent: debt bondage and renting out lots of lands to immigrants brought to Brazil to work, predominantly on coffee on plantations (Costa, 2009; Issa, 2017). Thus, practices from colonial slavery transformed into a system of labor exploitation that continued to influence the relationship between laborers and employers even after abolition. Present day forced labor exploitation reflects this division between landowners and members of society without resources. In many ways, the historically rooted model of exploitation facilitated a system of exploiting vulnerable workers for profit (Costa, 2009; Figueria, 2005). Any efforts to understand and combat forced labor exploitation must take into consideration these historically rooted ideals that oppression and exploitation are necessary to sustain the economy. 2.4.2 Poverty and Globalization Most exploited workers in Brazil originate from the poorest regions which have the fewest economic and employment opportunities as well as the worst living conditions. Individuals living in the northeast region of Brazil are particularly vulnerable to exploitation. More than two-thirds of households in this region earn less than a quarter of the minimum wage and roughly three quarters of those households are Black or mixed-race (Costa, 2009). 24 Globalization is characterized by free trade, utilization of cheaper foreign labor markets, and the movement of individuals across country lines to fill the demand for low-cost labor. Globalization also displaces low-skilled workers and depletes environmental resources, and many experts argue that it is has widened the income gap between the richest and poorest members of society (Brewer, 2009). Some argue that ?the lesser developed countries of the world have become the factories and workshops for the developed countries? (Brewer, 2009). When job markets are displaced to other countries, domestic workers are left with few options for survival. This results in huge rates of out-migration as people search for job opportunities. In 2015, 244 million people?3.3 percent of the world?s population? lived outside their country of origin; this represents a 40 percent increase since 2000. Most migrants cross borders in search of better economic and social opportunities, while others are forced to leave their countries because of conflict and war (United Nations Office on Drugs and Crime, 2017). Forced labor exploitation is a multibillion-dollar industry, driven by demand for cheap goods and services. Globalization increases the supply of vulnerable workers who are unable to compete in the rapidly changing workforce. Forced labor exploitation generates an estimated $51.2 billion worldwide: $9 billion from forced labor in agriculture; $8 billion from forced domestic labor; and $34 billion from other industries such as construction, manufacturing, and mining. Further, the costs associated with forced labor exploitation are estimated to be about $21 billion; $19.6 billion in underpaid wages and $1.4 billion in illegal recruitment fees (International Labor Organization, 2014). When this is translated to profits per victim, forced labor 25 exploitation in construction, manufacturing, mining, and utilities generates an annual $4,800 per victim, in agriculture forced labor exploitation generates $2,500 per victim annually, and in domestic work forced labor exploitation generates $2,300 per victim per year (International Labor Organization, 2014). Brazil has experienced massive economic growth in the past 50 years, making it one of the most developed South American countries (Georges & Maia, 2017). There is also a huge wealth and income gap in Brazil, with the wealthiest five percent of the population earning as much as the remaining 95 percent (Georges & Maia, 2017). After ten years of poverty reduction, Brazil experienced a financial crisis in 2015, leaving more than twenty percent of the country in poverty. (World Bank, 2017). 2.5 Gaps in understanding forced labor exploitation As described above, we have a somewhat robust understanding of the economic sectors in which forced labor exploitation takes place, the individual-level factors that increase vulnerability for forced labor exploitation, and the historical and political factors that facilitate the supply of and demand for exploited labor. There are several areas that remain less clear and which require further inquiry. First, there is limited information on the characteristics of forced labor exploitation; that is, the physical and psychological ways in which employers recruit and exploit workers. Without further clarification around the characteristics of forced labor exploitation, the conceptualization and thus measurement of the crime remains ambiguous. There is also limited information about how these cases are processed in the court system. Consequentially, almost nothing is known about how prosecutors decide whether to 26 bring a case for prosecution or how judges decide whether to convict. Each of these gaps is interrelated and contributes to our overall lack of understanding of this crime. Without a strong conceptualization of the behaviors used to perpetrate forced labor exploitation, investigators do not know what kinds of signs to be on the lookout for or what types of evidence they should collect if they suspect forced labor exploitation is present. If investigators do not collect the right types of evidence, it will be difficult for prosecutors to build a case that is strong enough to bring to court. Further, without strong evidence- for example, if there is no witness testimony or the prosecutor is not sure how to interpret the legal code- it becomes increasingly unlikely that a judge will convict. Many times, there is no conviction so the case is never classified as a forced labor exploitation case. This means that future court practitioners and researchers will not study the case when learning about forced labor exploitation, which diminishes the pool of available data. Further, those cases that represent forced labor exploitation in case law are only those which overcame each of these obstacles, suggesting that our understanding of forced labor exploitation is based on only the most exceptional cases, which may vary substantially from the most typical. There is limited information on the characteristics of forced labor exploitation in Brazil, and worldwide (Patricia Trindade Maranhao Costa, 2009; DW, 2014; International Labor Office, 2017a; Owens, 2015). The bulk of information on the characteristics of forced labor exploitation comes from small, purposive samples, anecdotes, or stories documented by journalists. These small samples do not allow for rigorous quantitative analysis, and their findings do not elicit generalizable 27 recommendations for policy and practice (Owens, 2015). In this study, I use a large administrative dataset that contains detailed information about each case to generate typologies of forced labor exploitation. I use data from interviews with labor inspectors, federal police, judges, and prosecutors to guide my analysis and to provide additional context to quantitative findings. Information from the interviews also is utilized to describe the forced labor exploitation context in the next chapter. There is also limited information on how different characteristics of forced labor exploitation are associated with sentencing outcomes (Kappelhoff, 2008; OSCE et al., 2008). This gap in understanding is due in part to overall low levels of prosecution for forced labor exploitation in Brazil and worldwide. The United Nations Office on Drugs and Crime (UNODC) finds that of the 155 countries included in their biannual trafficking in persons report, 63 (41 percent) reported no convictions for forced labor exploitation in 2018 (United Nations Office on Drugs and Crime, 2018). While this demonstrates that there are overall low levels of prosecution, it does not explain why. In one study that examined 140 human trafficking cases in the U.S., the authors found that most cases met legal requirements for prosecution, but in general, suspects were never charged. The authors cited several barriers to prosecution and conviction in these cases including subjective legal definitions and interpretation of case law, lack of specialized training for court actors related to human trafficking, and perceived credibility of victims (Farrell et al., 2014a). I use a mixed method approach to explore how different typologies of forced labor exploitation are associated with outcomes in the criminal and civil court systems at the national level in Brazil. I also 28 interview labor inspectors, federal police, prosecutors, and judges to learn how cases are prioritized for investigation and prosecution. The findings from the Farrell and colleagues (2014) study mentioned above highlight a pervasive pattern in court processing of forced labor exploitation cases: what constitutes forced labor exploitation is subject to discretion at several stages in the criminal justice system. First, an investigator must determine that a crime meets the requirements; then, the criminal prosecutor determines whether the case is viable for prosecution; and finally, the judge must decide whether the weight of evidence constitutes a criminal activity. There is no research to date, however, on how decisions are made at each of these stages. My study draws from interviews with actors at every stage of the criminal justice system to determine how decisions are made, how different elements of the crime are defined, and what kinds of evidence would need to be present to support claims that those elements are present. I then compare the subjective interpretation of the law across practitioners to identify areas of concordance and disagreement. There are a handful of studies, based in the United States, that examine the decision making of investigators, prosecutors, or judges in forced labor exploitation cases (A. Farrell et al., 2008, 2014a; Kappelhoff, 2008). In this study, I use focal concerns theory to guide these questions so that I can establish which factors are most relevant to facilitating or inhibiting case processing and guilty convictions. This theory has been applied in U.S.-based studies, and I update the framework based on the local context in Brazil (A. Farrell et al., 2014a; Spohn, 2014). Examining the factors that influence decision making is particularly important 29 in Brazil because the forced labor exploitation law leaves a large margin for subjective interpretation. 30 Chapter 3: Forced labor exploitation and the Brazilian context In this chapter, I describe the background on the introduction of forced labor exploitation legislation in Brazil. I then describe public policy changes that have been implemented to combat forced labor exploitation. Next, I describe criminal and civil case processing in Brazil. In this section, I draw from data gathered during interviews to narrate this process, but I more fully describe my qualitative sample and methods in Chapter 4. I then provide a theoretical framework for understanding prosecutorial and judicial decision making in forced labor exploitation cases. 3.1 Introduction of forced labor exploitation legislation in Brazil Several high-profile instances of forced labor exploitation garnered public attention in Brazil in the late 1980s. Notably, in 1989, 17-year old Jos? Pereira and his fellow worker ran away from armed men holding rural workers captive in the state of Par? in southern Brazil (Patricia Trindade Maranhao Costa, 2009). The gunmen killed Pereira?s colleague and shot Pereira, at which point Pereira pretended he was dead. After his body was dumped off the side of the highway, Pereira found help and was taken to the hospital. After his recovery, Pereira detailed the inhumane conditions that more than sixty other men were forced to live and work in to the Federal Police, who then freed the workers and gave them money to return to their home communities (Patricia Trindade Maranhao Costa, 2009). The Pereira case was widely publicized as it represented a failure from the Brazilian government to protect human rights and provide legal and safety protections in the workplace (Patric?a Trindade Maranh?o Costa, 2009). Accordingly, in 1994, the Pastoral Land Commission, Centre for Justice and International Law, and Human 31 Rights Watch filed a petition with the InterAmerican Commission on Human Rights at the Organization of American States. The petition highlighted violations of the right to life, liberty, and personal security as well as violations of the American Convention on Human Rights which prohibit slavery and servitude (Patric?a Trindade Maranh?o Costa, 2009; DW, 2014) (Patricia Trindade Maranhao Costa, 2009; DW, 2014). The petition argued that the failure of the Brazilian government to act against exploitation of workers facilitated violent conditions like those faced by Pereira. Proceedings lasted several years, and in 2003, Brazil signed a settlement agreement, accepting responsibility for Pereira?s case. In the agreement, Brazil committed to four types of action: 1. Public recognition of responsibility for violation of rights in the Pereira case. 2. Financial compensation to Pereira for damages [52,000 Brazilian Reais (R$)]. 3. Commitment to prosecute and punish responsible parties. 4. Implementation of preventive measures, including legislative amendments; measures to monitor and reduce forced labor and to raise public awareness around the issue. The Brazilian government officially recognized the existence of forced labor exploitation in 1995; since that time, Brazil has made steady progress in combatting this crime. The Pereira Case, however, served as a catalyst for a series of reform efforts coordinated by the Brazilian government, human rights groups, and the ILO to reduce and prevent forced labor exploitation. For example, it brought to public attention the need to combat forced labor exploitation more effectively, particularly in rural areas, and highlighted the need for a more comprehensive definition of forced 32 labor exploitation. Today, the ILO has repeatedly upheld efforts made by the Brazilian government and non-governmental organizations in the fight against forced labor exploitation as a model for other countries to replicate (International Labor Office, 2005, 2017b; Internationales Arbeitsamt, 2009). The success of these efforts is hard to overstate; between 1995 and 2018, 53,000 workers were rescued by the Ministry of Labor and employers found to be using forced labor exploitation have paid more than R $100 million in damages (Santos, 2019). The ILO also recognizes Brazil?s forced labor exploitation legislation as one of the most progressive in the world. Article 149 includes elements related to debt bondage, degrading conditions, and exhausting work hours. All of these elements go beyond the minimum standards in ILO convention number 29 which defines forced labor as the main element of contemporary slave labor (International Labor Organization, 2001). The Brazilian government has implemented several distinct public policies to combat forced labor exploitation over the past twenty-five years. In the sections that immediately follow, I describe relevant public policies aimed at combatting forced labor exploitation in Brazil as well as historical and legal context surrounding the implementation of those policies. These policies are summarized on a timeline at the end of the section (see Figure 4). Finally, I will describe case processing for forced labor exploitation cases in the criminal and civil court systems in Brazil and present a conceptual framework for studying forced labor exploitation case processing. 3.1.1 Special mobile inspection group In 1995, Decree 1.538/95 established the Executive Group to Eradicate Forced Labor (GERTRAF). The purpose of GERTRAF is to implement and oversee a 33 program to prevent forced labor exploitation, including coordinating with the ILO and prosecutors at the state and federal level to prosecute employers accused of violating forced labor exploitation legislation. One of the main activities implemented by GERTRAF was the special mobile inspection group (Cria o Grupo Executivo de Repress?o Ao Trabalho For?ado e D? Outras Provid?ncias, 1995, p. 538). The special mobile inspection group is comprised of labor inspectors. The group is typically supported by federal or military police, who accompany the team for security, and other support based on the characteristics of the case. For example, participants may include federal and civil prosecutors, environmental police, victim advocates, defense attorneys, judges, translators, and others. The composition of the groups was described to me as follows: The mobile group is formed by labor inspectors. In general, there are four fixed inspectors in each group. Everyone linked to labor inspection can also participate, people from the Public Ministry of Labor, there are also labor prosecutors, also, not in all actions, but in some, they forward to the criminal prosecutor to attend. The public defender, too. In practically all inspections they also send the federal police or the federal highway police, to provide security. And it depends on the need for action. Sometimes it involves more environmental crime, so IBAMA will go, or the pastoral land commission. Some more specific things then have the most specific organizations to whom contact is made. Depending on the state, the local police will go as well, but the people I mentioned before are usually the ones who participate in one or another action quite regularly.- LA002 The purpose of the special mobile inspection group is to respond to complaints about labor law violations. One of the priorities of the group is to free workers from situations of forced labor exploitation, thus complaints for this type of labor violation are among the highest priority for inspections. The group is based in Brasilia, the nation?s capital; however, some states with a particularly high level of complaints 34 have their own local groups. Operations are covert so that employers cannot hide evidence ahead of the inspection. 3.1.2 First National Commission for the Eradication of Slave Labor In 2002, the ILO began a technical cooperation project, ?Combating Forced Labor in Brazil? (Patric?a Trindade Maranh?o Costa, 2009). Also in 2002, Brazil launched the First National Plan for the Eradication of Slavery (CONATRAE). This replaced the Executive Group to Eradicate Forced Labor (GERTRAF). CONATRAE is the monitoring body created from the National Plan to Eradicate Slavery. The purpose of the plan is to implement 76 measures to guide Brazilian actions to combat forced labor exploitation. 3.1.3 Second National Commission for the Eradication of Slave Labor The first commission was viewed largely as a success, so in 2008, a new document was created that filled in gaps and created more attainable objectives. The overarching goal of the Second National Commission for the Eradication of Slave Labor continues to be to eliminate forced labor exploitation in Brazil (Patric?a Trindade Maranh?o Costa, 2009). 3.1.4 National Day to Combat Forced Labor Exploitation In 2009, Law 12064/09 established January 28 as a national day to combat forced labor exploitation. This law was created in response to a tragic crime that occurred in 2004. Three labor inspectors and their driver were murdered while conducting a routine inspection of a farm (Lacerda, 2020). Four men admitted to being involved 35 and were found guilty for ordering the assassination of the inspection team, but only one man received a prison sentence; the rest are free (Lacerda, 2020). 3.1.5 Unemployment Insurance Benefit for Victims of Forced Labor Exploitation In 2002, Brazil also enacted Law 10.608/02 which provides unemployment insurance for workers verified to have escaped from or been rescued from conditions of forced labor exploitation. Workers who are fired without just cause are also eligible for this temporary benefit. Qualified workers receive no less than their minimum pay for between three and five payments (Regula o Programa Do Seguro- Desemprego, o Abono Salarial, Institui o Fundo de Amparo Ao Trabalhador (FAT), e D? Outras Provid?ncias, 1990). For workers to qualify for unemployment insurance benefits, they must submit several documents, including: ? Work and Social Security Card Book. ? Conditions for Termination of Contract. ? Social Integration Program Card (PIS-PASEP) (Regula o Programa Do Seguro-Desemprego, o Abono Salarial, Institui o Fundo de Amparo Ao Trabalhador (FAT), e D? Outras Provid?ncias, 1990). 3.1.6 Clarified Legal Code In 2003, the government modified Article 149, which defines forced labor exploitation, to provide clarification around the different ways in which a worker could be exploited. Prior to the amendment, the law defined the offense as ?reducing someone to conditions analogous to slavery.? Following the 2003 modification, the law specifies actions including: debt servitude, degrading conditions, exhausting working hours, preventing the worker from leaving the workplace by restricting their use of any means of transportation, confiscating or withholding workers? 36 documentation, personal property, or maintaining manifest surveillance (Altera o Art. 149 Do Decreto-Lei No 2.848, de 7 de Dezembro de 1940, 2003). 3.1.7 Dirty List In 2004, Brazil implemented the Registry of Offending Employers through decree 540/2004, more commonly known as the ?Dirty List.? The Dirty List is a publicly available list of employers, including individuals and companies, found guilty of violating labor legislation. Error! Reference source not found. details the process for inclusion in the Dirty List. 37 Figure 3 Process for Inclusion in the Dirty List Special mobile If evidence of forced Complaint filed with inspection group labor exploitation, government or civil society investigates the charges filed with workplace Ministry of Labor and Employment If employer found Dirty List is published guilty they will be on Ministry of Dirty List is updated included on the Dirty Economy website and newspaper Reporter every 6 monthsList Brasil If an employer does not reoffend for two years and pays all fines they are removed from the list Companies on the Dirty List suffer substantial financial consequences. In 2010, the Central Bank of Brazil issued Administrative Rule no. 3876, which prohibits banks from granting rural credit to individuals and companies on the Dirty List. In 2014, Constitutional Amendment no. 81, determined that any properties where forced labor exploitation is identified would be taken away from the owner. Many international banks have also signed a pact to boycott companies on the Dirty List. Thus, when companies appear on the Dirty List, it can and does substantially affect business. For example, the day after popular Spanish clothing brand Zara appeared on the Dirty List, their shares fell nearly 4 percent on the Madrid Stock Exchange (Kelly, 2013). 38 3.2 Impact of public policies targeting forced labor exploitation Official data suggest that there was a significant improvement in combatting forced labor exploitation following the revision of Article 149. Statistics from the Ministry of Labor, displayed in Table 1, show significant increase in results of inspections after 2003, especially when compared with inspections in the 1990s. Table 1. Labor inspections conducted by the Public Ministry of the Economy between 1995 and 2013 Civil Fines Year # Operations # businesses inspected # workers rescued paid ($RS) 1995 11 77 84 Not available 1996 26 219 425 Not available 1997 20 95 394 Not available 1998 17 47 159 Not available 1999 19 56 725 Not available 2000 25 88 516 $ 472,849.69 2001 29 149 1305 $ 957,936.46 2002 30 85 2285 $ 2,084,406.41 2003 67 188 5223 $ 6,085,918.49 ?Law modified 2004 72 276 2887 $ 4,905,613.13 2005 85 189 4348 $ 7,820,211.26 2006 109 209 3417 $ 6,299,650.53 2007 116 206 5999 $ 9,914,276.59 2008 158 301 5016 $ 9,011,762.84 2009 156 350 3769 $ 5,908,897.07 2010 142 310 2628 $ 8,786,424.89 2011 170 341 2485 $ 6,159,707.42 2012 141 255 2750 $ 9,676,387.36 2013 179 300 2063 $ 8,236,288.02 Since 2003 there has been a significant increase in the number of inspections conducted by labor inspectors. For example, in 2002, there were 30 inspections, and in 2013, there were 179 inspections. There has also been a significant increase in the fines paid by employers; in 2002, employers paid R $ 2,084,406.41, and in 2013 they 39 paid a cumulative R $ 8,236,288.02. There was also an increase in the number of businesses inspected (85 in 2002 and 300 in 2013) and the number of workers rescued (from 737 on average between 1995 and 2002 to 3,379 on average between 2003 and 2013). Despite these improvements, there is no evidence that criminal prosecutions have increased or that punishment for employers found to be guilty of forced labor exploitation is more certain. Combatting forced labor exploitation remains a significant challenge. Despite significant policy and legal reform efforts focused on eliminating forced labor exploitation, there exist gaps in the Brazilian response to this crime. For example, in 2016, the Inter-American Commission on Human Rights recommended that to strengthen the response to forced labor exploitation, the government should create mechanisms for coordination between the civil and criminal court systems in order to overcome gaps in the investigation of specific crimes and to strengthen the legal response and enhance the certainty of prosecution and punishment for employers responsible for exploiting workers. Increased public policies and efforts to understand and categorize this crime can only help to a certain degree; if the justice system does not effectively respond to this crime, then hundreds of thousands of workers in Brazil will continue to be subjugated to criminally exploitive working conditions. 40 Figure 4. Timeline of forced labor exploitation policies in Brazil 3.3 Case processing of forced labor exploitation in the criminal and labor court systems The investigation and processing of forced labor exploitation cases in Brazil is a complex process. I asked about each step of this process during interviews. Below, I summarize the main steps of case processing, including: inspection, forwarding cases, the labor court system, and the criminal court system. I will more fully describe my qualitative sample and methods in Chapter 4. 3.3.1 Inspection Below is the process of conducting a ?typical? inspection as described by labor inspectors that I interviewed. 1. Filing a complaint 41 In Brazil, there are three primary ways that a forced labor exploitation case becomes known to investigative bodies. There are several channels for receiving complaints. Depending on the region, it could be made through the pastoral land commission?Dial 190 is also a channel that receives complaints?This month the Public Ministry of Work also launched a website where workers can file complaints. Anyone with access to the internet can file a complaint that way.- LA002 Complaints can be made through phone hotlines, WhatsApp lines, an internet portal, or directly to the ministry of the economy, ministry of labor, or pastoral land commission. Until recently, it was on-site, they had no online channel and now there is an online channel for complaints. We also receive complaints through channels from other ministries, and phone hotlines, 190 and others. There is also a telephone number of the state commission to eradicate work analogous to slavery, a cell phone that also receives texts through WhatsApp, sometimes they also enter the ombudsman's websites, it all comes to us.- LA004 The police may also identify forced labor exploitation through their routine patrols or as a result of investigating another type of crime. In this case, police would file a complaint directly with the Division of the Eradication of Slave Labor. 2. Prioritizing complaints After the complaint is filed, it is forwarded to the Division of Inspection for the Eradication of Slave Labor. This Division prioritizes cases and identifies a set of cases in the same geographic area then sends this set of cases to one of four mobile inspection groups. These complaints are then forwarded to the Division of Inspection for the Eradication of Slave Labor in Bras?lia, which analyzes the severity and urgency, and combines the complaints with others from the same region, then organizes the logistics to justify the departure of a whole group there. So, for example, if there are five or six complaints from the Mato Grosso region and you need a group to go, then they are forwarded to one of the four [mobile inspection] groups. -LA002 42 3. Preparing for the inspection The mobile inspections groups have a coordinator who determines the logistics of an inspection, including prioritizing the order to visit each site. This group of complaints is forwarded to the group?s coordinator and we are assigned a scheduled date for travel to investigate the group of complaints. The coordinator analyzes the logistics to determine where will go first. - LA002 The coordinator prioritizes cases based on severity and urgency. Urgency refers to whether the type of work being exploited is short-term- such as is the case in deforestation- or long-term- for example, in the production of charcoal or domestic work. Inspectors must respond to short-term work sites more quickly or they risk arriving to a site with no workers, no employers, and no evidence. Severity is a more subjective term but can refer to the population being exploited; for example, complaints that indicate children or marginalized groups are being exploited will receive priority. Severity can also refer to the degree of exploitation; imminent threats to the lives of workers will receive priority, as is often the case when there is the presence of weapons or physical violence. We try to give priority to child labor and slave labor, sometimes, the lack of people [on our team] causes us to take a long time to organize a team and sometimes we arrive late at that time and people are no longer there. Sometimes that slave labor and temporary work, it is the construction of a fence, or deforestation and when we arrive, the workers are no longer there. - LA004 We prioritize the most recent and the worst. The more perennial the activity, for example charcoal making, there is not so much problem in acting a while after the complaint. The faster the work, for example: deforestation, the inspection must be quick. Meeting all demands in the shortest possible time would be the best of all worlds-LA008 4. Conducting the inspection 43 During the investigation, the special mobile inspection group determines whether a case constitutes forced labor exploitation. Notably, the special mobile inspection group is trained by the Ministry of the Economy (Ministry of Labor before 2019). Thus, they are primarily concerned with identifying proof of a violation of labor laws; they are not criminal investigators. The burden of proof needed for a civil sentence is lower than the burden of proof needed for a criminal sentence, like in the United States court system. The labor inspectors take photos, videos, and written notes documenting evidence for each component of Article 149. Then we go to the establishment, do the physical inspection, interviews with employees and the employer and after all the analysis of what we collected during the inspection, we issue a report and make an infraction notice if there is any irregularity, we see with the employer to rectify the situation. This is a general situation.- LA003 The mobile inspection group may also identify forced labor exploitation during routine inspections of employer worksites. Figure 5 shows an example of one page of a report from 2016 that I received permission to publish. This page does not include identifying information. Individuals in the top right picture are members of the special mobile inspection group. Federal police typically accompany labor inspectors during investigations, but their role is to provide security to the labor inspectors. They do not conduct a criminal investigation. In their reports, the special mobile inspection group determines whether there is evidence of forced labor exploitation. 44 Figure 5. Page of a report from a labor inspector. Captions read (Top Left) Below, pictures from the structure where four workers lived. (Top Right) Signs of recent deforestation. (Middle Right) View of the external area of the sight where we found the living structure. 5. Victim rescue 45 If the labor inspectors identify forced labor exploitation, they notify the employer that they must make changes to comply with criminal and labor laws regarding workers? rights. If the employer does not agree to these conditions while the labor inspectors are there for the inspection, the mobile inspection group removes the workers and sets them up with accommodation, unemployment insurance, and other social benefits. Typically, labor inspectors travel for inspections for two weeks at a time and spend 2-5 days at a given job site. One labor inspector described this process as follows: If we identify [forced labor exploitation] we notify the employer that he has to remove these workers from the situation. This means finding a regular accommodation that minimally meets our standards for workers. If the employer does not comply with the notification, we send a report to the public ministry of labor to file a public civil action. We will then file, on the worker?s behalf, a job ?termination? as if he had been fired. This way he receives all of his labor rights and the prosecutor's office can then determine moral damages for that situation. In addition, we issue unemployment insurance for the worker so he receives, from the government, the equivalent of three minimum wages, and unemployment insurance for six months. We also forward the letters to the public assistance secretary, so that he has priority in registering for the Bolsa Fam?lia, which is an income program.- LA003 When the special mobile inspection group rescues workers from a situation of forced labor exploitation, they calculate the amount of working time for each worker, prepare and immediately cancel a contract for the worker, calculate what the appropriate labor compensation should have been, and issue a temporary work card. This allows the workers to receive unemployment insurance for up to five months (Patricia Trindade Maranhao Costa, 2009). 46 3.3.2 Forwarding the labor inspectors? findings After an inspection is completed, the coordinator for the mobile inspection group prepares a detailed report which is accompanied by any records of infractions or fines that were given to employers at the time of the inspection. This report also includes photographs, videos, and any witness testimony or other documents collected during the inspection. If the team finds evidence of a crime, the Labor Inspection Secretariat in Brasilia sends a copy of the inspection report to the Federal (Criminal) Proseuctor?s Office, the Public (Civil) Prosecutor?s Office, the Federal Police, the Regional Labor Precinct with jurisdiction in the state where the tax action was taken, and a federal government land authority (INCRA), if appropriate under Ordinance no.101 (January 12, 1996). The report is submitted to the Ministry of Labor. The Ministry of Labor then forwards the case simultaneously to prosecutors in the labor (civil) court and the criminal court systems. From here, civil and criminal prosecutors must build the case and work towards obtaining a conviction. Notably, actors from the civil and criminal courts do not typically communicate with each other about the cases, even though they are prosecuting the same case. Whether a case receives a conviction in labor court has no bearing on culpability in the criminal court system. 3.3.3 Labor court system In labor court, defendants, if convicted, may be required to pay fines, restitution to each victim, or pay back taxes and withheld wages. They may also enter into a plea agreement called a TAC (?termo de ajuste de conduta?). In the TAC, the defendant agrees to discontinue illegal behaviors and to repair damages (typically through 47 paying fines), and in return, the government does not take any further legal action. Decisions from the labor court can be appealed to Regional Labor Court, Superior Labor Court, and finally to Supreme Federal Court. 3.3.4 Criminal court system In criminal court, punishments for forced labor exploitation may include fines and two to eight years in prison. It is rare for a defendant to receive prison time. Forced labor exploitation may not be settled through a plea agreement in criminal court. Additionally, forced labor exploitation cases are not eligible for a jury trial. In Brazil, particularly serious crimes, such as forced labor exploitation are decided only by a judge. In an analysis of cases from the state of Minas Gerais, between 2004 and 2017, 373 reports were submitted by labor inspectors. Of these 373 reports, 157 identified forced labor exploitation, 79 were prosecuted in the criminal system, 21 were found guilty, and one defendant was given a prison sentence (Haddad & Miraglia, 2018a). In criminal court, the statute of limitations timer begins as soon as the defendant is made aware of the charges being filed against them. The statute of limitations for Article 149 is 12 years. While defendants may be arrested during the special mobile inspection investigation, defendants are not held in pre-trial detention and are free to return home or travel until the trial. There is typically no halt to their business (other than workers being rescued), and in theory, they could replace lost workers and continue business as usual. Federal prosecutors may require an investigation to be conducted by the federal police as a supplement to the report submitted by the special mobile inspection group. 48 A study conducted by the anti-slave labor and trafficking in persons clinic in the Law School at the Federal University of Minas Gerais (UFMG) examined the court response to forced labor exploitation cases in the state of Minas Gerais between 2004 and 2017. Of the 39 public civil actions, 21 public civil actions (53.8 percent) were partially upheld, and 13 (33.3 percent) were resolved through a settlement agreement. Of the 21 actions that were upheld, just nine acknowledged the existence of "degrading conditions" and another, similarly, mentioned "unworthy conditions" (Haddad & Miraglia, 2018b). Of the 79 criminal cases identified between 2004 and 2017, 32 received criminal sentences. In these 32 criminal sentences, 61 individual defendants were implicated. Of these 61 defendants, 21 defendants were convicted and 32 acquitted. Of the 21 defendants convicted, only three had received a sentence as of the end of the study period: in one of the cases, the statute of limitations expired and the defendant was free; in the second, a fine and a restrictive penalty were applied; in the third, the defendant was arrested and served prison time. This study is the only to date to map the criminal and labor responses to forced labor exploitation in Brazil. No such study has been conducted at the national level. Based on the findings in Minas Gerais, there is an indication that the court response to forced labor exploitation is not timely, and at least in the criminal justice system, there is a sense of impunity for this crime. If employers can profit from the exploitation of workers with relatively little risk of punishment, this may facilitate the continued participation in this crime. 49 3.4 Theoretical framework to understand court response to forced labor exploitation in Brazil. Focal concerns theory has been used to explain judicial and prosecutorial decision making (Albonetti, 1986; D. Steffensmeier et al., 1993, 1998; D. J. Steffensmeier, 1980). It emphasizes two major groups of factors said to influence judicial and prosecutorial decision making: legal and extralegal factors. When making decisions, judges and prosecutors are guided by three important groups of focal concerns: blameworthiness of the offender; the desire to protect the community from the offender; and concerns about the practical consequences of the sentencing decision, including social costs (D. Steffensmeier et al., 1998). Blameworthiness refers to the role the defendant had in the offense, the degree of harm the victim suffered, and whether the defendant admits responsibility. More blameworthiness is associated with an increased likelihood of punitive actions. Focal concerns theory describes community protection as the decision maker?s calculation about the potential danger to the public if the defendant reenters the community. For judges, this calculation often includes the severity of the crime, whether violence was involved, and whether they think the defendant will re-offend. The greater the potential threat to the community, the more punitive the sanctions. The final component is practical considerations which suggests that decision makers consider things like whether the defendant has children, who would care for them if the defendant is incarcerated, the capacity of prisons at the time of sentencing, or even the overall case load for the court. Overall, decision makers are influenced by the 50 organizational realities of the court system and the degree to which a sentence would negatively impact the defendant (Albonetti, 1986; D. J. Steffensmeier, 1980). While focal concerns is most often applied to judicial decision making in the criminal court system, it has also be used to explain decision making in other contexts including decision making of law enforcement officers, of court actors in specialized courts, and in civil courts (Alderman, 2017; Higgins et al., 2012; Ray & Dollar, 2013; Roberts, 2016). Focal concerns can also be applied to prosecutorial decision making. When applied to prosecutors? decisions about whether to pursue a case, the focal concerns are similar. Prosecutors are more likely to file charges if the crime is serious, the victim has objectively been harmed, and there is substantial evidence of culpability (Spohn et al., 2001). However, research suggests that concerns about the third group of focal concerns--practical consequences-- are different for prosecutors and judges. Research finds that prosecutors tend to focus primarily on the likelihood of conviction while judges are concerned primarily with social costs of sentencing. In the context of decision making for forced labor exploitation specifically, there is less research, but one study suggests that prosecutors? decisions are influenced by stereotypes of ?real crime and genuine victims? (Hawkins, 1981). Underlying this theory is the assumption that judges and prosecutors have limited time to devote to any one given case and that they expedite their decision making using a personal checklist of key factors. This checklist includes the three focal concerns as well as lived experiences with certain types of offenses or defendants, legal policies and procedures, and group stereotypes (Albonetti, 1986; D. Steffensmeier et al., 1993, 1998; D. J. Steffensmeier, 1980; Ulmer, 2012). Many 51 researchers have pointed out that an important limitation of focal concerns theory is that it does not detail the mechanisms by which the different factors that influence decision making lead to case outcomes or take into account court workgroup dynamics. Some scholars have tried to unpack this relationship and found that stereotypes are important to connect focal concerns with case outcomes (Albonetti, 1986; R. A. Farrell & Holmes, 1991; Ulmer, 2012). Below, I outline the ways in which focal concerns theory can be applied to forced labor exploitation and suggest potential mechanisms that link focal concerns with case outcomes. 3.4.1 Focal Concerns Theory and forced labor exploitation Forced labor exploitation is unanimously viewed as a severe crime that causes substantial harm to victims and poses a grave threat to society, so understanding how focal concerns come into play in decisions about forced labor exploitation is somewhat different than for other crime types. Whether forced labor exploitation is present or not is the pivotal decision that prosecutors and judges must make, rather than having to decide how severe the crime itself is. In cases of forced labor exploitation, the importance of avoiding uncertainty is even greater because court actors do not want to falsely accuse or convict a defendant of such a serious crime (with severe penalties), and on the other hand, they do not want to let a perpetrator of this crime back into the public to continue exploiting vulnerable community members. According to prior research on prosecution of human trafficking cases, a key consideration for prosecutors when deciding whether to bring a case forward is the perceived credibility of victims, victim cooperation, and negative attitudes about 52 victims of human trafficking (A. Farrell et al., 2008; Spohn, 2014). There are key differences between US-based studies, like the Farrell, McDevitt, and Fahy (2008) study, and the context in Brazil, however. First, US studies are based almost exclusively on cases of exploitation for sexual labor (i.e., sex trafficking). Victim credibility in these cases is arguably more critical because of stigma against sex work and stereotypes around women?s culpability for their victimization in sex crimes. Second, human trafficking cases in the US are eligible to receive a jury trial. In Brazil, however, the most serious types of crime are not eligible for a trial and the decision falls exclusively on the judge. Thus, while victim credibility is important, the dynamics of proving a credible victim to a judge are different than proving credibility to a jury. Because judges have ultimate decision-making authority for forced labor exploitation cases in Brazil, they must grapple with the equal importance of avoiding sentencing an innocent defendant and acquitting a guilty defendant. Focal concerns can be applied to understand how judges minimize uncertainty in these cases. 3.4.2 Avoiding uncertainty At its core, focal concerns is a theory about avoiding or eliminating uncertainty (Albonetti, 1986; Spohn, 2014). There are two primary factors that allow court actors to be certain about a case: the weight of evidence and alignment of details of the case with decisions from case law. In the absence of these legal factors, however, decision makers will turn to extralegal factors, notably the three groups of focal concerns. When considering what constitutes uncertainty from the perspective of prosecutors and judges, it is important to understand the types of information that 53 these groups use to make decisions. In the criminal justice system, a history of decisions about cases of the same type, referred to as case law, is often described as the most important basis for decision making. Robust case law allows prosecutors and judges to more heavily rely on legal factors when deciding a case. Forced labor exploitation cases, however, are rarely identified and even more rarely brought forward for prosecution. Further, there is no formal training either in school or on the job for Brazilian prosecutors or judges related to forced labor exploitation. They must rely on experience alone when making decisions, and these cases are rare so even individuals ?experienced? in forced labor exploitation cases have typically worked on fewer than ten total cases over the course of their career. One judge I interviewed with 25 years of tenure had only received one forced labor exploitation case in her career. This means that prosecutors and judges alike do not have the benefit of substantial case history from which to determine whether certain situations constitute ?real crime.? In the absence of a clear path to a decision based on legal factors alone, prosecutors and judges will turn to focal concerns and other extralegal factors to make decisions. There is also not a direct relationship between the weight of legal evidence and certainty of criminal culpability. As outlined in focal concerns theory, individual calculations related to focal concerns, as well as internalized stereotypes about groups of defendants and victims, often influence how prosecutors or judges interpret legal evidence. 54 3.4.3 Focal Concerns Blameworthiness. In an ideal world, characteristics of an offense would directly predict the weight of evidence available for a given case: investigators would document evidence that reflected the actual crime as it occurred. This evidence, in turn, would directly predict prosecutor?s certainty about whether a case is viable for prosecution and judges? certainty about criminal culpability. In the real world, however, the available evidence is not directly determined by characteristics of the offense. When examining the evidence that is available, prosecutors and judges alike will likely be more certain of culpability when elements such as physical violence, weapons, and manifest surveillance are present. These physical forms of coercion and control are easy to prove (e.g., through photographs) and are unarguably forms of control. A case may be less certain if there is no physical restriction of movement; for example, if there is psychological control, geographic isolation, or debt servitude. Other logistical factors that may also increase certainty of the crime are the number of victims and a shorter time between the investigation and the court date. Community protection. Judges and prosecutors may be more concerned with the threat to the community when forced labor exploitation cases include a greater number of victims or when more marginalized groups were victimized (such as children, women, or Indigenous persons). Additionally, they may be more concerned with the threat to the community when other crimes co-occurred with the offense and what threat those crimes pose to individuals or the environment. Practical considerations. There are several additional practical considerations that come into play with forced labor exploitation cases. First, judges 55 may consider the defendants socioeconomic status. An individual farm owner would be devastated by the fines and prison sentence associated with a guilty conviction whereas a large corporation would likely be able to continue business as usual. Prosecutors may also consider the time that elapsed since the investigation because a greater amount of time results in a decreased likelihood of identifying witnesses to give statements. Also, the more time that elapses, the less reliable individual recounts of the events will be. Prosecutors may be less likely, then, to move forward with older cases. Individual experiences and stereotypes Perhaps the most important mechanism by which prosecutors and judges apply their calculation of focal concerns to case outcomes is their own understanding or experiences with the types of situations presented in the case. There is a substantial wealth gap in Brazil, with the top five percent of earners having as much wealth as the remaining 95 percent. Judges and prosecutors are overwhelmingly from middle- or upper-class backgrounds. Thus, these individuals are unlikely to share lived experiences with victims or perpetrators of forced labor exploitation. When there is little or no knowledge or exposure, decision makers will fall back on stereotypes or assumptions about the individuals involved. For example, a defining characteristic of forced labor exploitation cases in Brazil is degrading conditions (Patric?a Trindade Maranh?o Costa, 2009), yet there is no definition of what degrading conditions means. Since judges and prosecutors are unlikely to have experiences with the types of work involved in forced labor exploitation cases, or with poverty and marginalization in general, stereotypes likely 56 affect whether judges and prosecutors perceive different situations as degrading. Most individuals that become judges and attorneys do not have lived experiences with poverty. However, deep poverty is a reality for many individuals subject to forced labor exploitation. Judges and prosecutors, therefore, have difficulty distinguishing what they perceive to be normal living and working conditions for people in poverty versus criminally degrading living and working conditions. 3.4.4 External Influences Finally, there are a group of factors that influence case processing that are outside the control of prosecutors and judges. These factors can influence the collection of evidence, speed of case processing, and ultimately, final outcomes in a case. Political restraints relevant to the context in Brazil include resources allocated to the special mobile inspection unit and federal police to conduct inspections, establishment and maintenance of specialized units to combat forced labor exploitation, trainings (or lack thereof) of court actors on forced labor exploitation, and time that elapses between investigation and prosecution due to the court?s existing case load. Priorities of labor inspectors vs. priorities of criminal justice system. Another unique feature of forced labor exploitation case processing is that labor inspectors, rather than police, are responsible for nearly all investigation and collection of evidence. Labor inspectors, however, work for the Ministry of the Economy. They receive extensive formal training on forced labor exploitation, unlike judges and prosecutors, prior to beginning their careers on special mobile inspection units. Labor inspectors? goals and motivations are not intentionally aligned with or 57 tailored to the goals of the criminal justice system; yet, they serve as the gatekeepers to the types of cases that are processed and the evidence available. For example, a prosecutor or judge may prioritize cases that most closely align with the types of cases that have received convictions in the past (based on case law) or they may prioritize cases that meet the stereotypical understanding of what constitutes forced labor exploitation (rural, isolated, with presence of armed guards). Labor inspectors? goal is not to secure a conviction in criminal or labor court, but to protect the lives and wellbeing of individual workers; however, they can impose civil fines on employers. Labor inspectors therefore prioritize cases based on urgency and severity assessed from the information available in complaints. I used focal concerns theory to guide my analyses, including to develop and adapt data collection protocols, inform model specification, and provide a conceptual framework from which to understand qualitative findings. This study fills several important gaps in the literature on forced labor exploitation. First, I leverage a large administrative dataset that contains detailed information from the investigation stage through the sentencing and adjudication stages in both criminal and labor court systems in Brazil. Second, this is the first study to identify latent typologies of forced labor exploitation and to then explore how those typologies of forced labor exploitation are associated with sentencing outcomes in the criminal and labor court systems. Third, this is one of only two studies to examine different case outcomes for forced labor exploitation cases in Brazil, and the only study to do so at a national level (Haddad & Miraglia, 2018b). Fourth, this is the first study in Brazil, and one of a few globally, that examines how 58 court actors make decisions about whether to bring a forced labor exploitation case forward for prosecution, and whether to find the defendant(s) guilty (A. Farrell et al., 2014a; Go?dziak & Bump, 2008; Kappelhoff, 2008; Laan et al., 2011). Finally, this study sheds light on the types of evidence needed to prove different elements of forced labor exploitation in court. 59 Chapter 4: Research Questions, Data, and Methods The proposed study leverages a robust quantitative dataset as well as original data collection with labor inspectors, police, and prosecutors to achieve three primary objectives: ? Describe characteristics of forced labor exploitation cases forwarded for processing in criminal and civil courts; ? Analyze differences in types of forced labor exploitation by key characteristics; and ? Examine the sentencing outcomes for different types of forced labor exploitation cases. To achieve these objectives, I examine the research questions outlined in Table 2. Table 2. Research questions, data sources, and method. Objective Research Questions Data Source Method O1: Describe What are the characteristics of characteristics of known forced labor exploitation forced labor cases in Brazil? exploitation cases Administrative Latent class forwarded for Data analysis processing in criminal and civil courts O2: Analyze What factors influence the type Qualitative Qualitative interviews analysis differences in of forced labor exploitation activities used (e.g., geography, Administrative Multinomial typologies of economic sector, co-occurring Data logistic forced labor exploitation by key criminal activities, size of regression characteristics operation)? What are the outcomes of cases Qualitative Qualitative O3: Examine the sentencing prosecuted with forced labor interviews analysis exploitation charges? Administrative Multinomial outcomes for Data logistic different types of regression forced labor exploitation cases How are different typologies of Qualitative Qualitative forced labor exploitation cases interviews analysis associated with criminal Administrative LCA with sentences? Data distal outcome 60 4.1 Administrative data I used a dataset developed by the Anti-Slave Labor and Trafficking in Persons clinic (the clinic) in the law school at the Federal University of Minas Gerais in 2018- 2020. This dataset links information from labor inspection reports with case processing and outcome data. The dataset includes data on the sentences and judgments handed down within the scope of the five Federal Regional (criminal) Courts and the 24 Regional Labor (civil) Courts, in addition to the Superior Court of Justice, Superior Labor Court, and the Supreme Federal Court, therefore encompassing the entire Brazilian territory and all judicial bodies. To develop the dataset, the clinic team requested a list of all criminal actions and public civil actions from the Federal Public Ministry and the Public Labor Ministry, respectively. The team then reviewed the courts? websites to identify any other lawsuits that were missing from the lists and then requested data from those cases specifically. Upon review of these lawsuits, the clinic team discovered that most criminal actions were initiated in 2008 and most civil lawsuits began in 2012. Thus, they limited their sample to cases initiated in the criminal court system in 2008 or later and in the civil court system in 2012 or later. The team also thought it would be analytically advantageous to limit the sample to criminal cases that were initiated in 2008 or later because all cases would be reviewed against the revised version of Article 149 (revised in 2003). Establishing 2008 as the initial time frame also facilitated access to data on case outcomes, given that not all sentences and judgments handed down in the early 2000s are available on the institutional websites of the regional federal and labor courts. 61 The clinic team searched for procedural changes (e.g., if the case was rejected or dropped for any reason) as well as sentencing decisions for each case on the courts? websites. Some cases were confidential and, therefore, it was not possible access them or to see their procedural progress. Additionally, some cases, despite containing information on all the procedural changes, did not present the full detail about sentencing. In cases where there was not complete data for a case, the team searched on sites like JusBrasil, which is a privately run website that makes court data publicly available. When the team could not find the missing case information on JusBrasil, they made direct contact with the court via email and telephone. When the team could still not fill in details after these attempts, they excluded the case from the dataset. I do not have information on precisely how many cases were excluded for this reason. For cases in the labor court system, the team excluded all cases that were not public civil actions (i.e., cases brought forward by the government). That is, any individual lawsuit from a single employee brought against their employer is not included in the data. The team also excluded any cases that did not include charges under Article 149 of the penal code (forced labor exploitation). During the initial review of the data, the clinic team identified a few public civil actions with incomplete data (for example, data on appeals only and nothing prior). Upon further exploration of the Public Labor Ministry website, the team identified an additional 23 public civil actions that took place between 2008 and 2011 that were related to incomplete cases that were included in the original sample frame of cases from 2012 62 onward. These 23 cases are included in the final dataset despite having a start date prior to 2012. In some instances, it was not clear whether a case should be included in the dataset. For example, in criminal cases that did not end up moving to sentencing, the court?s decision to receive the complaint was not available. In this situation, there was no data on the crimes reported in the complaint. In these cases, the team chose to trust the Public Federal Ministry that these contained charges for Article 149 and missing data was listed for other crimes in this complaint. A similar situation happened with public civil actions; when information on the initial complaint was missing the team trusted the list provided by the Public Ministry of Labor and indicated that the complaint was for Article 149. The team matched this court data with labor inspection reports for each case, when available. Inspection reports provide robust detail about the situation at the jobsite itself. Labor inspection reports are frequently the basis for complaints made in court. The team was able to match court data to an inspection report for 67.6 percent of all cases. More detail on data entry procedures is available in Appendix A. 4.1.1 Sample There are 1,764 cases in my analytic dataset. Roughly 83 percent (N=1,464) are cases processed in the criminal court system, and nearly 25 percent (N=432) are cases processed in the labor court system. About eight percent (N=132) of cases were processed in both the criminal and labor (civil) systems. Cases brought to the labor court system that immediately settled out of court, such as through a TAC (Conduct Adjustment Term), are not included in the analytic sample. 63 About 75 percent of criminal actions were preceded by a labor inspector?s investigation relative to just 50 percent of public civil actions. This difference is likely due to the fact that the Public Ministry of Labor institutes separate investigative procedures to measure forced labor exploitation that do not rely on labor inspectors. 4.2 Qualitative data I conducted 28 interviews with six groups of key informants: federal judges, federal prosecutors, labor judges, labor prosecutors, labor inspectors, and federal police (see Table 3). Interviews took place via Zoom, Google Meetings, and WhatsApp video and lasted an average of 54 minutes. Table 3. Summary of recruited participants Totals Contacted Denied Completed Response rate Federal judge 8 1 6 75% Federal prosecutor 11 0 6 55% Labor judge 2 0 2 100% Labor prosecutor 5 0 4 80% Labor inspectors 8 1 7 88% Federal police 3 0 3 100% I had a high response rate for each group, ranging from 55-100 percent. Only two people declined to participate. Both decliners (a federal judge and a labor prosecutor) indicated that they had a bad experience with researchers in the past and no longer were willing to participate in research studies. The federal judge indicated that he agreed to participate in a study and found out that it was a covert political attack and the requesters used information he provided to make threats against him and his family. The labor inspector who declined did not provide detail. 64 4.2.1 Recruitment I used the administrative data to identify prosecutors and judges who have experience with forced labor exploitation cases. My point of contact at the clinic at UFMG, Dr. Carlos Haddad, then emailed these individuals and invited them to participate in the study. He emailed around five individuals per week beginning the last week of February 2020. I followed up if a participant responded to schedule the interview. I also followed up after one week if they did not respond at all to ask if they were interested in participating. Since Dr. Haddad works in the federal court system, it was more difficult to identify the contact information for the judges and prosecutors in the labor court system. To improve recruitment of participants from the labor court system, we asked labor court participants to refer others to the study. To recruit labor inspectors, Dr. Haddad emailed or messaged via WhatsApp the coordinator of the national mobile inspection group and each of four existing state inspection groups in the first week of June. I then asked these individuals to participate in an interview and to refer other interested labor inspectors to participate. To recruit federal police officers who have been involved in investigation of forced labor exploitation cases, Dr. Haddad emailed his two contacts at federal police headquarters. Again, we asked these individuals to participate and to refer any other police with experience on forced labor exploitation cases. This resulted in three interviews with police officers. These officers were unable to refer any additional colleagues as they did not know anyone else with experience investigating forced labor exploitation cases. 65 4.2.2 Interview topics Interviews with judges and prosecutors focused on how respondents make decisions about viability and culpability of cases, how they characterize different elements of forced labor exploitation, and what kinds of evidence prosecutors and judges think are needed to prove different elements of forced labor exploitation in the labor and criminal court systems. Interviews also probed on what judges and prosecutors perceive to be the importance of the reports provided by labor inspectors, when they think a police investigation is needed, and what instruments they think are most effective to prevent and punish trafficking. Interviews with police and labor inspectors focused on how they characterize different elements of forced labor exploitation and what kinds of evidence they would document to prove the existence of those elements. I also asked these groups who they think should be involved in investigation teams. For all groups, interviews probed on the kinds of training respondents receive related to forced labor exploitation and the types of training they think would be ideal. Interviews also probed about what kinds of things could be done to improve the investigation or prosecution of forced labor exploitation cases. Interview protocols are available in Appendix B. 4.3 Analytic methods 4.3.1 Objective One: Describe characteristics of forced labor exploitation cases forwarded for processing in criminal and civil courts The goal of objective one is to provide detailed information about the 66 characteristics of forced labor exploitation cases, including detailing the ways employers induce, recruit, transport, and obtain workers as well as the forms of force, fraud, and coercion that employers use to compel forced work. I achieve this objective using a combination of qualitative and quantitative analyses. During interviews with all stakeholder groups, I asked respondents to describe the characteristics of forced labor exploitation cases they had been involved in. I also asked respondents to specifically define degrading conditions and an exhausting workday, as these are the elements of Article 149 that are open to the most subjective interpretation. To answer research questions in Objective One, I rely primarily of deductive analysis based on codes I established a-priori. Specifically, I developed nodes for each element of Article 149 and coded each time these elements were discussed in the interviews: debt servitude, degrading conditions, forced labor, exhausting workday, restriction of freedom or transportation, violence, and withholding documents. While coding interviews using this pre-established framework, I also used an inductive approach of open and axial coding to allow new themes to emerge. A common emerging theme, for example, was the level of necessity for each different elements of Article 149. I developed a coding scheme to identify whether respondents said the element was legally required to constitute forced labor exploitation, whether the element was not required but a case almost certainly would not move forward without it, and elements that, while not required, if present the case would be much more likely to result in a guilty conviction in either court system. I also conducted latent class analysis (LCA) using the administrative data the 67 clinic developed to identify typologies of forced labor exploitation cases. LCA models, estimated through structural equation modeling frameworks (Warren et al., 2015), examine relationships in data by identifying a set of mutually exclusive, unobserved subgroups that account for the distribution of cases occurring within a cross-tabulation of categorical variables. As a result, LCA models allow me to examine patterns of characteristics that commonly group together among forced labor exploitation cases. I used LCA to identify subgroups of forced labor exploitation cases based on a series of observed and measured characteristics and assign cases to one of these subgroups. I first identified latent subgroups based on categorical variables that measure the different characteristics of cases as aligned with the legal definition of forced labor exploitation (Article 149) and the ILO framework for forced labor exploitation which includes three key features of forced labor exploitation: provision of work or service, use of threat or force to compel work, and involuntariness of work. I also used findings from the qualitative analysis to inform the variables that went into the model. For example, in the interviews, participants brought up the presence of weapons and recruitment fraud as important elements of forced labor exploitation that are not specified in the penal code. Characteristics of the provision of work include arduous working days and degrading work conditions. Factors related to use of threat or force to compel work include debt servitude and fraudulent recruitment. Finally, factors related to involuntariness include restriction of movement, confiscating documents, manifest surveillance, and physical violence and/or use of weapons. 68 LCA assumes homogeneity conditional on subgroup membership (Warren et al., 2015). For this analysis, I assumed that members of the same latent subgroup share observed patterns among observed variables entered into the model (Eggleston, Laub, and Sampson, 2004). I ran the LCA model for one to five subgroups. I then compared measures of model fit to determine the most appropriate number of subgroups, including the likelihood ratio chi-square statistic, the Akaike information criterion (AIC), the Bayesian information criterion (BIC), and the entropy statistic; recent research finds that the BIC may be a superior measure of model fit (Yang, 2006) . I also used the Lo-Mendell-Rubin Adjusted Likelihood Ratio Test to compare models with different numbers of classes, indicating a significant difference between n subgroups and n-1 subgroups. Given that these criteria typically do not point to a single best model (number of subgroups), I also examined the models qualitatively to determine the final solution. 4.3.2 Objective Two: Examine factors that predict membership in each forced labor exploitation profile The goal of Objective Two is to identify the factors that are associated with membership in each of the typologies of forced labor exploitation identified in Objective One. Specifically, I asses the factors that influence the type of forced labor exploitation such as economic sector of the work, co-occurring criminal activities, and size of the operation. During interviews, I asked respondents questions about the economic sectors associated with the forced labor exploitation they have been involved in, the geographic areas where these cases take place, and the crimes that most often occur 69 alongside forced labor exploitation. I also asked about the structure of forced labor exploitation operations including the hierarchy of employers and the number and characteristics of exploited workers. My analysis to respond to research questions in Objective Two started with deductive coding using a pre-established coding scheme for each of these elements: co-occurring crimes, location, economic sector, and victim characteristics. The bulk of the analysis, however, was inductive, using open and axial coding to allow new themes to emerge. For example, when asking about the economic sector of cases, nearly all respondents categorized sectors as either urban or rural. From there, I probed further to get a classification of each specific work activity as either rural or urban. I provide more details about emergent themes in the results section that follows. I also examined factors associated with the typologies of forced labor exploitation cases identified in objective one by including covariates in the LCA model to identify which variables significantly predict subgroup membership. I did this using the bias-adjusted three-step approach developed and refined by Vermunt (2010), Bakk, Tekle, and Vermunt (2013), and Vermunt and Magidson (2020). This three-step approach is preferred over a one-step approach in which the latent class model and latent class regression model are joined together using a maximum likelihood estimator for several reasons. First, cases have some probability of belonging to all identified classes, but in the one-step model, it is assumed that cases belong only to the class for which they have the highest posterior probability of membership. In the one-step method, there is no adjustment for this inherent measurement error. Second, the latent class regression model can affect latent class 70 formation when they are simultaneously estimated. This means the latent class variable generated is based not only on the indicator variables, but also on the covariates, which can cause the latent class variable to lose or, at the very least, change meaning. Each time a covariate is added to the regression model, the latent class model needs to be re-estimated (Vermunt, 2010; Bakk, Tekle, and Vermunt, 2013; Vermunt and Magidson, 2020). The one-step approach also raises additional issues such as whether the influence of covariates on the latent class model should be considered when deciding on the appropriate number of latent classes. The inclusion of covariates simultaneously with the estimation of the latent class variable also goes against the accepted logic of specifying a classification model prior to introducing covariates (Vermunt, 2010). To overcome the limitations of the one-step approach, I use the three-step approach developed by Vermunt and colleagues, summarized in Figure 6. Figure 6. Summary of three-step approach Relate latent classification scores saved in step two to Assign cases to latent covariates of interest, classes (using posterior correcting for probabilities) classification error Identify the best fitting latent class model using goodness of fit indicators I first estimated LCA models using only latent class indicator variables, as described in objective one, above. In the second step, I created the most likely class variable, ??, for each case using latent class posterior probabilities that are calculated 71 in the first step. For each case, ?? is the class, c, for which ??(?? = ??|??) is largest, meaning it is the most likely class to which the case belongs. In this formula, U represents the latent class indicators (degrading conditions, exhausting work day, debt servitude, restricting transport/movement, withholding documents, recruitment fraud, guards/surveillance, and weapon), and C is the latent class variable being estimated (type of forced labor exploitation) (Nylund-Gibson, Grimm, and Masyn, 2019). The equation for the uncertainty of the predicted class, N, is: 1 ????1,??2 = ??(?? = ??2|?? = ??1) = ? ??(????= = ??2|????) ????1 ????=??1 In this equation ????1 is the number of cases assigned to class 1 based on the most likely class variable N. ???? is the most likely class variable for the i-th observation. ???? is the ?true? latent class variable for the i-th case, and Ui is the indicator variables for the i-th case. The probability that the true typology of forced labor exploitation for a given case will be a specific class given the indicator variables for that case ??(????= = ??2|????), and the measurement error associated with that calculation is computed from the LCA model in step one. Finally, the most likely class variable, N, is regressed on predictor variables, accounting for misclassification (i.e., measurement error) in step two, with the likelihood of misclassification fixed at the probabilities obtained in step two. In this third step, the covariates are introduced but the measurement relationship between the true forced labor exploitation typology, C, and the most likely typology calculated in the LCA model, N, are fixed to account 72 for misclassification error (Vermunt, 2010; Nylund-Gibson, Grimm, and Masyn, 2019; Bakk, Tekle, and Vermunt, 2013). 4.3.3 Objective Three: Analyze how case profiles are associated with sentencing outcomes The goal of Objective Three is to examine how different types of forced labor exploitation cases are associated with outcomes in the labor and criminal court systems. To achieve this objective, I answer two primary research questions. First, I describe the outcomes of cases prosecuted with forced labor exploitation charges using descriptive statistics from the administrative data and analysis of qualitative data. In the interviews, labor inspectors and police noted that they do not get information about sentencing outcomes for their cases, so I relied on information from prosecutors and judges for this objective. I asked prosecutors and judges what makes it more likely that a case will receive a conviction, including the characteristics of the case and the types of evidence needed to prove each characteristic. Most respondents from the criminal system reported that these cases rarely receive convictions; in these situations, I probed on why the respondents thought that was the case. I also asked respondents what they though the most effective instruments for punishing and preventing forced labor exploitation were. For this objective, I relied almost entirely on inductive analysis. The only pre-established codes I used were a large catch all for prevention and punishment instruments and another for barriers to conviction or investigation. 73 To answer the second research question in this objective, using the administrative data, I examine how latent classes of forced labor exploitation cases are associated with civil and criminal court outcomes. There are a variety of ways to include a distal outcome in a latent class model, but current literature points to two options that are appropriate for reducing measurement error: an ?automatic? maximum likelihood three-step approach and a manual maximum likelihood three- step approach (Vermunt, 2010; Nylund-Gibson, Grimm, and Masyn, 2019; Bakk, Tekle, and Vermunt, 2013; Vermunt, and Magidson 2020). I provide results using both. Each of these methods assess the relationship between latent classes and the distal outcome by calculating the mean of the outcome for each class and then looking for significant differences in the distal mean for each class. This process involves three steps, which are similar to the steps for identifying predictors of latent classes described above. The first step in each approach is to identify the latent class model; the second and third steps vary slightly for each approach and are described below. The logic behind the automatic and manual maximum likelihood (ML) three- step approach is identical and similar to the method implemented (Nylund-Gibson, Grimm, and Masyn, 2019). First, I identify the latent class model and save posterior probabilities and most likely class assignment. In step two, I direct Mplus to compute conditional probabilities for the most likely class assignments given true latent class membership. In practice, these are estimated classification errors for most likely class assignment. Then in step three, the most likely class assignment is used as an 74 indicator of the latent class variable, but the classification errors are fixed at the values computed in step two. To implement the automatic ML approach, I use the auxiliary option in MPlus to indicate that there is a distal outcome I would like to include in a three-step model. Using this option, MPlus automatically corrects for measurement error inherent in the predicted latent class when calculating the distal outcome mean. The drawback of this method is that it cannot simultaneously control for additional covariates and calculate the mean of the distal outcome, so the model does not control for covariates (Nylund- Gibson, Grimm, and Masyn, 2019). In the manual ML three-step, I specify the fixed classification errors in the model command, which allows me to also include covariates. These approaches are somewhat different than the traditional approach of assigning the most likely class to each case based on the highest marginal probability of class membership and then running a traditional regression model. I also run this traditional model for comparison. For civil cases, I was interested in three outcomes: 1. A nominal variable indicating the outcome of the case (conviction, agreement, or neither). 2. Whether the case was added to the Dirty List. 3. Whether a TAC was signed. 4. A continuous variable indicating the amount of collective moral damages ordered. 5. A continuous variable indicating the amount of individual moral damages ordered. However, there are only 211 cases processed in the labor court system with non-missing data for the latent class indicators (missing 221 cases). Thus, there is not a large enough sample to assess the first three outcomes. Thus, for civil case 75 outcomes I provide a simple crosstab to show variation in outcomes by latent class. For the fourth and fifth outcomes listed, I use the approaches described above. For criminal cases I examine five outcomes: 1. A dichotomous variable indicating whether the case received a conviction. 2. A dichotomous variable indicating whether the case was acquitted. 3. A continuous variable indicating the length of the case from investigation to sentencing. 4. A continuous variable indicating the sentence length in months for guilty convictions. There are 1,075 criminal cases with non-missing data for the latent classes. 76 Chapter 5: Results ?We are faced with a visceral reality in Brazil: poverty derived from the inaction of the state?The suffering of the worker is the tip of the iceberg of a sick society that has forgotten or never even known values of fraternity, of the right to a dignified life, the de-concentration of wealth? - LA008 This chapter presents results from the analyses and responses to the research questions outlined in Chapter 4. This chapter begins with descriptive information about qualitative interviews. I then review key variables relevant to my analysis, beginning with what I found in the qualitative data relating to those variables, then describing the variables from the administrative data. 5.1 Qualitative data As shown in Table 4, 18 of the 28 respondents were male: this includes four of six federal judges, all three federal police, four of six federal prosecutors, two out of seven labor inspectors, both labor judges, and three out of four labor prosecutors. Table 4. Respondent sex and role Respondent Type Female Male Total Federal judge 2 4 6 Federal police 3 3 Federal prosecutor 2 4 6 Labor inspector 5 2 7 Labor judge 2 2 Labor prosecutor 1 3 4 Total 10 18 28 I also recorded the length of time the respondents served in their career, shown in Table 5. Three respondents had worked for less than five years, seven for 5- 10 years, seven for 10-15 years, and 11 for 15 years or more. Importantly, many 77 respondents noted that forced labor exploitation cases were rare, so length of tenure does not directly correlate to more experience with forced labor exploitation cases. Table 5. Respondent tenure Respondent Type <10 years 5->10 years 10-<15 years 15+ years Total Federal judge 1 1 1 3 6 Federal police 1 2 3 Federal prosecutor 1 3 2 6 Labor inspector 1 4 1 1 7 Labor judge 2 2 Labor prosecutor 2 1 1 4 Total 3 7 7 11 28 Detailed information about each respondent is available in Appendix C, Table 1. 5.2 Administrative data In this section, I provide descriptive information about the cases in the civil and criminal systems. I exclude 126 cases that had multiple procedural outcomes and thus appeared twice or more in the data, keeping only the first instance of these cases to avoid counting the same characteristic multiple times for the same case, leaving a sample of 1,638 cases. A case could have multiple criminal outcomes if more than one defendant were charged for the same incident. Many of the quantitative variables described in the descriptive summary that immediately follows are highly correlated and thus not all can be used for the analysis (e.g., State and City). Following the descriptive section, I provide a summary of the variables used in each analysis. I then present findings from my analyses by objective, presenting qualitative findings first, followed by quantitative. 78 5.3 Descriptive information 5.3.1 Economic sector In this section, I first describe qualitative findings related to the economic sectors in which forced labor exploitation takes place and then provide summary statistics for variables related to economic sector from the administrative data. Qualitative findings I asked participants about the industries where forced labor exploitation takes place. These discussions ended up being intertwined with discussions about urbanicity. Every participant distinguished between rural and urban forced labor exploitation. One labor inspector (LA002) summarized the economic activities that commonly involve forced labor exploitation in urban and rural areas: Regarding an economic activity, [forced labor exploitation] is still quite prevalent in the rural area. There is even greater volume in activities such as cattle, deforestation, coal, some crops such as coffee, carnauba, as well as some activities that are far from large city centers? But it is also very common to find [forced labor exploitation] in cities, in urban centers. In civil construction and in sewing workshops it is very common. It also appears in the countryside, with mining. These are some activities that, due to the type of labor needed, end up leading to greater vulnerability of the workers. Respondents mentioned rural forced labor exploitation cases most often. As shown in Table 6, 18 out of 28 respondents indicated having experience with urban cases, while nearly all (26) indicated having experience with rural cases. The two respondents who indicated not having experience with a rural case were a federal judge and a federal prosecutor. Some of the industries in rural areas mentioned were cattle farming, charcoal production, agriculture (e.g., soybeans, coffee, sugar cane), and mining. 79 Table 6. Respondent experience with urban and rural cases Respondent Experience with Urban Experience with Rural Type Cases Cases Federal judge 3 5 Federal police 1 3 Federal prosecutor 2 5 Labor inspector 7 7 Labor judge 1 2 Labor prosecutor 3 4 Total 18 26 Many participants who mentioned experience with urban cases clarified that most of the cases they had were in rural sectors; for example, one federal judge (FJ003) noted, ?I have received complaints from cases in the urban area? but most of the cases are related to agriculture.? Another judge (FJ004) noted that she only had rural cases throughout her entire career: ?Only rural. I never took any urban slave labor.? This judge further specified that most cases took place ??on livestock and agriculture farms. Most of the time it was on farms that mix the two activities.? Another federal prosecutor noted: More frequently in [the state of] Amazonas is rural slave labor, that is, there are farms, boats, or small distant communities that are exploited by local businessmen?but the focus of inspections has been disproportionately focused on these rural cases even though our state has one of the largest industrial centers in Brazil. -FP003 As one labor judge (JT001) noted, ?The three cases I had here [in this city in the state of Minas Gerais] were coffee producers and in [another city in the state of Minas Gerais] I had one case with quartz extraction.? Participants also mentioned that sometimes employers engaged in forced labor exploitation to clear land so they could begin farming. As one federal judge said: 80 To better explain the region, here there are large farms for planting soybeans and cotton, mainly. So, several cases I took had to do with workers called ?stick pickers?, I don't know if you are aware of that [term]. Because the region here has native undergrowth, these workers are hired to clear the land of this undergrowth. Land cleaning cases, not related to the main activity of the company, but in relation to the isolated part of the land and the contracting of a specific cleaning task. -FJ003 A labor inspector (LA003) described the same type of work: ?Here in [the state of] Rond?nia, most of the cases that we find with slave labor are in the activity of deforestation, usually in the opening of the passage of a cattle farm.? Participants also discussed some of the unique challenges inherent to working in rural and urban areas. In rural areas, for example, investigations are often more difficult because the owner of the operation is rarely present. Instead, a middleman may be there. In many instances, no one present even knows the owner. A federal judge provided an example similar to those given by many participants: We had a situation in [Northern State], where there was a report of the crime of illegally extracting precious stones, the police arrived at the place, these people who were there, were arrested as if they were extracting illegal stones. When they arrived at the prison, they discovered those people were not responsible and were just workers who were in slave labor. We called and talked to the prosecutor?we were able to talk to everyone and get their testimonies there at the last minute and then we managed to rescue them and they left for their city, so we never had contact again. -FJ006 Labor inspectors, however, tended to agree that rural forced labor exploitation was easier to investigate. One labor inspector (LA004) noted, ?Rural forced labor exploitation is still the majority. There are cases in civil construction and now also domestic work, but as it is more sensitive and difficult for us to investigate, we have done only two rescues of domestic workers in the state.? This same labor inspector 81 also said that there are some rural areas that are easy to get to, so their team routinely investigates in those areas: ?For example, here in [the state of] Bahia we have the issue of cocoa production, which is an activity that we consider very accessible in the southern region of the state, and that is why it is an activity that we constantly monitor regardless of complaints.? Another labor inspector mentioned that both rural and urban investigations have their own challenges and dangers: On the rural side, we always need to have a larger team, because the risk to the inspection team due to violence and aggression is greater. But sometimes, here in the urban environment, we happen to have certain activities that are also socially precarious areas, an environment with paramilitary organizations that we call militias in Brazil, are areas that are dominated by drug dealers, and they do not know that we are investigating forced labor exploitation, they may think that we are investigating their business, which is drug trafficking or selling properties in reserved urban areas known here as grilagem, so they are also very sensitive to possible violence. -LA006 Quantitative findings I analyzed the primary economic activity for civil and criminal cases. Of these cases, 491 were missing data on economic sector. Of the 491 cases missing data on economic sector, 479 (97.56 percent) were not investigated by labor inspectors and thus did not have a linked inspection report, demonstrating the importance of the inspection reports in describing the facts of the case (said differently, demonstrates the insufficiency of court records alone to describe details of cases). Figure 7 shows the economic sector of cases. The largest share of cases were in livestock or fishing (26.62 percent), followed by agricultural activities (18.19 percent) which included harvesting sugar and rubber as well as other products like coffee, grains, soy, and other produce. 82 Figure 7. Economic sector of forced labor exploitation cases, N=1,147 livestock, fishing 436 agriculture, sugar, rubber 298 charcoal 125 construction 85 forestry 77 manufacturing 38 mining/extraction 36 urban services-? 27 other 25 Given that all interview respondents conceptualized economic sectors by whether they were urban or rural, I also categorized economic sector by whether the industry took place in rural areas. Non-rural (i.e., urban and suburban) activities included cases that took place in the transportation/delivery, hospitality, civil construction, energy, parking, domestic work, and manufacturing industries. All other industries were classified as rural (see Appendix A for more detail). Overall, about 85 percent of cases took place in a rural industry. As shown in Figure 8, however, the proportion of cases taking place in the rural sector has declined each year, from 100 percent in 2004-2006 to just 68 percent in 2018 and 73 percent in 2019. This pattern parallels a pattern in the overall economy of Brazil, which has been shifting toward more urban activities. 83 Figure 8. Forced labor exploitation identified in rural sectors has declined over time 120% 100% 99% 100% 100% 93% 100% 92% 83%95% 80% 90% 94% 95% 88% 81% 75% 73% 60% 68% 40% 20% 0% 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 5.3.2 Geographic Location Qualitative findings Interview respondents were distributed in a diverse range of geographic areas, with eight respondents in the Center-West region, five in the North, eight in the Northeast, and five in the Southeast (see Table 7). Notably, many participants had worked in a variety of locations throughout their careers; Table 7 shows only their location as of the interview date. Two participants, one labor inspector, and one labor prosecutor were members of teams that worked on cases throughout the country. Table 7. Respondents' current geographic location Respondent Type Center-West North Northeast Nationwide South East Total Federal judge 2 1 2 1 6 Federal police 2 1 3 Federal prosecutor 2 2 2 6 Labor inspector 1 1 2 1 2 7 Labor judge 1 1 2 Labor prosecutor 1 1 1 1 4 Total 8 5 8 2 5 28 84 A lot of the discussion around geography in interviews focused on the different economic sectors that commonly used forced labor exploitation in each area, as discussed in the economic sector section above. However, another common theme in discussions about geography were the characteristics of victims in each region. Several respondents noted the relationship between vulnerability of workers and the economic sectors that usually include forced labor exploitation; for example, a federal prosecutor (FP006) mentioned that most cases take place ??in the rural sector. The less qualified the workforce, the easier it is to exploit.? Another federal prosecutor elaborated on this concept, describing that the pervasiveness of poverty combined with a lack of job opportunities in some rural areas of the country leave forced labor exploitation as the only viable way for workers to earn a living: In the case of degrading work, there is the problem that workers do not recognize themselves as victims because they are living in conditions that they are already used to. There is a lot of poverty in the South of Cear?, Maranh?o and Piau?, in fact many times the victims thank God that they can work. And the wages are not always necessarily low, for cutting wood the wages are reasonably high, they earn up to R $ 3,000.00 to R $ 4,000.00 [USD$ 600- USD$ 800] per month, so they accept these conditions. -FP002 Nine participants mentioned forced labor exploitation the Amazon. The Amazon rainforest extends over eight countries and nine states in Brazil. Forced labor exploitation cases in the Amazon were described as some of the worst abuses of human rights as well as the least likely to be investigated. A federal prosecutor explained that in the Amazon, there is a provision that allows an individual to use an area for farming if the individual can prove they have occupied the area for a certain number of years. In order to occupy the area, however, the individuals first must clear out the trees and other forestation: 85 It may be ruled, as a provisional measure, for you to legally use land in the Amazon. This provisional measure encourages land grabbing, because for you to get regularized in the Amazon, you have to prove that you occupied the area and the occupation is done by deforestation, so the first absurdity of slave labor is that [exploited workers] carry out deforestation for illegal land grabbing. -FP001 When discussing forced labor exploitation in urban areas, ten respondents brought up the exploitation of immigrants, particularly from Haiti, Venezuela, Bolivia, and China. One federal labor inspector, for example, said: A pasteleria is a type of snack bar. So, you have a snack production activity where you can have a Chinese immigrant there, living in that place, working every day, from dawn until the last hours of the night, and then we identify it as an exhaustive journey--the person?s basically living in a prison. This worker is vulnerable because he doesn?t speak the language and doesn?t know the situation of being exploited by the owner of this establishment all the time, without communication, without any type of complaint. -LA006 A handful of respondents also indicated that immigrants were victimized in rural areas of the country as well. One federal prosecutor described the connection between economic sector, geography, and victim-level risk factors like race and nationality: The connection between slave labor in the Amazon and slave labor in Mato Grosso do Sul is immigrants. We have difficulty investigating these crimes because immigrants have so many vulnerabilities. In the Amazon as a whole, you have a lot [of immigrant workers], especially in [the state of] Par?, coming from [the state of] Maranh?o. You also have racial vulnerabilities because most of these people are Quilombolas from the interior of [the state of] Maranh?o, who end up moving to the interior of Par? and Mato Grosso do Sul. In Mato Grosso do Sul, there is also something known as a Paraguayan job, which is a slave job both linked to farms, cutting eucalyptus, and also linked to sex slave labor. -FP001 State Information about the state where each case was identified is available for 1,638 unique cases. As shown in Figure 9, the majority of cases were identified in Para (20.57 percent), Mato Grosso (11.36 percent), and Minas Gerais (9.77 percent). 86 Figure 9. Number of cases identified in each state and cumulative percentage of cases, N=1,638 City Data on the city where each case was identified is available for 1,638 unique cases. The top 20 cities account for roughly 17 percent of all identified cases. The top city, Sao Paolo, has 48 cases, representing 43 percent of cases in the state of Sao Paolo. Nine of the top 20 cities account for 41 percent of cases in Para, which is the state with the highest identified number of forced labor exploitation cases. Appendix C, Table 2 provides detailed information on the cities where forced labor exploitation is identified. 87 Region Brazil is divided into five geographic regions, described below. A map depicting these regions is available in Appendix A. ? North: Acre, Amap?, Amazonas, Par?, Rond?nia, Roraima, Tocantins ? Northeast: Alagoas, Bahia, Cear?, Maranh?o, Para?ba, Pernambuco, Piau?, Rio Grande do Norte, Sergipe ? Center-West: Goi?s, Mato Grosso, Mato Grosso do Sul, Distrito Federal (Federal District) ? Southeast: Esp?rito Santo, Minas Gerais, Rio de Janeiro, S?o Paulo ? South: Paran?, Rio Grande do Sul, Santa Catarina As shown in Figure 10, the highest share of cases was identified in the Northern region (33 percent), followed by the Southeastern (21 percent), Center-West (20 percent), Northeastern (17 percent), and Southern (9 percent) regions. Figure 10. Proportion of force labor exploitation cases identified in each region, N=1,638 South 9% Center-West 20% Southeast 21% Northeast 17% North 33% 88 5.3.3 Number of workers Number of workers reached Data on the number of workers reached at each job site is available for 1,091 cases. Of the 577 with missing values for number of workers reached, 97.4 percent (N=562) are cases that are not linked to a labor inspector?s report, again demonstrating the utility of labor inspectors? reports in providing details about the cases. In total across 1,091 cases, labor inspectors found 75,971 workers at the job sites, or an average of 71.6 workers per job site. Number of workers rescued When labor inspectors determine that a worker is subject to forced labor exploitation at a job site, they remove the worker from that situation. As I learned in the qualitative interviews (described below), it is common for employers to have a set of staff who are employed formally and legally as a cover for staff who are subject to forced labor exploitation. Thus, the number of workers rescued from situations of forced labor exploitation is typically only a share of the workers reached. Data on number of workers rescued is available for 968 cases. Again, 562 missing cases (83.9 percent) were not investigated by labor inspectors. In total across 968 job sites, 20,174 workers were rescued, an average of 20.8 workers per job site. On average, about 74 percent of workers in inspected sites where forced labor exploitation was found were removed or rescued from the situation. 5.3.4 Number of workers rescued who received unemployment As discussed in Chapter 3, Law 10.608/02 allows rescued workers leaving situations of forced labor exploitation, regardless of citizenship or immigration status, to receive unemployment payments. In line with this regulation, more than 90 percent 89 of workers rescued each year received unemployment. As shown in Table 8, across the cases with non-missing data, 75,791 workers were reached, and 20,174 workers were rescued. Across the 866 cases with non-missing data about receipt of unemployment, a total of 15,213 rescued workers received unemployment, or an average of 17.6 workers per job site. Table 8. Number of workers reached, rescued, and receiving unemployment N (non-missing) Total across all cases Min Max Average Number reached 1061 75,971 1 3971 71.6 Number rescued 968 20,174 1 1003 20.84 Number receiving unemployment 866 15,213 1 922 17.57 5.3.5 Other crimes that co-occur with forced labor exploitation Qualitative findings I asked participants about the other types of crimes that tend to be committed alongside forced labor exploitation. They mentioned crimes against the environment, worker enticement, human trafficking, and the use of violence, weapons, and private imprisonment. Notably, respondents also mentioned a variety of white-collar crimes that facilitate forced labor exploitation such as document forgery, use of shell companies, and falsifying work documents (see Table 9). 90 Table 9. Other crimes committed alongside forced labor exploitation, interview data Human Enticement Environmental Child White Physical Trafficking of Workers Crimes Labor Collar Violence Number of respondents 15 12 8 5 7 6 Federal judges 1 3 3 2 1 1 Federal Prosecutors 5 4 2 1 3 1 Police Officers 2 1 0 0 1 1 Labor judges 0 0 0 0 1 0 Labor prosecutors 2 3 0 1 0 1 Labor inspectors 4 1 3 1 1 2 Article 207 Enticement of Workers Twelve participants mentioned that domestic movement of workers into other states was a common co-occurring crime with forced labor exploitation. Domestic relocation of workers for fraudulent reasons is a crime under Article 207 of the penal code. As one federal judge (FJ001) described of a case she recently worked on: ?The [forced labor exploitation] case involved internal trafficking, which is taking people from poor states in Brazil and bringing them here [to an urban area in the state of Minas Gerais].? Another federal judge (FJ003) also said that Article 207 often co- occurred with forced labor exploitation and that Article 207 is usually only reported in conjunction with forced labor exploitation: ?Usually here, it is only slave labor that is reported or refers to Article 207 of the Penal Code, which is to entice workers from another location in the national territory.? Article 149a Human Trafficking Fourteen respondents mentioned human trafficking or sex trafficking as crimes that co-occur with forced labor exploitation. One labor inspector (LA004) said, ?The most common [co-occurring crime] is human trafficking. In fact, in 100 percent of the slave labor cases, trafficking also occurs because they are vulnerable people...? As one labor inspector explained, the definition of forced labor exploitation 91 in Brazil was developed based on the UN Palermo Protocol to combat trafficking in persons so there is a lot of overlap between human trafficking and forced labor exploitation: The concept of slave labor in Brazil comes from the Palermo protocol and in our penal code, it was adopted as one of the modalities, reception, transportation, all those concepts of human trafficking, for purposes of labor analogous to slavery. Our concept of slave labor in Brazil is broad, so when we inspect it, we use the concept of analogous slave labor more than that of human trafficking, although both are present in most cases. -LA001 I asked respondents who brought up human trafficking to explain the difference between human trafficking (Article 149a) and forced labor exploitation (Article 149). Respondents unanimously agreed that human trafficking must include relocation of workers away from their home community. One police officer recognized that he did not know the legislative differences but described his conceptualization as follows: You have to move people to characterize human trafficking, I don't know what the legislation is like, but what I see is.... if you hire people promising something that would be far from fulfilled, then you have a crime of enticement and human trafficking?you hire a person, promising them that they will have accommodation with good working conditions, hygiene, and health, you promise her certain wages but beforehand, you already know that you will not fulfill that. So human trafficking is a crime that usually precedes slave labor. Now when you have a situation where workers in the region or the municipality itself go find a crime of isolated slave labor, you will not have the solicitation, you will not have this migration of workers for this other purpose. -PO003 A federal prosecutor (FP001) also discussed the same concept that the police officer above mentioned which is that if the worker willingly goes to the job site, it is forced labor exploitation only, but if they are somehow coerced into traveling for the 92 job, it is human trafficking: ?There is always a fine line between consent and human trafficking, which ends up making the fight more difficult.? Crimes against the environment Eight respondents mentioned crimes against the environment regularly co- occur alongside forced labor exploitation, particularly illegal deforestation or land clearing and illegal mining. One federal judge (FJ006) explained, ?We are focused very much as if slave labor were a one-off thing but it is not only exploiting people but also destroying the environment.? White collar crimes (document fraud, forgery, money laundering) Seven respondents mentioned that white collar crimes co-occur with forced labor exploitation. One federal prosecutor (FP006) mentioned: ?You?re going to have money laundering? money laundering is pretty common.? One way that business owners launder money is through individuals referred to as ?laranjas? whose names are used to open bank accounts through which to funnel money. One federal judge (FJ006) noted, ?Today we have many people who act as laranjas.? A federal prosecutor (FP002) said that they are beginning to work with banks to identify laranjas and other suspicious activity: ?In Brazil 15 years ago, nobody knew about money laundering, now we talk about it, if a bank teller is making a withdrawal of R $ 150,000,000, R $ 200,000.00 and does not notify [the ministry of labor] we say that when this happens it is an indication of money laundering, do not close your eyes.? Other forms of white-collar crime primarily dealt with forging the work cards of exploited workers to either avoid paying legally mandated benefits or to withdraw the worker?s government benefits. A labor judge (JT001) recalled a case that involved insurance fraud and possibly embezzlement: ?In the case that I worked?there was a 93 complaint that the farmer? remained with the worker's citizen card to be able to withdraw unemployment insurance, so there was?a civil action and a criminal action also regarding the crime of embezzlement or something like that.? Child Labor Five respondents mentioned child labor co-occurred with forced labor exploitation. One labor inspector (LA003) said child labor is common in rural work: ?Child labor is usually found on farms. Generally, the son of the worker who is also there and the father takes him to work, it is a systemic situation.? A labor prosecutor recalled a case of child labor and forced labor exploitation in the urban sector. He began by explaining that large corporations often subcontract through several layers of factories to avoid being tied to worker exploitation and then described a case he had in one of the largest cities in the country: Several stores in S?o Paulo, called Zara, contracted to four levels of factories, the fourth was exactly where the Bolivians were, which was a house, transformed into a sewing factory, where they worked about 18 hours a day, they worked, slept, and ate in the same environment, children as well? producing a frightening number of pieces at prices that we don?t believe, but pants that are sold at R $ 100.00, R $ 200.00, R $ 300.00 being produced at R $ 1.00 or R $ 2.00. -LP004 Homicide, torture, and illegal use of weapons Five respondents said that presence of weapons and physical violence, including crimes as serious as torture and homicide, co-occurred with forced labor exploitation. One federal prosecutor summarized the different crimes that other respondents mentioned: Well, slave labor is a major social problem, so it usually includes physical violence, some situations of torture, some situations of a crime provided for in our legislation which is the crime of soliciting workers to work in another part of the national territory, often includes the possession of weapons without 94 authorization, so they are actually crimes related to the use of violence and the solicitation of workers who are in situations of vulnerability. -FP007 Two participants also mentioned situations where labor inspectors were murdered during their investigation, but most respondents noted that homicide was rare but happened occasionally; for example, one federal prosecutor (FP006) said, ?in Par?, threats of homicides are common. The threat is often acquitted but there are homicide cases as well.? Quantitative findings I examined the other crimes that labor inspectors identified alongside forced labor exploitation. Only cases which were inspected by labor inspectors have valid data for these variables. As shown in Table 10, the most common crime that inspectors identified was a violation of Article 203, which is the use of force or fraud to violate a worker?s rights that are guaranteed by labor laws. The next most common co-occurring crimes were a violation of Article 132, putting someone?s health or life in imminent danger, and a violation of Article 207, fraudulently recruiting workers for a job within the national territory of Brazil; both of these crimes were identified in roughly six percent of cases. About nine percent of cases reference an ?other? type of co-occurring crime. These include offenses such as possession of an unregistered or illegally obtained firearm, bribery in an international business transaction, homicide, prostitution, and others. 95 Table 10. Other crimes identified alongside forced labor exploitation Variable Obs Mean SD Min Max 149a: Human Trafficking 1,072 1.03% 10.08% 0 1 132: Put someone's health or life in imminent danger 1,075 5.77% 23.32% 0 1 203: Violation of workers? rights guaranteed by labor legislation via fraud or violence 1,075 10.98% 31.27% 0 1 206: recruitment fraud-international 1,075 0.09% 3.05% 0 1 207: recruitment fraud-domestic 1,075 5.58% 22.97% 0 1 Other crime 1,072 8.58% 28.02% 0 1 There are few places where administrative data does not align with what interview respondents were the most common co-occurring crimes; namely the interview respondents reported that human trafficking is highly prevalent while the administrative data shows that just one percent of cases involve human trafficking. When I asked a labor inspector (LA004) how they determine whether human trafficking is present in addition to forced labor exploitation, she responded by saying it is not the role of the labor inspector to determine this: ?We put the facts in the report and those who are going to indict the crime? if they understand that there is only one crime, they will only charge for one crime.? Prosecutors may choose not to indict cases on human trafficking if they think they have a strong enough case against forced labor exploitation and/or if they do not have sufficient evidence to prove human trafficking is present. A labor prosecutor commented on the difficulty of distinguishing between the crime of enticement /solicitation of workers (207) and human trafficking (149a): The immigration law of 2016 and 2017 introduced Article 149 A to the penal code, which deals with human trafficking and that with some modalities?but due to a breakdown of the penal code, there is immense confusion about how to frame, if enticement, if trafficking, if the crime of slave labor absorbs the crime of Article 149 A, which is trafficking... in my opinion, making an institutional self-criticism, this is a very serious failure of our judicial public policy. -TP005 96 Also missing from the administrative data are the white-collar crimes mentioned by interview respondents. Some of these crimes that respondents described, such as unemployment fraud, are included in Article 203. It is possible that the ?other crime? category includes other forms of white-collar crimes brought up in the interviews. Another co-occurring crime that interview respondents mentioned that is absent from administrative data are crimes against the environment. One quote from the interviews sheds light on another possible explanation: that when environmental crimes are present, the prosecution focuses on this crime alone rather than as a charge that accompanies forced labor exploitation: The evidence of environmental crime is just easier to collect [than forced labor exploitation]. Also, environmental crimes are defined in our constitution, so it is one of the few types of crime in which Brazil allows legal entities [rather than individuals] to be held accountable. It is not that investigation [of environmental crimes] is easier, but in this context, collecting the evidence is perhaps easier. - FP009 5.3.6 Characteristics of Forced Labor Exploitation In this section, I describe the different elements that characterize forced labor exploitation. First, I provide findings from the interviews. Then, I summarize relevant variables from the administrative data. Qualitative Findings I asked interview participants how they would define or characterize forced labor exploitation, as well as what kinds of physical evidence could be used to establish the existence of each of the different characteristics that participants mentioned. This discussion yielded important information about the types of cases 97 known to practitioners as well as the barriers to establishing proof for certain aspects of forced labor exploitation. General definitions of forced labor exploitation did not vary much across practitioner types or across practitioners in the criminal and labor systems. All 28 participants provided a definition of forced labor exploitation that is in line with Article 149 of the penal code. As one labor prosecutor described, ?We have four characteristics [that can define forced labor exploitation]: exhausting work hours, degrading conditions, forced labor, and debt servitude.? In addition to these elements specified by the legislation, workers described the presence of weapons, physical violence, and restriction of freedom as important methods that employers use to maintain control over workers. Participants also indicated that while typically constituting a small share of their overall caseload, forced labor exploitation ?is a huge social problem?, in their areas and nationwide. Exhausting work When describing exhausting work, nearly all respondents referenced the legal definition. For example, one federal judge (FJ008) said, ?A workday is eight hours, and according to the law two extra hours are allowed, which would be ten hours.? Upon further probing (e.g., ?Is it possible to have exhausting work hours with fewer than ten hours??), however, respondents indicated that what constitutes an exhausting work day depends on the type of work. As one federal prosecutor reported, ?It depends on the nature of the work; even six hours could be exhausting work.? A federal judge (FJ006) expressed a similar sentiment: ?A person can be working inside for ten hours at a desk in the air conditioning and that isn?t exhausting, but he could 98 be working outside for four hours and it would be exhausting.? The same judge also noted that type of work may not matter given the environmental conditions: [The state] where I work, the heat is extreme, around [104 F] on a regular basis. On a day-to-day basis rural workers are not supposed to work between 11 am until around 3 or 4pm because of the heat. So, when you put a person in an exploitive situation who are forced to work 12, 13 hours in an oven like that, that alone is an exhausting workday. -FJ006 Labor inspectors, federal police, and prosecutors indicated that exhausting work was present in almost every case of forced labor exploitation, but, as one criminal prosecutor (FP002) put it, ?Proving the exhausting workday in rural work, in my opinion, is practically impossible.? Another federal prosecutor (FP003) described, ?An exhausting workday is difficult to prove because you need victim testimony, you don?t know where or if the employer noted the workday in their logs, it could not be written down anywhere.? Another federal prosecutor (FP002) similarly stated, ?How will you prove that the person is there for 10 or 11 hours a day? There is no way to prove it easily.? Another federal prosecutor mentioned that victims are often afraid to admit to working too many hours. This prosecutor provided an example of a case where he was able to secure a sentence that recognized an exhausting workday: I had a case where a woman had multiple jobs on the same farm, this worker did not have her own form of transportation. She arrived at work at 4:00am, she probably woke up at 2:00am to leave home at 3:00am. She started with milking the cows, taking care of the cattle, and then they made her do a series of other jobs until 10:00pm when she was brought home?it was the only case that I managed to prove this issue of an exhausting journey, it was a case I remember in detail because it caught my attention. This person is not inside developing software, the person is doing manual labor. -FP001 99 Another federal prosecutor reported that in urban areas, it is easier to get evidence of an exhausting workday because timekeeping is likely to be done via a computer: I think it is easier to obtain this evidence in the urban environment, that you can do some kind of inspection with technology that allows you to prove the time that people are working without relying on testimonial evidence. Depending on testimonial evidence in this type of crime, depends on several people willing to tell the same story while knowing that the state will not be able to protect them later. -FP002 Degrading Conditions As one labor investigator (LA002) described, degrading conditions are the most prevalent characteristic of forced labor exploitation: ?Currently the most common of the modalities is that of degrading conditions. It is what appears most among the complaints, mainly related to the way the worker is housed, the way he is eating, the water, the sanitary conditions, the working condition is what has a lot to do with the degradation of life and the worker's home there.? Most respondents noted that there was not a clear definition of what constitutes degrading conditions, but that it is usually a situation you know immediately upon seeing. Despite lacking a common legal definition, most participants provided nearly identical descriptions, using phrases such as treating a human like an ?object? or ?animal?, or ?lacking human dignity.? As one labor prosecutor reported ?[Degrading conditions] are when the worker is malnourished, when he doesn?t get any kind of protection from occupational hazards. The water is not potable, he sleeps poorly, his shelter is awful. They treat that person like a thing, like an animal.? Another labor prosecutor provided a more detailed description: Degrading conditions are very serious, extreme conditions that result in a lack of hygiene, poor health, and safety concerns. Things like a lack of potable water for these workers to drink, bathe with, or to cook food. These workers have to go to the bathroom outside in the trees. Or they may be 100 exposed to lethal pesticides or other occupational health risks... And many employers take the workers far from their homes, so the workers have a house at the jobsite?the classic canvas shed or mud houses?and there?s no guarantee that that house will not collapse. When you have this systematic violation of basic hygiene and living conditions, we consider that a degrading condition. Respondents also noted that to characterize degrading conditions under the forced labor exploitation statute, Article 149, a combination of serious factors typically needs to be present. As one criminal prosecutor recalled of his prior cases: In most of [my] cases there were a combination of degrading conditions; workers lived in the middle of the bush, often shared living spaces and drinking water with the animals and did not have protective equipment. [In my experience] the cases were usually of combination of degrading conditions. Some respondents also noted that it is hard to determine degrading conditions in rural areas because the standards for quality of life are different in remote areas than in the rest of the country. As one federal judge (FJ006) said, ?I cannot demand from a rural area that they have sewage treatment, because, in fact, neither do the nearby cities.? Debt Servitude Participants also agreed on a general definition for a third component of forced labor exploitation, debt servitude. Respondents noted that debt servitude is a form of control that employers use to keep workers in situations of forced labor exploitation. One federal prosecutor described debt servitude as follows: Control can be physical, but it can also occur through other means, for example, in historical slavery you had armed guards. I have had a few cases with armed guards that were there to keep workers from leaving. More common, though, is debt servitude, where the person takes on debts and the employer tells them they cannot leave until the debts are paid. This is an old 101 system? but it shows that there are physical and abstract forms of keeping control [over workers]. - FP001 Another federal prosecutor (FP009) mentioned that failing to pay workers, even in the absence of formal debt servitude, also characterizes forced labor exploitation: ?Remuneration is an important feature. If you have slave labor on a farm where the person does not receive payment and only has food and stuff, this also characterizes [forced labor exploitation].? Respondents also noted that debt servitude can be one of the easier characteristics of forced labor exploitation to prove because there is typically physical evidence, like a notebook where employers keep track of the debts. One federal prosecutor explained: In the case of the investigation of the lumber company that I participated in, we apprehended the notebook that had the value that people owed and showed the truck system, that they buy [necessities] from the employer, that notebook helps a lot because it is a typical example of slave labor through debt bondage. - FP002 A truck or barter system is a form of debt servitude in which the employer pays workers with some form of money substitute that can only be used at a company owned store to buy items that are incredibly marked up in value. Physical violence and restriction of freedom of movement I also asked participants if physical violence or physically preventing a worker from leaving the jobsite were required to characterize forced labor exploitation. All participants indicated that it was not. For example, one labor prosecutor responded as follows: No, under Brazilian law, there does not need to be physical violence or restriction of freedom to come and go [to constitute forced labor exploitation]. It may be that the worker can come and go freely, but because the workers is not able to find any other job, is having financial difficulty, 102 there is a lack of opportunities in general, it can lead the worker to accept being subjected to an exhaustive and degrading condition. There can be forced labor exploitation without any kind of violence. -TP001 Despite this consensus that physical violence and restriction of movement are not required under law to characterize forced labor exploitation, participants reported that cases were unlikely to receive a criminal conviction in the absence of evidence of violence, weapons, or physically prohibiting workers from leaving the worksite. As one federal prosecutor noted: ?But without a doubt, not having physical violence or restriction of freedom makes it difficult to convince the judge of the crime, which goes against the concept of contemporary slavery. We cannot rely on the original abolitionist laws, we need to see what contemporary slavery would be today, and follow the current legislation that defines it. -FP006 Quantitative findings I assessed information about characteristics of forced labor exploitation using two sets of variables. First, I looked at whether the labor inspector identified each characteristic of forced labor exploitation in the report. As shown in Table 11, there are 1,072-1,075 non-missing cases for each of these characteristics. For each variable, 99 percent (all but one or two) of cases are missing data because they were not investigated by labor inspectors. One characteristic, violence, is a variable I created by combining forced labor, presence of guards or other ostensive surveillance, and physical violence. When analyzing the open-ended variables describing each of these characteristics, I found there were similar patterns and descriptions. For example, ?work activity developed under threat,? ?without the possibility of voluntarily leaving 103 the place where the activities are carried out,? and ?violence? were prevalent in the description of all of the variables. Further, respondents in the interviews (described above) tended to group these characteristics together. As described in Table 11, degrading conditions is the most prevalent characteristic, identified, in 94.4 percent of cases. Debt servitude was identified in 23.3 percent of cases, exhausting workday in 18.4 percent, and other characteristics such as forced labor, restriction of transportation, armed guards, presence of guns or other weapons, withholding personal documents, and recruitment fraud were identified in less than ten percent of cases. Table 11. Characteristics of forced labor exploitation identified in inspection reports Variable Obs Mean Std. Dev. Min Max Degrading conditions 1,075 94.42% 22.97% 0 1 Exhausting workday 1,073 18.36% 38.73% 0 1 Debt servitude 1,074 23.28% 42.28% 0 1 Restrict transport 1,075 7.91% 27.00% 0 1 Withholding documents 1,074 6.61% 24.86% 0 1 Recruitment fraud 1,074 8.38% 27.72% 0 1 Ostensive surveillance 1,067 2.81% 16.54% 0 1 Presence of weapons 1,075 3.72% 18.94% 0 1 The administrative and interview data are, for the most part, well aligned. Both data sources indicate that degrading conditions are the most prevalent characteristic of forced labor exploitation. Although interview respondents indicated that exhausting workday was potentially as prevalent as degrading conditions, that we only see about one in five cases in the administrative data have the presence of an exhausting workday is also in line with what interview respondents said about this characteristic being incredibly difficult to prove. In the same vein, though Debt 104 servitude was not described by interview participants as being too common, we see that about eight percent of cases in administrative data contain debt servitude. This may be because, as interview participants explained, nearly every situation where debt servitude is present is proven via the existence of debt workbooks. 5.3.6 Outcomes and case processing In this section, I first describe the tools that interview respondents nominated as most effective for combating forced labor exploitation. I then provide summary information from the administrative data for each tool. Qualitative data gives information about why the tool us useful while administrative data provides information on the frequency each tool is used. Qualitative findings I asked judges and prosecutors what they thought were the most effective tools to combat forced labor exploitation, either preventively or punitively. Although I did not explicitly ask this question of police officers and labor inspectors, a handful of respondents raised the topic. Respondents? answers fell into a few categories, summarized in Table 12: fines from the labor court system; agreements (TACs) in the labor court system; criminal sentencing; the Dirty List; improved investigation; and non-legal interventions, such as increasing public awareness about forced labor exploitation and providing services to victims so they do not return to similar exploitive situations after their rescue. 105 Table 12. Tools to combat forced labor exploitation mentioned by interview participants Criminal Dirty Non- Civil Action List Fines legal* Prison action TAC Number of respondents 5 11 9 10 3 4 7 Federal judges 2 6 4 2 1 2 Federal Prosecutors 2 3 1 1 Police Officers** 1 1 Labor judges 2 1 1 1 1 2 2 Labor prosecutors 2 4 1 1 2 Labor inspectors** 1 1 * Includes preventive actions like public awareness campaigns and victim services to prevent re- victimization **Participants were not asked a specific question about instruments to combat forced labor exploitation Most participants suggested that a combination of approaches was needed. For example, one labor prosecutor (TP004) said, ?There are three parts investigation, prevention, and victim assistance that are necessary and have to go together, there is no use for inspection only if there is no prevention, there is no use for inspection and prevention if there is no assistance part.? In general, however, respondents described fines imposed either by labor inspectors onsite or during labor court processing as the primary way to combat forced labor exploitation. These fines, respondents said, serve the purpose of reducing the profitability of exploiting workers. As one federal judge explained: The criminal sentence is not a preventive instrument, it will at most, suppress and not prevent?The heavy fines from labor inspectors are a great disincentive to slave labor, because the producer will think that it is not worth disrespecting the rights of workers at the risk of paying a very large fine. And of course, to damage the reputation of the rural producer, and that is a job for the rural producer union?If you were able to associate these companies with slave labor, what would happen? They would lose market abroad and we would have the rural segment itself against these entrepreneurs because the sector's image is damaged. -FJ004 106 Another federal judge had similar views, describing criminal outcomes as serving the purpose of revenge, while the labor court system did more to prevent the crime because of the heavy fines imposed: Let's imagine that [a landowner] is sentenced to 5, 6, or 7 years, what does society gain from this? I would say that often criminal law is an instrument of revenge for society and the result of criminal law is very limited, it does not serve to make social policy, it does not serve to make society's relations more just?I think the administrative law, these Public Ministry TACs, public civil actions, this can produce much better effects for society than the criminal process? [The landowner] has to lose everything he earned and even more. - FJ005 When respondents from the criminal justice system mentioned their preference for labor court outcomes, I asked whether labor court outcomes would be sufficient to prevent forced labor exploitation or whether a criminal outcome was needed. They all replied that cases must also be judged in the criminal system. One federal judge (FJ003) replied, for example, ?It depends on the case but simply improving the working conditions will not simply erase that a crime was committed. So, if the case is only dealt with there [in the labor court], in my opinion is that would not be enough.? Respondents from the labor court system also said that a combination of approaches were needed and often included the criminal justice response as an important component of a multi-pronged approach. One labor court judge, for example, said that labor court penalties only prevent future reoffending if there is a parallel criminal conviction. He also noted that an important component of the legal 107 response in both systems is the general deterrent effect that consequences can have on other landowners: Look, I think that only the work of the Public Ministry of Labor scares people only when there is a criminal offense too, then I think it is enough to restrain this type of attitude. Now, with the indemnity for collective moral damages [in the labor court system], there will be a financial penalty that is also another disincentive for them to continue in that practice. I believe that in the cases in which I worked, that situation of slave labor was resolved, and it also served to restrain other farmers because the news is disseminated and this ends up also affecting other potential perpetrators of this crime. -JT001 Another labor court judge similarly noted that outcomes in the criminal and labor systems, along with awareness campaigns, are all needed to combat forced labor exploitation: First, the most effective instrument in my point of view is the awareness of entrepreneurs or individuals who hire someone to work with the need to comply with the laws, but that is difficult. The most effective instrument we have are heavy convictions in public civil action, conviction in collective moral damage and heavy fine because that's a way of preventing them from continuing to fail to comply with the legislation. The criminal aspect is also important because when the president of a company is convicted of the crime of slave labor, he will answer criminally for that? so these three mechanisms are important to make companies aware that it is necessary to comply with the laws because the workers are performing the service for them, and a human being has to be treated as a human. ? JT002 A labor prosecutor (TP001) also said that criminal sentencing, specifically prison time is needed: ?Prison is necessary because if the subject feels that paying only the fine is sufficient, sometimes the exploitation brings a greater financial benefit than what he will pay.? Dirty List 108 The Dirty List was the instrument brought up by the greatest number of respondents (N=12). Many participants mentioned that the financial consequences of being on the Dirty List has a deterrent effect on employers and landowners who may consider exploiting workers for profit. As one labor prosecutor (TP001) noted, ?The Dirty List is a very good tool because if the person is prohibited from taking a loan, he is prohibited from exporting his products and selling, of course he will think more than once before exploiting a worker.? Respondents also mentioned that the Dirty List can cause long lasting organizational change that benefits workers, as explained by one labor court judge: The Dirty List is very important, because companies start to work more safely, they start to provide a safer work environment for workers, they start to comply more with work safety standards, not only the labor legislation but also safety standards. For you to have an idea, the ministry of labor today has, if I'm not mistaken, there are 36 work safety rules. Some types of activities have specific rules, for example, slaughterhouses, the slaughterhouses have specific rules for him, the worker in the countryside, has specific rules to be fulfilled by the farmers? With this dirty list that exists, companies are more attentive to these rules to prevent them from going on the Dirty List. -TJ002 Some respondents in the criminal system mentioned that the Dirty List was an even better preventive tool than a criminal sentence, such as this federal judge (FJ006): ?[The Dirty List] is incredibly strong in combating slave labor, I would say it is stronger than criminal conviction.? Another federal judge noted the Dirty List is a good tool because the consequences are often more severe than the consequences in a criminal sentence: [The Dirty List] is a stronger instrument even than the employer's condemnation [in criminal court] because often the criminal penalty will not become a high penalty. So, the inclusion in the Dirty List will cause a blemish 109 on his image, he will not get credit for loans, so it is an instrument that in practice, inhibits slave labor more than criminal conviction. -FJ008 This judge also noted a second benefit, brought up by several other participants, which is that the requirements for inclusion on the Dirty List are easier to meet than the requirements for a criminal conviction: The inclusion in the list, it seems to me is easier, because the requirements to include in the list are more flexible than the requirements for configuring the crime. Because inclusion in the list is done by the Public Ministry of Labor, it is not the criminal judge who determines inclusion in the list. -FJ008 Another benefit of the Dirty List that was frequently mentioned was that the Dirty List allowed individual consumers to be informed about their purchases and where they choose to spend money. As one federal prosecutor (FP009) described, ?I think [the Dirty List] is very important for the consumer ? to know the origin of the products they consume.? Non-legal interventions to combat forced labor exploitation Several participants also suggested that non-legal interventions were among the most effective for preventing and intervening in forced labor exploitation. Some specific examples included public awareness campaigns, research studies of the root causes of forced labor exploitation, and services for victims. For example, one labor prosecutor (TP001) discussed the need for prevention efforts targeted at the general public and employers: First, we need to have more resources so that we could go to companies to explain slave labor, hold seminars, seek training for employers, and clarify for the country as a whole, even training in schools, so that people understand what work analogous to slave labor is and how to respect workers. The country needs to develop, but people need to better understand the human condition and gain respect for the human person and workers. - TP001 110 Another labor prosecutor gave a compelling explanation about the types of misinformation that need to be dispelled around exploited labor: Child labor is often culturally accepted, here in the Northeast. If you start talking about child labor, the people will say that I have to leave the children working, rather than stealing and using drugs. So?we want to make people understand that you are not helping the child by giving them an early job, you are causing serious consequences in their future, because they will be functional illiterates, they will continue in that vicious cycle of poverty and they will not evolve because they did not have a simple educational formation? slave labor happens in the same way, one of the characteristics of the worker who is placed in this situation is low education, they are workers that we call marginalized, not in the pejorative sense, but in the sense that they are on the margins of society, and as if they were invisible workers, many of them have no documents, anyway. -TP004 The labor prosecutor quoted above described that larger shifts in societal attitudes are needed to combat forced labor exploitation. A federal police officer responded along the same lines, indicating that educational campaigns are needed but also larger changes that reduce marginalization of vulnerable individuals: Extinguishing a crime is a utopia. It is impossible... What we have to create in my point of view, nationwide? is basically two things: first is to reduce the ignorance on the part of the worker about the legal protection that he is entitled to... Then there are macro issues, the issue of social inequality in the country, it is obvious that when people are looking for subsistence, they are no longer concerned with dignity, they are concerned with a plate of food. - PO001 Another labor prosecutor indicated that public education can also help to combat negative stereotypes about individuals living in poverty: ?many people understand that [forced labor exploitation] doesn?t even exist, that it?s the prosecutor?s office, the ministry and that it doesn?t exist, that 111 workers are there willingly, I?ve heard a lot people that it was useless to put a bed for them because they didn?t know how to sleep in bed, there was no use putting a bathroom because they didn?t even know how to use the bathroom, for you to have an idea of what the society?s enlightenment is, preventive is very important. -TP004 While public awareness campaigns can help to prevent forced labor exploitation, it could also help to address a primary barrier to convictions: lack of witness testimony. Eleven respondents indicated that a barrier in obtaining convictions for their cases was that the victims themselves did not think they had experienced forced labor exploitation. As one federal prosecutor (FP002) explained, ??it is difficult because it involves a cultural issue as well, we are in a very unequal country? there are few people with a lot of money and many people without any money so it is difficult for society to see it as a degrading situation because they are used to it.? Another federal prosecutor (FP001) similarly noted, ?A worker subject to slave labor is not in a situation so different from what he was used to before arriving in that situation, these people even say so themselves. When you place degrading work with the same concept of slave labor in an extremely poor region, there is a rejection of the idea of slave labor.? A labor inspector (LA005) also said, ?Nowadays, the vulnerability of the human being is increasingly so extreme that the working conditions, although very precarious?do not bother this worker in the way as it should, because he came from a much worse situation?he is a slave but he has food, he has a bed to sleep on, so it's OK.? Improved Investigation 112 All respondent types except for judges mentioned the importance of improving investigations in order to combat forced labor exploitation. A police officer (PO002), for example, said that the best way to combat forced labor exploitation is to improve investigations: ?For us [the federal police] it is very clear that the most effective is improve investigations so that it leads to an increase in the convictions.? A federal prosecutor (FP002) mentioned that investigations are done from the perspective of the labor court system because labor inspectors do the bulk of investigations, which makes it hard to establish a sufficient burden of proof for criminal cases. Further, police are not trained to handle forced labor exploitation cases: ?We have this problem that the investigation has always been from the labor point of view. The second problem is the inability of the federal police to handle this type of investigation.? As one labor inspector explained, research into forced labor exploitation can help to improve investigations: [We need more] studies of the supply chain where slave labor exists. For example, slave labor in a charcoal factory-- If we studied the entire chain and saw who and the company that is contracting the company that is using slave labor, we could hold them responsible [at the top of the supply chain]. -LA003 Several participants indicated a desire to more thoroughly investigate the complete supply chain, including companies that drive demand for products made with forced labor exploitation. One federal prosecutor explained: I have no way of criminally getting the conviction of the exporter or whoever is receiving [goods produced with forced labor exploitation], but an investigation that takes into account the necessary bank secrecy breaches, 113 money laundering, and then you have a series of frauds that accompany money laundering too?This could be investigated concurrently to get the whole picture, then you could be able to prove the degrading work. - FP002 A labor inspector described a similar issue: We still lack specific intelligence and tracking work?sometimes we plan an action in a specific region, in a specific economic activity, tracking the entire chain, and the result affect a greater number of workers. But in general, we answer specific complaints, we put out fires, we don't solve the problem. I think that planning and intelligence training is important. -LA003 Quantitative findings Dirty List The Dirty List was created by Ordinance 1.234 passed by the Public Ministry of the Economy in 2003. The United Nations recognizes the Dirty List as a best practice, and it is one of the primary instruments to combat forced labor exploitation in Brazil (Nacoes unidas no Brasil, 2013). The Dirty List identifies employers who submit workers to forced labor exploitation. One of the most severe consequences of the Dirty List is that once an employer is added to the Dirty List, the employer is prevented from taking public lines of credit (Portaria MIN No 1.150 de 18/11/2003, 2003). Data on inclusion in the Dirty List is available for 736 of the total 1,764 cases in the data (N=1,028 missing). Of the 736 cases with non-missing data for the Dirty List outcome variable, 651 cases have defendants who were included in the Dirty List, representing 36.9 percent of criminal and civil cases. Looking at the distribution of cases included on the Dirty List by state, there is a pattern that parallels the overall number of cases per state provided in Figure 11. For example, Para has 152 cases included in the Dirty List, representing 23.4 percent of the 651 cases added to the 114 Dirty List during the study period, followed by Minas Gerais, which has 77 cases representing 11.8 percent of cases, and Mato Gross, which has 76 cases representing 11.7 percent of cases. Parallel to this trend, Para, Minas Gerais, and Mato Grosso represent 20.6 percent, 9.8 percent, and 11.4 percent of cases of forced labor exploitation, respectively. Figure 11. Distribution of cases included in Dirty List by state (N=651) Public Ministry of Labor Case Processing The Public Ministry of Labor uses public civil actions to hold employers charged with forced labor exploitation accountable. There are 432 public civil actions in the dataset; all were filed between 2012-2019. There are 24 Regional Labor Court jurisdictions, described in Appendix C, Table 3. 115 As with the number of cases overall, we see the Regional Labor Court in Belo Horizonte, Minas Gerais filed the most public civil actions, representing 14.4 percent of all actions filed, followed closely by the Regional Labor Court in Para and Amapa, which represent 13.9 percent of actions filed (see Figure 12). Notably, although Sao Paolo was fourth for number of cases identified overall, the Regional Labor Court that has jurisdiction over Sao Paolo filed just 15 public civil actions, the tenth highest overall, or about 3.5 percent of all actions filed. Figure 12. Number of Cases by Regional Labor Court (N=432) Request for moral damages There are two types of moral damages that can be requested in the public court system, either as part of an adjustment of conduct (TAC) or public civil action sentence. One labor prosecutor I interviewed described the purpose of these fines clearly: 116 One of the things we also do is to seek compensation, this compensation often goes toward the community, health, hospitals to compensate for the damage resulting from slave labor. It should go to a fund but we don?t have a fund especially for that and so we often send it back to the community like building schools, public hospitals, congresses, courses and everything that aims to repair this damage. In addition to this indemnity, we also seek compensation in favor of the worker who was subjected to this condition, this money goes to him. There are two indemnities for moral damages, one for collective moral damages, which we revert to society and this one of individual moral damage that goes to the workers themselves. -TP001 Individual moral damages Individual moral damages are paid to individual workers. Of the 432 public civil actions filed, access to initial petitions for moral damages was missing for 75 cases. Therefore, there was data on whether an individual moral damage claim was made in 357 cases. Of these 357 cases, individual moral damages were requested in 75 (21.0 percent). The amounts in Table 13 below represent the amounts paid to all workers. If more than one worker was identified as being victimized by the employer, then these amounts would be divided by all workers. The employer who paid $12,000,000, for example, was required to pay $2,000,000 to each of six workers. Unfortunately, data on the number of workers included in the award are provided in write-in boxes in court records and missing the majority of the time. Table 13. Amount of individual moral damages requested by Regional Labor Court Regional Labor # Court Jurisdiction cases Min (RS$) Max (RS$) Average (RS$) Total all cases (RS$) Rio de Janeiro 3 $ 15,000 $ 150,000 $ 71,667 $ 215,000 Grande S?o Paulo 2 $ 10,000 $ 100,000 $ 55,000 $ 110,000 Minas Gerais 11 $ 3,000 $ 160,000 $ 53,091 $ 584,000 Rio Grande do Sul 1 $ 46,850 $ 46,850 $ 46,850 $ 46,850 Bahia 16 $ 5,000 $ 550,000 $ 150,938 $ 2,415,000 Pernambuco 0 . . . . Cear? 0 . . . . Par? e Amap? 4 $ 30,000 $ 93,351 $ 52,088 $ 208,351 117 Paran? 2 $ 6,000 $ 50,000 $ 28,000 $ 56,000 Distrito Federal e Tocantins 2 $ 20,000 $ 100,000 $ 60,000 $ 120,000 Amazonas e Roraima 4 $ 20,000 $ 220,000 $ 105,000 $ 420,000 Santa Catarina 2 $ 310,000 $ 750,000 $ 530,000 $ 1,060,000 Para?ba 0 . . . . Acre e Rond?nia 4 $ 10,000 $12,000,000 $ 3,028,750 $ 12,115,000 Munic?pios do estado de S?o Paulo 2 $ 10,000 $ 65,000 $ 37,500 $ 75,000 Maranh?o 2 $ 30,000 $ 50,000 $ 40,000 $ 80,000 Esp?rito Santo 2 $ 20,000 $ 30,000 $ 25,000 $ 50,000 Goi?s 5 $ 8,000 $ 65,000 $ 27,000 $ 135,000 Alagoas 1 $ 100,000 $ 100,000 $ 100,000 $ 100,000 Sergipe 1 $ 323,840 $ 323,840 $ 323,840 $ 323,840 Rio Grande do Norte 1 $ 1,000,000 $ 1,000,000 $ 1,000,000 $ 1,000,000 Piau? 1 $ 10,000 $ 10,000 $ 10,000 $ 10,000 Mato Grosso 6 $ 30,000 $ 150,000 $ 88,333 $ 530,000 Mato Grosso do Sul 0 . . . . Total (Brazil) 72 $ 27,973 $ 1,411,360 $ 3,000 $ 19,654,041 One way to estimate the per person amount of moral damages paid is to use the number of workers rescued in a case as a proxy for the number of workers paid out individual moral damages. This will likely overestimate the number of workers involved in the suit, however, and therefore underestimate the per person payout. There are 45 cases with valid data for number of workers rescued and amount of moral damages requested. The minimum per person is $208, the maximum per person is $175,000 and the average per person in $18,505. Right away, it is clear these numbers are inaccurate because we know in at least one case, six workers received $2,000,000 each. In that specific case, data on the number of workers rescued was missing because the case was not inspected by labor investigators and therefore did not have a linked investigation report. Collective moral damages Collective moral damages are described as a ?transindividual? or meta- payments related to the diffuse and collective rights of the community of workers. 118 The individual labor prosecutor in charge of the case decides to what organization these payments will be funneled, but it is typically a community organization or academic institution. For example, the anti-slave labor and trafficking in persons clinic that I partnered with at UFMG is often the recipient of these payments from the Regional Labor Court in Belo Horizonte. In my interviews, I learned that during the pandemic many labor prosecutors designated collective moral damages to COVID-19 relief programs. Collective moral damages are much more common than individual moral damages. Collective moral damages were requested in 339 public civil actions (78.5 percent). Prosecutors did not request collective moral damages in just 17 (3.9 percent) of public civil actions. There is missing data on whether collective moral damages were requested in 75 cases (17.6 percent). Table 14 provides information on the amount of collective moral damages paid in 333 public collective actions with non- missing data. The total collective moral damages paid amount to nearly $1.2 billion, and the average amount paid across all regions was about $3.6 million. Table 14. Amount of collective moral damages requested by Regional Labor Court Regional Labor # Court Jurisdiction cases Min Max Average Total all cases 11 $ 20,000 $ 20,000,000 $ 2,310,909 $ 25,420,000 Rio de Janeiro 10 $ 50,000 $ 2,000,000 $ 495,000 $ 4,950,000 Grande S?o Paulo 41 $ 10,000 $ 10,000,000 $ 1,448,000 $ 59,368,000 Minas Gerais 6 $ 100,000 $ 5,000,000 $ 1,050,000 $ 6,300,000 Rio Grande do Sul Bahia 31 $ 20,000 $ 3,434,000 $ 630,774 $ 19,554,000 5 $ 100,000 $ 1,270,000 $ 534,000 $ 2,670,000 Pernambuco 119 Cear? 5 $ 200,000 $ 1,000,000 $ 647,696 $ 3,238,480 47 $ 30,000 $ 10,000,000 $ 978,511 $ 45,990,000 Par? e Amap? Paran? 10 $ 20,000 $ 2,000,000 $ 724,500 $ 7,245,000 Distrito 20 $ 20,000 $ 100,000,000 $ 6,741,430 $ 134,828,592 Federal e Tocantins 15 $ 1,000 $ 5,000,000 $ 976,733 $ 14,651,000 Amazonas e Roraima 9 $ 100,000 $ 5,000,000 $ 1,359,489 $ 12,235,402 Santa Catarina1 Para?ba 2 $ 300,000 $ 1,000,000 $ 650,000 $ 1,300,000 18 $ 50,000 $ 100,000,000 $ 11,891,667 $ 214,050,000 Acre e Rond?nia 18 $ 50,000 $ 500,000,000 $ 28,321,667 $ 509,790,000 Munic?pios do estado de S?o Paulo Maranh?o 22 $ 30,000 $ 19,000,000 $ 1,591,364 $ 35,010,000 2 $ 200,000 $ 500,000 $ 350,000 $ 700,000 Esp?rito Santo Goi?s 22 $ 10,000 $ 10,000,000 $ 1,372,626 $ 30,197,772 Alagoas 2 $ 200,000 $ 1,000,000 $ 600,000 $ 1,200,000 Sergipe 4 $ 30,000 $ 3,774,069 $ 1,498,017 $ 5,992,069 3 $ 1,000,000 $ 5,000,000 $ 2,333,333 $ 7,000,000 Rio Grande do Norte Piau? 5 $ 50,000 $ 1,000,000 $ 510,000 $ 2,550,000 19 $ 150,000 $ 6,000,000 $ 962,842 $ 18,293,990 Mato Grosso 6 $ 10,000 $ 20,000,000 $ 3,568,333 $ 21,410,000 Mato Grosso do Sul Total (Brazil) 333 $ 1,000 $ 500,000,000 $ 3,555,388 $ 1,183,944,304 Request for judge to declare the presence of forced labor exploitation The public labor prosecutor can request that the courts recognize the presence of forced labor exploitation in the sentence. This judicial declaration has a few outcomes, including making the employer eligible to be included on the Dirty List, 1 Data said ?minimum amount requested? for one case. In consultation with the law clinic, we assigned a value of $1,000 RS$ to this case. 120 making workers eligible to receive social assistance benefits, and affording victimized immigrant workers without legal status rights to residency and social benefit programs. As shown in Table 15, there is non-missing data about whether the prosecutor requested a judicial declaration acknowledging the presence of forced labor exploitation for 338 of the 432 public civil actions in the data. As shown in the table below, of the 338 cases with non-missing data, prosecutors requested this declaration in just 22 (5.1 percent of all public civil actions filed). Notably, among the 22 cases where prosecutors requested a judicial declaration, judges granted a declaration in just 4 cases (18.2 percent of requested); however, judges granted 44 declarations in cases where the prosecutor did not make a request (13.9 percent of not requested cases). This pattern seems to suggest that prosecutor?s request for a declaration has no bearing on the judges? decisions to make a declaration. Table 15. Number of public civil actions in which declaration of forced labor exploitation was requested and granted # Granted # Not Granted # Missing Total Requested 9 4 9 22 Not Requested 44 99 173 316 Missing 10 19 65 94 Total 63 122 247 432 Sentencing in public civil actions Typically, sentences for public civil actions require that employers found guilty of forced labor exploitation make changes to the work environment and treatment and/or payment of their workers moving forward. Additionally, civil 121 sentences carry large fines and/or forfeiture of assets; these fines are in addition to the individual and collective moral damages that the public prosecutor for the Ministry of Labor requests. Figure 13 below shows a flow chart for public civil actions. Figure 13. Public civil action flow chart 192 cases referred directly 1,672 labor inspections from Ministry of Labor 1,332 not referred for 432 referred for public public civil action civil action (240 received labor inspection) 171 Full/Partial decisions against defendant At any point in the processing of a public civil action, however, a defendant can choose to settle, making an agreement instead of receiving a sentence. If a defendant chooses to settle, they are in a better position to negotiate reduced fines. As shown in Figure 13. below, 171 cases (39.6 percent) received a full or partial conviction, 174 (40.3 percent) ended in an agreement, 18 cases (4.2 percent) were not convicted/acquitted, 16 cases (3.7 percent) were dropped (typically this happens when the judge deems there is insufficient evidence to move the case forward), and 53 cases (12.3 percent) were still in progress as of December 2019. A full agreement or full conviction means that the judge accepted all of the Ministry of Labor?s requests 122 at sentencing (i.e., individual and collective moral damages), whereas in a partial agreement or conviction, there may be reduced fees. Of the convictions, only 24 of 171 were full convictions; the rest were partial. Conversely, only four of 174 agreements are partial agreements. This is somewhat misleading, potentially, because prosecutors may lower their ?full ask? when a defendant agrees to settle. Figure 14. Sentencing outcomes of forced labor exploitation cases processed in civil court system (N=432) in progress 53 case dropped 16 acquitted 18 full/partial agreement 174 full/partial conviction 171 0 50 100 150 200 Court recognition of elements of forced labor exploitation Table 16 shows the characteristics of forced labor exploitation that were recognized in any type of civil outcome (i.e., full/ partial sentence or agreement) as well as the characteristics that were recognized in the labor inspector?s report for cases that ended in a sentence or agreement. Overall, civil courts recognized the presence of each characteristic in about one-third to one-half as many cases as did labor inspectors. Forced labor exploitation was present in the inspection reports of all cases in the data but recognized in just about one-third (34.05 percent) of civil case 123 outcomes. Similarly, an exhausting workday was recognized by inspectors in 22.78 percent of cases and just 8.7 percent of case outcomes. In nearly all cases (90 percent), inspectors identified degrading conditions, but the civil courts recognized this element in just 44.62 percent of cases. The only characteristic that does not follow this pattern is forced labor, which was identified in about one percent of inspection reports but in about five percent of civil case outcomes. Table 16. Characteristics of forced labor exploitation recognized in case outcome and in inspection report for public civil actions ending in an agreement or sentence Recognized by Civil Court Prevalence in Inspector Report Obs Mean SD Obs Mean SD Forced labor exploitation 185 34.05% 47.52% Exhausting workday 184 8.70% 28.25% 180 22.78% 42.06% Degrading conditions 186 44.62% 49.84% 180 90.00% 30.08% Debt servitude 186 1.08% 10.34% 180 17.78% 38.34% Forced labor 186 5.38% 22.62% 180 1.11% 10.51% Recruitment fraud 185 3.24% 17.76% 180 10.56% 30.81% Restriction of transportation 186 0.00% 0.00% 180 8.89% 28.54% Ostensive surveillance 183 0.55% 7.39% 180 3.98% 19.60% Withholding documents 186 1.61% 12.63% 180 8.33% 27.72% Case processing length Figure 14 shows the stages of civil case processing. To determine the case processing time for public civil actions, I created a variable that measured the length between the date the public civil action was filed in civil court and the date of the last court action taken for that case using the following variables: sentence date, date the court decided not to issue a sentence/convict, date the case was dropped, or issuance of the final agreement. 124 Figure 15. Stages of public civil action processing File a Appeal Inspection public civil Sentence Final action (if any) sentencing Case processing length is not available for cases that are still in progress. In some cases, the sentence was appealed. In these cases, I use the final sentence date of the last appeal. There are 377 cases with non-missing data for the timeline variables needed to construct the case processing length variable. Of these 377 cases, the mean case processing time is 504 days. The minimum time is 15 days and the maximum time is 3,506 days or about 9.6 years. Figure 15 shows average case processing time for public civil actions by outcome. As could be expected, the cases that end in an agreement are shorter, on average, than sentenced cases. 125 Figure 16. Mean number of days case processing time for public civil actions, by case outcome Full/partial conviction (N=171) 724.17 All (N=377) 504.41 Full/partial agreement (N=173) 293.24 Case dropped (N=15) 140.60 Time trends Figure 16 shows the number of public civil actions that began each year and the number that reached final adjudication each year. Between 2013 and 2016, there was a sharp period of growth in the number of cases initiated and closed. Figure 17. Initiation (N=431) and final adjudication (N=295) of public civil actions, by year Initiation Adjudication 80 76 70 63 60 58 51 55 50 46 48 40 42 45 37 38 49 48 30 20 6 8 15 2 6 1510 0 0 3 6 7 2 2006 2008 2010 2012 2014 2016 2018 2020 126 Criminal Sentencing This section focuses on an analysis of the 1,464 criminal actions in the data. The criminal court system in Brazil is divided among five federal districts, described in Appendix C, Table 4 and Figure 1. As with the number of cases overall, and in line with patterns from civil court regions, we see the federal court district of Brasilia, which includes the states of Minas Gerais and Par?, filed the most public criminal actions, representing 73.57 percent of all actions filed (see Figure 17). Figure 18. Number and cumulative percent of criminal actions by federal court district Sentencing in Public Criminal Actions Unlike in the civil court system, criminal cases may have multiple outcomes because each case may have multiple defendants. Unfortunately, the outcome variable does not consistently label which outcome goes with which defendant. Further, the names of the defendants listed in the outcome variable do not always align with the names in the defendant variable. It is, therefore, not advisable to conduct a defendant level analysis of criminal case outcomes. Instead, I present the frequency of each outcome across all cases, (see Table 17). 127 Table 17. Criminal action outcomes Variable Obs Mean Std. Dev. Min Max Frequency Absolved/ acquitted 1,464 38.25% 48.62% 0 1 560 In progress 1,464 30.46% 46.04% 0 1 446 Convicted 1,464 21.04% 40.77% 0 1 308 Statute of limitations 1,464 9.43% 29.23% 0 1 138 Defendant died 1,464 4.92% 21.63% 0 1 72 Suspended 1,464 2.80% 16.50% 0 1 41 Extinct 1,464 2.25% 14.85% 0 1 33 Case rejected 1,464 1.57% 12.44% 0 1 23 Meeting 1,464 1.57% 12.44% 0 1 23 Archived 1,464 0.20% 4.52% 0 1 3 Incompetence 1,464 0.20% 4.52% 0 1 3 Dropped because of double jeopardy 1,464 0.14% 3.69% 0 1 2 Injunction granted 1,464 0.07% 2.61% 0 1 1 Reasons for Acquittal The most common case outcome is acquittal. Judges use the following criteria to determine whether acquittal is appropriate. These criteria are defined by Article 386 of the penal process, below: Art. 386. The judge will acquit the defendant, mentioning the cause in the dispositive part, provided that he recognizes: I - it is proven that there is no such crime; II - there is no proof of the existence of the crime; III - does not constitute a criminal offense; IV - it is proven that the defendant did not collaborate on the criminal offense; V - there is no evidence that the defendant participated in the criminal offense; VI - there are circumstances that exclude the crime or exempt the defendant from penalty (arts. 20, 21, 22, 23, 26 and ? 1 of art. 28, all of the Penal Code), or even if there is founded doubt about its existence; VII - there is insufficient evidence for the conviction. Regarding number four, the Brazilian penal code distinguishes collaborating with other people to commit a crime and participating in the criminal offense. In order to prove that someone was collaborating with other people to commit a crime, 128 four elements must be present. First, more than one person must have been involved in the crime. For establishing guilt, typically one defendant is identified as the organizer/leader and the others as participants. Second, there must be proof that each participant?s actions or lack of actions had a direct impact on the criminal outcome when looking at the causal chain of events. Participants whose behaviors are not directly tied to a causal link are not considered part of the criminal team. Third, there must be some kind of ?psychological or normative? bond between the actors to be considered a criminal collaboration. The example given to justify this is that if Person A and Person B shoot at Person C and Person A fatally shoots Person C while Person B misses, then Person B is still guilty of homicide. If Person A and Person B have a subjective bond, they are both guilty of homicide, whereas if there is not a connection between them, they will have distinct sentencing outcomes. The fourth requirement is simply that criminal collaboration is not present if the participants are connected by a subjective bond, but the defendant did not know they were assisting in the crime. Each of the reasons noted in Article 386 corresponds to the justifications for acquittal with the same number below: I - denial of existence of the crime II - insufficient evidence of the crime III -does not constitute a crime IV - denial of collaboration V - insufficient proof of criminal activity VI - exempt VII - insufficient evidence Data on reason for acquittal is not consistently available and in some cases is available for only one of the defendants. In other cases, it is not stated to which defendants each reason is applied. However, behavior not constituting a crime is the 129 most often cited, followed by insufficient evidence of a crime. This is in line with findings from the interviews. In the interviews, prosecutors mentioned that the most common reason for acquittal in their cases was also that the denial of the existence of the crime. As one federal prosecutor (FP009) said, ?One of the ways you can, for example, check what are the difficulties encountered in your case is to read precisely what leads to acquittal. Usually the judge writes ?the prosecutor doesn?t adequately describe the conditions that constitute a crime.?? Court recognition of elements of forced labor exploitation in criminal cases As in the labor court system, judges in the criminal court system can acknowledge characteristics of forced labor exploitation in the final sentence for a case processed in the criminal court. In Table 18 below, we see that despite degrading conditions being prevalent in nearly all cases (94.4 percent), just over one-third of cases recognize degrading conditions in the sentence (34.5 percent). Just nine percent of sentences recognized exhausting workday or debt servitude, and less than five percent of sentences recognized another characteristic. Physical violence and presence of guns or other weapons was not an option as a characteristic to be recognized at sentencing. Table 18. Characteristics of forced labor exploitation recognized at criminal sentencing Variable Obs. Mean Std. Dev. Min Max Degrading conditions 861 34.84% 47.68% 0 1 Exhausting workday 861 8.59% 28.04% 0 1 Debt servitude 861 8.59% 28.04% 0 1 Restrict transport 861 2.79% 16.47% 0 1 Withhold documents 861 1.97% 13.92% 0 1 Recruitment fraud 861 1.86% 13.51% 0 1 Ostensive surveillance 861 3.83% 19.21% 0 1 130 Co-Occurring crimes Table 19 shows the other crimes which were indicted and convicted along with forced labor exploitation. Most co-occurring crimes for which the federal prosecutor brought charges were not sentenced at final disposition. Co-occurring crimes for which defendants were found guilty include threatening to cause unjust and serious harm, property crimes, not paying minimum wage, domestic recruitment fraud, maintaining a brothel, and falsifying documents. A complete list of other crimes indicted, including those that did not receive a sentence, are available in Appendix C, Table 5. Table 19. Other crimes indicted and sentenced Indicted N=1,464 Sentenced, N=308 Std. Std. Variable Mean Dev. Freq. Mean Dev. Freq. Article 147-Threatening, by word, writing, gesture, or any other symbolic means, to cause unjust & serious harm 0.48% 6.90% 7 0.97% 9.84% 3 Article 171-Crime against property 0.41% 6.39% 6 0.32% 5.70% 1 Article 203- Not paying minimum wage 28.55% 45.18% 418 3.57% 18.59% 11 Article 207- Recruit workers through fraud to take them to another location within country 14.62% 35.34% 214 4.87% 21.56% 15 Article 229-Maintain, on their own account or as a third party, an establishment in which sexual exploitation occurs, whether or not there is an intention for profit or direct mediation 0.34% 5.84% 5 0.32% 5.70% 1 Article 297- Falsify, in whole or in part, a public document, or change a true public document 25.20% 43.43% 369 11.36% 31.79% 35 Article 299-To omit, in a public or private document, a declaration that should be included in it, or to insert or cause to insert a false or different declaration 0.89% 9.38% 13 0.65% 8.05% 2 Punishments for Criminal Sentences If an individual is convicted for forced labor exploitation, they will be ordered to pay a fine based on their salary. Additionally, the court will sentence them to 131 prison time. However, individuals typically only serve out their sentence in a detention facility (i.e., prison) if the sentence length is four years or greater; otherwise, defendants serve out their time in community supervision. I attempted to search for publicly available arrest warrants for defendants associated with cases in this data, as this would indicate that the individuals had been arrested to be taken to a detention facility. Arrest warrants were rarely available; I found no warrants for cases with sentence length less than four years and only 11 warrants for individuals with sentence lengths of four years or greater. After consulting with my host, we decided that it is most accurate to assume that only defendants with sentence lengths greater than four years served prison time. This assumption likely overestimates the number of cases that resulted in prison time and thus represents an understatement of the impunity associated with forced labor exploitation cases. The mean sentence length associated with the crime of forced labor exploitation (Article 149) is 56. 6 months (4.72 years); the range is 14-236 months. Figure 18 shows a flow chart beginning with the number of labor inspections and ending with the number of cases associated with prison time. Prison time is assumed for sentence length four years or greater. 132 Figure 19. Criminal action processing flow chart 1,672 labor inspections 893 police investigations 1,131 inspections identified 333 cases received forced labor exploitation police (560 received police investigation but investigation) no labor inspection 1,464 unique criminal court cases 308 convicted in first degree for 111 sen tenced to mor e than four years Criminal case processing length Figure 19 shows the stages of criminal case processing. To determine the case processing time for public criminal actions, I created a variable that measured the length between the date the public criminal action was filed in civil court and the date of the last court action taken for that case using the following variables: sentence date, date the court decided not to issue a sentence/convict, or date the case was dropped. 133 Figure 20. Stages of criminal case processing Police File a public Appeal Inspection Investigation criminal Sentence Final (if requested) action (if any) sentencing There are 1,037 cases with non-missing data for the timeline variables needed to construct the case processing length variable. Of these 1,037 cases, the mean case processing time is 1,710.4 days (4.7 years). The minimum time is 92 days, and the maximum time is 9,129 days or about 25 years. Figure 20 shows average case processing time for public civil actions by outcome. As could be expected, the cases that end in an agreement are shorter, on average, than sentenced cases. Figure 21. Mean number of days case processing time for criminal actions, by case outcome Case Rejected (n=14) 1443.3 Defendant Died (n=61) 1795.5 Statute of Limitations (n=127) 2660.7 Acquitted (n=559) 1556.5 Convicted (n=308) 1851 0 500 1000 1500 2000 2500 3000 134 5.4 Variables used in analyses for objectives one through three Table 20, below, summarizes the variables used in each of the analyses that follow. Objective One is descriptive and does not include any dependent variables. Instead, I list the independent variables that contribute to the formation of the latent class variable typologies of forced labor exploitation. These indicator variables include dichotomous indicators for whether the following elements of forced labor exploitation were present: degrading conditions, exhausting work, debt servitude, restricting movement, withholding documents, recruitment fraud, ostensive surveillance, and presence of weapons. Objective Two examines which factors predict the most likely latent class based on posterior probabilities of latent class membership. I restricted independent variables for Objective Two to factors related to the case itself, not any variables related to case processing. I use number of individuals on the inspection team as a proxy for the severity of the crime and therefore community protection because interview respondents indicated that more individuals participate in the inspections when a more serious threat is anticipated. The number of workers rescued is included as a proxy for the size of the operation. I also included the presence of other crimes that co-occurred with forced labor exploitation and whether the crime took place in a rural sector (see Figure 21). Whether a crime took place in the rural sector is a proxy for several practical considerations. Courts are less likely to keep in contact with victims rescued from rural areas, and respondents described individuals living in rural areas as living in greater poverty. Thus, as is described later, decision makers were more hesitant to label these farm managers. 135 Figure 22. Theoretical map Community protection Number rescued Number on inspection team Blameworthiness of the Focal Concerns offender Evidence of additional crimes Practical considerations Rural sector Objective Three predicts different case outcomes in the criminal and labor court systems. For civil cases, I examine whether the case was added to the Dirty List, whether a TAC was signed, and the outcome of the criminal case. For these three outcomes, there is not enough valid data for all latent classes for a full analysis, so I present cross tabs of the variable summary by predicted latent class membership. I also examine the amount of collective and individual moral damages and for these two outcomes I am able to conduct a full analysis with predictors, including controls for whether there was a judicial declaration acknowledging presence of forced labor exploitation, and region. For criminal cases, these outcomes include whether the case was convicted, whether it was acquitted, case length, and sentence length. In these models where the 136 primary independent variable is the most likely latent class, I also include controls for whether a police investigation took place, the number of other indicted crimes, and the court region. Table 20. Summary of variables used in analyses Objective Dependent Variable(s) Independent Variables O1: Describe characteristics of ? Degrading conditions forced labor exploitation cases ? Exhausting work forwarded for processing in ? Debt servitude criminal and civil courts ? Restriction of N/A transportation or movement ? Withholding documents ? Recruitment fraud ? Ostensive surveillance ? Presence of weapons O2: Analyze differences in ? Most likely latent ? Number on inspection typologies of forced labor class team exploitation by key ? # workers rescued characteristics during inspection ? Report Identified Evidence of additional crimes: ? 132 ? 203 ? 207 ? Other ? Rural sector O3: Examine the sentencing Civil Outcomes outcomes for different types of ? Dirty list ? Most likely latent class forced labor exploitation cases ? TAC signed ? Declaration of forced ? Outcome of labor labor exploitation case ? Region ? Individual moral damages ? Collective moral damages Criminal Outcomes ? Convicted ? Most likely latent class ? Acquitted ? Whether a police ? Length of case investigation took place ? Sentence length ? Number of other indicted crimes ? Criminal court region 137 5.5 Objective One: Describe characteristics of forced labor exploitation cases forwarded for processing in criminal and civil courts 5.5.1 Latent Class Analysis Latent class models were based on all cases with non-missing data for any of the characteristics of forced labor exploitation. As mentioned above, however, only cases that were inspected by labor inspectors had valid data about the characteristics of forced labor exploitation. Thus, the analysis is limited to those cases that entered the criminal and/or civil court systems that had been investigated by labor inspectors. I examined model solutions for one to five classes. Table 21 describes goodness of fit indicators for each class solutions. Table 21. Goodness of fit indicators for latent class analysis (N=1,075) Adjusted VLMR PBLRT No. Classes AIC BIC Adjusted BIC p- value* Entropy R2 p- value* 1 4919.99 5014.61 4954.27 0.00 0.73 0 2 4698.85 4783.51 4729.52 0.00 0.70 0 3 4637.04 4766.52 4683.94 0.25 0.81 0 4 4629.57 4803.87 4692.71 0.20 0.88 0 5 4631.52 4850.65 4710.89 0.18 0.88 0.29 Numbers are bolded for the class solution indicated by each test **PBLRT is the Parametric Bootstrapped Likelihood Ratio Test For AIC, BIC, and Adjusted BIC, lower numbers represent better fitting solutions. The AIC points to a four-class solution as best fitting, and the BIC and Adjusted BIC point to the three-class model as best fitting. The Adjusted Vuong-Lo- Mendell-Rubin Likelihood Ratio (VLMR) and the Parametric Bootstrapped Likelihood Ratio Tests (PBLRT) compare the Nth Class to the N-1 class. If the p- value is significant, the Nth class is better fitting. The Adjusted VLMR points to the 138 two-class solution as best fitting while the PBLRT points to the fourth class as best fitting. Finally, higher numbers for entropy r-squared indicate better fitting solutions. Again, the entropy statistic indicates that the four- or five-class solution is best fitting. As is often the case, the fit statistics do not point to a single best solution. After examining the solutions qualitatively, I decided the that the three-class solution was the best model. Figure 21 shows the characteristics and probabilities of engaging in each different behavior associated with forced labor exploitation for cases in each of the latent class profiles (also shown in Table 22). The figure and the table provide the same information. Figure 23. Latent classes of forced labor exploitation 100.00% 90.00% 80.00% 70.00% Class 1, 13.21% Class 2, 83.81% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Cases in all classes have a high probability of having the presence of degrading conditions (99 percent, 93 percent, and 97 percent, respectively). Cases in Class 1, which represent about 13 percent of all cases, have a high likelihood of debt 139 servitude (72 percent) and moderate likelihood of recruitment fraud (38 percent). Cases in Class 2, which represent the about 84 percent of cases, are unlikely to have any characteristic aside from degrading conditions. Finally, cases in Class 3, which represent just 3 percent of cases, are characterized by a high likelihood of the presence of armed surveillance (72 percent) and weapons (100 percent). Accordingly, I named the classes Degrading Conditions and Debt Servitude, Degrading Conditions, and Degrading Conditions and Weapons and Surveillance, respectively. Table 22. Latent class marginal probabilities for characteristics of forced labor exploitation, 3-class solution Class 1 Class 2 Class 3 Degrading Conditions and Degrading Degrading Conditions and Weapons Debt Servitude Conditions and Surveillance Class Count and Proportion N=142, 13.21% N=901, 83.81% N=32, 2.98% Marg Std. Margin Std. Err. in Err. Margin Std. Err. Degrading 93.20 Conditions 99.90% 2.00% % 0.90% 96.70% 5.70% 11.40 Exhausting Work 49.90% 12.50% % 1.40% 24.90% 7.60% 11.70 Debt Servitude 72.40% 10.40% % 2.50% 52.90% 22.40% Restricting 2.90 Transport/Movement 26.40% 4.40% % 1.30% 30.00% 19.50% Withholding 3.10 Documents 20.30% 7.20% % 0.90% 22.90% 25.30% 1.30 Recruitment Fraud 37.70% 6.80% % 1.40% 30.60% 10.30% Ostensive 0.20 surveillance 3.50% 1.80% % 0.20% 72.50% 73.00% 1.00 Weapons 0.80% 14.40% % 0.70% 100.00% 0.00% Although cases that enter both the criminal and civil systems are identical, I was not sure if cases that were only handled in civil or criminal court were distinct enough to merit separation. As a sensitivity check, I conducted the LCA models and included whether the case was processed in the criminal justice system or the civil system. Neither variable was a significant covariate in the models, indicating that it is 140 not appropriate to conduct separate LCA models for criminal and civil cases (see Appendix D, Table 1). These latent classes align well with the description of the characteristics of forced labor exploitation described by interview participants. For Class 1, Degrading Conditions and Debt Servitude, it is not surprising that other forms of control are not present. As the prosecutor quoted earlier stated, debt servitude is an effective means of control that often is sufficient on its own to keep workers in an exploitive situation. Another labor prosecutor also described how debt servitude can be used to maintain control over workers and how it goes hand-in-hand with fraudulent recruitment: Often the worker, especially when he is fraudulently recruited, arrives at the company with some debts to pay because he had to leave some money with the family, he had to pay the ticket to be able to be there and he gets stuck with the employer?With debt bondage the worker often has a moral obligation?he does not want to leave there owing to the employer so he gets trapped because he has a debt.- TP001 For Class 2, as described earlier, interview participants noted that degrading conditions were by far the most prevalent characteristic of forced labor exploitation. Further, respondents noted that often intangible forms of isolation (e.g., geographic, linguistic, social) in combination with degrading conditions are enough to keep workers stuck in these situations. For example, one labor inspector (LA003) described vulnerabilities of immigrant workers submitted to degrading conditions: ??the fear that because they are illegally here in Brazil, the fear that they will return to their countries where they left, they end up submitting to any situation, they end up working for food and housing.? A federal judge also described the psychological control employers can maintain through degrading conditions alone: 141 The employer restricts his freedom but not necessarily by violent means. The worker has to be without food, in an extremely dirty environment, with nowhere to sleep, with venomous animals, without safety equipment to protect the use of pesticides, without drinking water, without bathing, in such a way that he gets sick and does not get out of that situation there is a deprivation of his freedom, but by psychological and emotional means. -FJ008 Class 3 contains the most forms of force, fraud, and coercion and, in line with findings from the interviews, includes the lowest proportion of cases. Interview respondents indicated that although things like physical violence, armed guards, and use of weapons came up, these more overt forms of control were rare. As one federal prosecutor (FP006) explained, ??the cases I had never included restriction of freedom to come and go or severe violence?? 5.6 Objective Two: Examine factors that predict membership in each Forced Labor Exploitation Case Profile 5.6.1 Quantitative Findings To better understand the factors that predict each class of forced labor exploitation cases, I conducted multinomial logistic regression models with the latent classes as dependent variables and relevant covariates as independent variables. In this analysis, I include only covariates measured before case processing to gain a better understanding of the characteristics associated with the crime of forced labor exploitation itself, rather than characteristics associated with the investigation and prosecution of the crime. Some variables occur temporally after investigation, but I include them as I think they represent relevant proxy variables. For example, number of workers rescued is a good proxy variable for the size of the operation. 142 As shown in Table 23, there are no significant differences in predictors of Class 1 versus the reference, Class 3. For Class 2, however, there are differences relative to Class 3. Both the number on the inspection team and the number of workers rescued have a significant p-value, but the odds ratios are very close to one. The odds are slightly lower that a case in Class 2 would also include a violation of Article 207 or another crime type. The odds are 1.6 times higher that a case in Class 2 would include work in the rural sector relative to cases in Class 3. Table 23. Tests of categorical latent variable multinomial logistic regression using the 3-step procedure (N=944) Two-Tailed p- Estimate OR SE value Class 1: Degrading Conditions and Debt Servitude Number on inspection team -0.03 0.97 0.024 0.246 # workers rescued during inspection 0.00 1.00 0.002 0.371 Report Identified Evidence of additional crimes: 132: danger to life or health of others -0.216 0.81 0.76 0.776 203: use of fraud or violence to deny labor rights 1.21 3.37 1.051 0.248 207: relocating workers under false pretenses 0.30 1.34 0.607 0.626 Other -1.45 0.23 0.77 0.06 Rural sector -0.38 0.68 0.629 0.547 Intercept 2.71 0.69 0 Class 2: Degrading Conditions Number on inspection team -0.06 0.95 0.02 0.007 # workers rescued during inspection -0.04 0.96 0.01 0.005 Report Identified Evidence of other crimes: 1.00 132: danger to life or health of others -1.501 0.22 0.80 0.06 203: use of fraud or violence to deny labor rights 0.83 2.29 1.13 0.464 207: relocating workers under false pretenses -1.63 0.20 0.74 0.027 Other Crimes -1.86 0.16 0.74 0.012 Rural sector 1.60 4.96 0.68 0.019 Intercept 3.81 0.728 0 Class 3: Degrading Conditions and Weapons and surveillance (reference) Indicates p< .10 p<.05; indicates p<.01 143 It makes sense that Class 1 and 3 have fewer significant differences than do Class 2 and 3 because Class 1 and 3 have similar marginal probabilities of most of the characteristics of forced labor exploitation, but Class 3 includes weapons and ostensive surveillance while Class 1 does not. That Class 2 is more likely than Class 3 to involve a rural sector also aligns with qualitative findings from the interviews and prior literature. When individuals are geographically isolated, as is often the case in rural work in Brazil, employers do not need to implement a variety of coercive techniques to keep employees at work. Employers can keep the employees stuck in degrading working conditions because the employees have no recourse to leave. That Class 2 is less likely to be associated with a violation of Article 207, enticing workers in order to take them from one state to another, is also interesting. This suggests that victims in Class 2 are more likely to be exploited in their home state. However, in the interviews, respondents said that employers typically will offer high interest loans to help employees move states for a job opportunity, a system that typically constitutes debt servitude. Thus, it is logical that violations of Article 207 are more likely in Class 3 where debt servitude is more prevalent. 5.7 Objective Three: Examine the sentencing outcomes for different types of forced labor exploitation cases In this section, I begin by describing the factors that influence decision making and case outcomes as described by interview participants. I then proceed to the quantitative findings. 144 5.7.1 Qualitative findings on outcomes of forced labor exploitation cases Interviews did not explicitly focus on the actual outcomes of forced labor exploitation cases in general; however, the topic did emerge in a few interviews. When discussions about case outcomes came up, respondents mentioned that securing convictions in criminal court is incredibly difficult. One federal judge (FJ001), with more than 15 years of tenure, recalled only one case ever making it to sentencing, and she acquitted the defendants: ?The only one that ended up going on to sentencing was from a large supermarket chain, it was urban slave labor and ended in the acquittal of everyone, without conviction for anyone.? Conversations about impunity also came up somewhat regularly. If this topic came up (topic came up with eight respondents), every respondent except one said there was a high degree of impunity for forced labor exploitation. The one federal prosecutor (FP003) who did not believe there was impunity said the following: ?The feasible cases that are filed, they will not see impunity, [impunity] exists as in all processes but it is not so big that it is so alarming?. What exists a lot is the black box, I think there are many cases that are not brought to justice, the state's capacity to investigate those cases is still not enough.? This prosecutor said that there is an average amount of cases for which guilty defendants go unpunished, but the real impunity is for those cases that are identified. Rather than ask about specific case outcomes, I asked how respondents make decisions about cases, including questions about prioritizing cases for labor inspectors, police officers, and prosecutors, and decisions about disposition and sentencing for cases for judges. I grouped factors that influence decision making into 145 categories based on focal concerns theory. Below, I describe the factors that influence respondents? decision making or their perceptions of others? decision making. I group these factors by focal concern categories. Community protection With what ethics and morals are you going to build a nation based on the exploitation of a person? ? PO001 Focal concerns theory describes community protection as the decision maker?s calculation about the potential danger to the public if the defendant reenters the community. Factors related to community protection that came up in the interviews included decisions around whether decision makers thought a crime took place at all, the severity of the crime, and whether violence or restriction of freedom were involved. Finally, a few respondents mentioned that criminal punishments do little to protect the community because victims often return to an exploitive situation whether or not one particular employer is punished. Seventeen respondents mentioned that Article 149 is subjective and leaves too much open to interpretation. Subjectivity in the law was described as being associated with cases being dropped or acquitted. Respondents never said the subjectivity led to innocent individuals being convicted. For example, one federal judge (FJ006) said: If it is not an absolutely extreme situation, I acquit the defendants. I try to escape from subjectivisms, from those situations where some will find it a crime, others will think it is not? I think the legislation should be clearer in this regard. - FJ006 A police officer also said that subjectivity in the law makes investigations difficult and that this subjectivity means that only the most extreme cases will be prosecuted: 146 It is complex to do the investigation with crimes like slave labor, for example, because it is very open to interpretation. Often, whether there is any criminal punishment depends very much on an almost extreme characterization of the crime within the police investigation. -PO001 Several respondents in the criminal and labor court systems mentioned that the issue is not with the law itself but the lack of training for judges. For example, when asked what the main challenges are for obtaining convictions, one federal prosecutor (FP006) said, ?A lack of understanding among judges about what constitutes modern slavery.? Another federal prosecutor (FP007) similarly stated, ?Some judges understood that?in cases that for us [prosecutors] were very serious and characterize slave labor, did not characterize slave labor and thus acquitted the defendants.? A federal judge (FJ006) even stated, ?The federal judge is not a judge who has his training focused on labor law?this ends up getting in the way of judgment.? In fact, every federal judge said that they did not receive formal training related to forced labor exploitation. A newer judge (FJ008) said that judges were expected to rely on their own experiences: ?I have only my life experience and experience as a judge to fall back on, but I?ve only got two years of experience as a judge.? Nearly every federal judged echoed this idea that they are meant to use their experience to make decisions about a case. Even the most tenured judges thought this system was flawed because they infrequently came across forced labor exploitation cases. One judge further noted that perhaps judges are not in the best position to draw on life experience when making decisions because their life experiences are very different from the experiences of people involved in forced labor exploitation: Basically, relying on using the life experience that the judge has, it does bring some difficulties because [the forced labor exploitation case] ends up being experiences that we do not have very close knowledge of. For example, maybe 147 it is difficult for you, who has an urban life, to know about facts that occur in rural life? It is a big challenge because we have to look at a reality that we don't know about and are obliged to learn about. -FJ005 In line with the idea that judges aim to avoid uncertainty about the presence of an actual crime, some respondents mentioned that judges will not convict unless there is physical violence or restriction of freedom present. For example, a federal prosecutor (FP009) said, ?We face this issue that many judges end up understanding that in order to have modern slave labor, the person would have to be prevented from coming and going.? A labor prosecutor (TP005) also mentioned, ?In my view, in Brazil there is a misconception that exploiting slave labor requires an iron ball in your foot. In reality, even when slavery was allowed in Brazil, it was not like that.? A prosecutor from the labor court system (TP001) gave a specific example of a case they viewed as characterizing forced labor exploitation, but the judge did not agree: For example, we found a situation of a worker who stayed 24 hours at work with a 24-hour rest. A 24-hour day, in my view, this characterizes an exhaustive day, especially workers who use machines to work, in the testimony of workers, they went so far as to say that they dozed off because they were sleepy and nearly ran over their colleagues, this led to exhaustion but the court understood that there was no violence, that they were free to come and go, that they were earning better and so it did not characterize slave labor. -TP001 One federal judge gave an example of the type of decision making that frustrated prosecutors: The victim does not know what reducing a similar slave condition is, they are very simple and humble people, but it is important that I know if they felt deprived of their freedom, if they thought they could leave whenever they wanted. If they did not feel deprived of his freedom, even though the working conditions were bad, but he thought he could go home, there is no slave labor. So his opinion is important. - FJ008 148 While there were concerns about judges? willingness to convict in the absence of extreme or violent conditions, some respondents also noted that there are plenty of cases that simply are not severe enough to be classified as forced labor exploitation. For example, one labor judge (JT001) gave a specific example of a case in which he acquitted the defendants because there was not sufficient evidence of a crime: ?The biggest complaint there was that there was no water supply but it was found that there was a very good water mine there and that they drank water from that mine. The food too, it was said that they had no food but they said they took food from home and then a cook was hired to cook for them.? A police officer also mentioned that few cases are convicted, but that it does more harm than good to inaccurately classify any case as forced labor exploitation; Often [our investigation] does not lead to any conviction, either the prosecutor himself or the judge do not believe a crime took place? I understand though. When you say that everything is slave labor, but in the end it doesn?t really characterize slave labor, it ends up discrediting the work. - P003 Blameworthiness Focal concerns related to blameworthiness of the offender that came up in the interview include the role the defendant had in the offense, whether the defendant admits responsibility and whether the victim believes they were exploited. A large proportion of the interviews focused on the hierarchy of forced labor exploitation operations, the different individuals involved in recruiting and exploiting victims, and whether each of these individuals was truly guilty of a crime themselves or another victim. In rural work in particular, there are several individuals who may be involved in the labor operation itself. One judge described the different roles as follows: 149 We typically indict the farm owner, the farm manager, foreman, and the one considered to be a ?gato.? And then we have a serious problem. Sometimes the owner of the farm lives in S?o Paulo, BH or Rio Grande do Sul and has a farm of 400 hectares, 5,000 hectares here in Mato Grosso and he usually has other activities besides the farm. -FJ004 Several participants discussed the difficulty of holding landowners accountable. For example, one federal judge (FJ006) explained, ?One person uses 3 or 4 people as people who are formally in charge of that business to hide their presence.? A federal prosecutor elaborated on this problem: What happens is that [the owner] can say that he has nothing to do with it and that the person who hires the farm workers is a person he hired, who is a third company who went to get labor in another state that it is very common for rural workers on a given property to be from another state and then the owner of the property says that it has nothing to do with it and that he has no responsibility.-FP009 A situation that respondents noted is particularly difficult is investigating up the supply chain to identify the proprietor who is ultimately responsible for using exploited labor, especially in larger operations. As one federal prosecutor (FP003) explained, ?When crimes occur in more complex production chains you have more difficulty in demonstrating that the owner of the services was aware and that these crimes occurred with their suppliers or its business partners.? A labor prosecutor (TP001) expressed a similar frustration: ?It is often difficult to identify who the central beneficiary really is, whether he has benefited directly or indirectly.? Two respondents mentioned that it is easy to find the proprietor. These respondents both described situations that involved smaller operations in which the owner lived or worked on the jobsite. For example, one judge described that it is easy to find the landowner but more difficult to find the ?gato? or the individual responsible for recruiting workers: 150 Defendants are not difficult to locate. Because they are owners of the farm, it is easy to locate them on the farms, they are people who declare income. It is sometimes difficult to find the labor intermediary, gatos, who are also poor people who do not declare income tax and have no registration with public agencies. -FJ008 The idea that gatos were difficult to locate came up frequently among respondents in the criminal justice system. Labor inspectors, however, had different views about the culpability of gatos and other middlemen, who they noted were often victims themselves: For example, I can consider that the owner of the workshop is a worker in relation to the top company in the supply chain, but the police understand that he has to be arrested, even though he is a victim of human trafficking too?My understanding is that if the worker is a victim first, then their other crimes would not be considered? there are several immigrants here in S?o Paulo who are arrested for drug trafficking and many of them were victims of human trafficking. So, when applying criminal law, you can cause serious violations of human rights. - LA005 Difficulty locating or proving the culpability of proprietors and business owners went hand-in-hand with another theme in the interviews, which is that defendants rarely admitted culpability. During conversations about whether defendants admitted guilt, I asked respondents if they thought the defendants truly did not think they committed a crime or if they were simply trying to avoid charges. A labor prosecutor explained that in the case of large brands that contract out labor to smaller businesses, they may know that forced labor exploitation takes place but genuinely do not believe they can be held responsible for it: The big brands sometimes understand that they would have no responsibility for these people, because they had contracted the work to another company, but in this situation that we call responsibility for the production chain, he has to have the responsibility to know where the material it is coming from, because he is selling that product, he already knows that it could not be produced for that price without being in breach of labor legislation?This was very difficult for the people 151 involved, sometimes the judiciary, the companies themselves had difficulty understanding that they were responsible for this. ? TP004 In general, conversations about whether the business owner recognized their culpability were intertwined with conversations about whether the victim recognized themselves as having experienced forced labor exploitation and the cultural acceptance of poor working conditions in general. Unlike other research on decision making of court actors which finds that defendants who do not accept responsibility are seen as a more blameworthy, conversations about whether the defendants accepted guilt were met with understanding from interview respondents who were willing to write it off as a larger scale issue that individuals do not know more about labor rights. For example, one federal judge (FP008) said, ?We need to prove that the person who committed the crime was aware of [the crime], we have to prove intent, it may happen that the owner of the farm is not convicted. And I don't see a problem with that either.? Another judge (FJ001) mentioned that all of the rescued workers in one case testified on behalf of the defendant: ?In this case, it was very curious because the defense arranged as a witness all the workers appointed by the Public Ministry as working in degrading situations and all testified in favor of the company.? Another judge mentioned that a case can be convicted regardless of whether the victim agrees that a crime occurred, but the lack of victim testimony can make it impossible to prove a crime existed: I do not consider that because [the worker] thinks [the conditions] are normal, we cannot move to a sentence. I don't think [the worker?s] opinion matters much. For example, if a victim of a homicide attempt has an intimate relationship with her abuser, when analyzing this, the crime does not cease to exist because of 152 that. But that can influence the process if the victim won?t testify and they even continue working in the same situation. -FJ003 Another federal judge said that having victim testimony makes it easier to convict: What would be easier would be the situation in which the victim of slave labor would recognize and affirm this, but this is not the case, either because of a situation where the person feels intimidated or it stems from the economic and social situation of that person, they are very poor people who have had a very difficult life. -FJ006 Practical considerations Conversations related to practical considerations included topics related to the amount of time that elapsed between the investigation, indictment, and sentencing and whether the defendant is an individual farmer or part of a lager corporation. Fourteen respondents mentioned that slow case processing times are a barrier to securing convictions in forced labor exploitation cases. As a federal judge explained: Just for you to have an idea, my penultimate decision is a case that in theory occurred in June 2002, before the 2003 reform... this has dragged on for years in the investigation, it arrives at the public ministry, then there is more time until it arrives for judgement, and then it is another many years that the process is going on. Those people who committed these crimes, there is no longer any idea where they are, to find these people is very difficult. -FJ005 Another federal prosecutor (FP001) explained that long sentence lengths can make it difficult to secure a conviction: ?As a rule, we end up not convicting people for this crime? The cases in which the defendant doesn?t end up in prison? are crimes that advance more slowly.? Another judge (FJ006), when asked about best practices for forced labor exploitation cases, said, ?Always try to be as quick as possible in this type of process. It is difficult to think of another best practice?.? One reason that case length can influence the outcome of a case is that the longer the case lasts, the less reliable the evidence is. As one federal prosecutor 153 (FP006) said, ?There are situations that you will need to hear a witness, or even the defendant died, disappears or changes the region and this ends up making the process more difficult.? A primary issue with long cases is that victim testimony becomes difficult to secure. As one federal judge said: It doesn't just depend on the judge, there are several obstacles that happen because the more time passes between the occurrence of the fact and the trial, this ends up making it difficult to analyze the evidence. Sometimes we are unable to get testimony from witnesses and victims, and this ends up making it very difficult. -FJ003 Because of difficulties identifying witnesses to testify years after the crime occurs, a few labor inspectors advocated for letting their interviews with workers during the inspection be used in place of victim testimony. One labor inspector described several benefits that would come from this practice: It makes no sense for you to look for a worker 5,6,7,8 years later to interview them once they?ve been removed from that situation. Today we have footage of everything, we film the testimonies and I believe that it would speed up investigations, streamline criminal execution and be more effective. The sensitivity that [labor inspectors] have when interviewing is different. A worker within that court environment already feels different too?And the way they ask questions in the court environment, sometimes they are already denying questions, for example: ?Your boss didn?t treat you bad, right?? Then he is already induced to respond in this way. He is also intimidated by the environment. -LA004 Another practical consideration that respondents brought up related to decision making in forced labor exploitation cases is related to characteristics of the defendant, specifically whether they are an individual land owner or part of a larger organization. 154 One labor prosecutor (TP003) noted that in the case of individual farm owners, people can have conflicting feelings about whether severe punishment is needed: ?Of course, it is not an easy decision that begs the possibility of the farmer or employer being arrested in the future.? This labor prosecutor also mentioned that sometimes it is hard to see the defendant as a criminal in cases of forced labor exploitation: ?it?s an ideological issue, so when you have a crime of trafficking, a crime of rape, robbery, we claim that whoever practices this crime is a criminal. In the case of slave labor, we do not title it as a criminal, we title it as an employer or a farmer, so we were unable to overcome this barrier and identify those who practice slave labor as the criminal. External factors Nine respondents raised the issue of political prioritization and funding for forced labor exploitation investigations as a barrier to investigating and prosecuting forced labor exploitation. Respondents explained that different politicians can greatly shift the amount of money allocated to forced labor exploitation operations which is highly disruptive. For example, one federal judge (FJ006) said, ?They closed a lot of offices because of a lack of money, investigations simply aren?t happening.? A federal prosecutor (FP004) simply noted, ?I believe that there is a deficit of labor inspectors for this activity but I cannot say for sure.? ?The Ministry of Economy, it has not been replacing the vacancies of the labor inspectors? What is natural? When a civil servant retires, so as not to cause problems in the public service, there should be a replacement of that civil servant who left. But they say there are budgetary issues because when the person retires, he remains on the payroll for years, even if he dies the pension keeps going to the family?-FP004 A labor inspector (LA004) provided a similar observation: ?When I arrived at the ministry, we had 45 inspectors to serve the entire state of Rond?nia, and today for you to have an idea, we are 15 inspectors.? A labor prosecutor (TP004) indicated that 155 politicians intentionally defund the labor inspectors because they receive money from corporations who are guilty of forced labor exploitation: ?There are the political issues, the pressure because the labor inspectors mess with companies a lot, with pol itical sponsors.? However, no other respondents made this explicit connection. A labor inspector also said that forced labor exploitation is a political topic, and every institution wants to be seen as responsible for leading the fight to eliminate it and do not collaborate: So what we have today instead of an interinstitutional relationship is the institutions that want to play their primary role, without taking into account everything else that was already done before, how public policy was built. I think this is one of the biggest challenges today, and in this political moment now that we are living, it is worse.- LA005 Quantitative findings If case outcomes are associated with underlying characteristics of forced labor exploitation cases, I would expect to find that cases with a stronger evidence and those in which the blameworthiness of the offender is more objectively observable would be associated with more severe civil and criminal justice system outcomes. Specifically, I expect Class 3, Degrading Conditions and Weapons and Surveillance, to be associated with more severe outcomes than the other classes. Based on findings from the qualitative interviews, I also expect the Class 1, Degrading Conditions and Debt Servitude, would be more likely to receive a conviction because respondents noted that debt servitude is one of the easier forms of forced labor exploitation to prove. Below, I describe findings for civil and criminal outcomes. There were some outcomes that I was interested in exploring for which there was not enough variation by class or that had too much missing information to allow for a formal analysis. I review these outcomes first and then present findings from the regression models. 156 Respondents noted that the Dirty List was an important instrument to deter and punish forced labor exploitation. I looked at a cross tab of whether a case was included in the Dirty List by class (Table 24). Not surprisingly, all cases in Class 3 for which information about the Dirty List was available were included in the Dirty List. Class 1 and 2 had a comparable distribution, with the majority being included (97 percent and 95 percent, respectively). Table 24. Cross tab of latent classes and inclusion in dirty list, column frequencies (N=672) Dirty C1: Degrading Conditions and C2: Degrading C3: Degrading Conditions and Weapons Tot List Debt Servitude Conditions and Surveillance al N 3 29 0 o 3.26% 5.19% 0% 32 Y 89 530 21 es 96.74% 94.81% 100% 640 92 559 21 Total 100% 100% 100% 672 Civil For civil court processing, I was interested in examining three outcomes: 1. A nominal variable indicating the outcome of the case (conviction, agreement, or neither). 2. A continuous variable indicating the amount of collective moral damages ordered. 3. A continuous variable indicating the amount of individual moral damages ordered. For outcomes of public civil actions, I excluded all cases that were still in progress (N=53), leaving a nominal three-category outcome variable: full or partial conviction, full or partial agreement, and no conviction or agreement. However, this left a small sample size for this analysis (N=211) that yielded unstable estimates and standard errors in the distal outcome models and the multinomial logistic regression models. Table 25 provides a crosstab of different public civil action outcomes for each class. 157 Table 25. Cross tab of latent classes and public civil action outcomes, column frequencies C3: Degrading Conditions Tot C1: Degrading Conditions and Debt C2: Degrading and Weapons and al Servitude Conditions Surveillance Full/partial 14 62 1 77 conviction 53.85% 34.64% 16.67% Full/partial 9 90 4 103 agreement 34.62% 50.28% 66.67% 2 5 0 7 Not Convicted 7.69% 2.79% 0% 1 22 1 24 In Progress 3.85% 12.29% 16.67% 26 179 6 211 Total 100% 100% 100% While these models would not converge in MPlus, I ran the analyses in Stata as a multinomial logistic regression with conviction as the reference class. I included the latent classes as covariates (class one used as the reference class). This model does not account for measurement error in latent class assignment. As expected, several of the estimates and standard errors were extreme. These results are available in Appendix D, Table 2. When I asked respondents about what the most effective tools to punish forced labor exploitation were, many responded that fines were among the most important. For smaller enterprises, at least, fines were often enough to force the employer to shut down their business or to make it unprofitable to continue. Table 26 shows the results of automatic and manual three-step ML approach to examining distal outcomes (individual and collective moral damages ordered). In the manual three-step models, I controlled for region and whether the labor prosecutor was a member of the inspection team. I also controlled for number of workers rescued, given that this number is often directly correlated with total damages ordered. Individual moral damages typically reflect the amount in backpay that workers are owed while collective moral damages are more often determined 158 based on the gravity of the crime. Thus, I expected to find a stronger difference in collective moral damages by latent class. Table 26. Significant differences between class-specific means for individual moral damages and collective moral damages Approach Class Mean (SE) Class 1 Class 2 45.3965 C1 DC +Debt Servitude (9.293) 0 Manual ML 3-stepb 8.6207 C2 Degrading Conditions (3.791) 36.776*** 0 128.415 Ind. Moral C3 DC+ Weapons (96.701) -83.0187 -119.795 Damages a 88.536 C1 DC+ Debt Servitude (22.257) 0 Automatic ML 3-step 15.706 C2 Degrading Conditions (3.609) 7 2.830*** 47.354 C3 DC+ Weapons (2.325) 41.182* -31.648*** 75.165 C1 DC+ Debt Servitude (11.547) 0 Manual ML 3-stepb 22.338 C2 Degrading Conditions (4.960) 52.8276*** 0 530.883 Coll. Moral C3 DC+ Weapons (121.782) -455.72*** -508.55*** Damages a 121.723 C1 DC+ Debt Servitude (28.087) 0 Automatic ML 3-step 25.333 C2 Degrading Conditions (3.328) 96.39*** 0 462.940 C3 DC+ Weapons (209.442) -341.217 -437.61** a Dependent variables scaled down by a factor of 100. b Model controls for region, number of workers rescued, and whether labor inspector was present at inspection *p<.1 **p<.05 ***p<.01 In the manual three-step model, the Degrading Conditions and Weapons class had a mean of RS$12,841.50 in individuals damages ordered compared to RS$4,539.65 in the Degrading Conditions and Debt Servitude class and RS$862.07 in the Degrading Conditions class. Although the difference in mean between Degrading Conditions and Debt Servitude and Degrading Conditions is lowest of the pairwise combinations in the manual three-step model (RS $3,677.58), it was the only significant difference in the manual three-step model. In the automatic three-step model for the distal outcome individual moral damages, I could not control for the number of workers rescued (or any other covariates). In this model, Degrading 159 Conditions and Debt Servitude had the highest mean individual moral damages ordered (RS$ 8,853.60), and Degrading Conditions class still had the lowest (RS$1,570.60). In the automatic three-step model, the difference between Degrading Conditions and Debt Servitude and Degrading Conditions (RS$ 7,283), as well as the difference between Degrading Conditions and Degrading Conditions and Weapons (- RS$3,164) is significant. The difference between Degrading Conditions and Degrading Conditions and Weapons is likely significant in the automatic model and not the manual model because there are fewer parameters in the automatic model due to the absence of covariates. For collective moral damages in the manual and automatic three-step models, there is a similar pattern with Degrading Conditions and Weapons having the highest mean of collective moral damages and Degrading Conditions having the lowest mean. In the manual three-step method, there is a significant difference between the mean of Degrading Conditions and Debt Servitude and Degrading Conditions (RS$5,282.76). There is also a significant difference between the mean of the Degrading Conditions and Debt Servitude class and the Degrading Conditions and Weapons class (- RS$ 45,572) and Degrading Conditions and Degrading Conditions and Weapons (- RS$ 50,855). In the manual three step approach, most likely class membership is recalculated because of the inclusion of covariates. There were slight changes in class counts and proportions but overall, the pattern from step one persisted (see Appendix D, Table 3). I also conducted a regression of individual and collective moral damages on most likely class membership for comparison. These models do not account for 160 uncertainty in latent class assignment. The results from this model are available in Appendix D Table 4. There are not significant differences in the amount of individual damages ordered for Degrading Conditions and Debt Servitude or Degrading Conditions versus Degrading Conditions and Weapons. For collective moral damages, however, cases in the Degrading Conditions and Debt Servitude class are ordered to pay about $RS 94,009 less in collective moral damages relative to Degrading Conditions and Weapons, and cases in Degrading Conditions are ordered to pay $RS105,690 less in collective moral damages relative to Degrading Conditions and Weapons. Criminal Outcomes A recurrent theme in interviews was that slow processing time for criminal actions resulted in impunity for employers engaged in forced labor exploitation. Respondents noted that the more time that elapsed, the harder it was to identify victims to provide testimony and the less reliable the testimony of victims and investigators was because they were less able to recall specific information from the crime. To explore this in the administrative data, I examined how case processing time varied by latent class. I examined how criminal case outcomes are associated with different latent classes of forced labor exploitation. I looked at two dichotomous outcomes: whether the case was convicted and whether the case was acquitted. These are separate, dichotomous outcomes rather than a single nominal outcome because one case can have multiple outcomes for different defendants. I also examined two continuous outcomes: sentence length in months and case processing length in days. Results from 161 the manual and automatic maximum likelihood methods for distal outcomes in latent class models are shown in Table 27. 162 Table 27. Significant differences between class-specific means for criminal case outcomes: Convicted, acquitted, sentence length, case length Approach Class Mean (SE) Class 1 Class 2 0.416 C1 DC + Debt Servitude (0.158) 0 Manual ML 3-stepa 0.187 C2 Degrading Conditions (0.075) 0.228 0 0.291 Convicted C3 DC + Weapons (0.135) 0.125 -0.103 0.429 C1 DC + Debt Servitude (0.083) 0 Automatic ML 3-step 0.198 C2 Degrading Conditions (0.028) 0.231** 0.282 C3 DC + Weapons (0.103) 0.147 -0.084 0.495 C1 DC + Debt Servitude (0.118) 0 Manual ML 3-stepa 0.522 C2 Degrading Conditions (0.087) -0.027 0.414 Acquitted C3 DC + Weapons (0.144) 0.081 0.109 0.423 C1 DC + Debt Servitude (0.053) 0 Automatic ML 3-step 0.439 C2 Degrading Conditions (0.019) -0.016 0 0.372 C3 DC + Weapons (0.103) 0.051 0.067 85.621 C1 DC + Debt Servitude (28.722) 0 Manual ML 3-stepa 75.479 C2 Degrading Conditions (28.462) 10.1422 Sentence 123.268 Length C3 DC + Weapons (26.605) -37.648* -47.790** (months) 55.508 C1 DC + Debt Servitude (7.11) 0 Automatic ML 3-step 50.834 C2 Degrading Conditions (2.701) 4.674 107.794 C3 DC + Weapons (20.843) -52.286** -56.96*** 1511.451 C1 DC + Debt Servitude (209.207) 0 Manual ML 3-stepa 1221.579 C2 Degrading Conditions (141.654) 289.873* 1651.199 Case Length C3 DC + Weapons (323.062) -139.748 -429.62 (days) 2098.891 C1 DC + Debt Servitude (136.269) 0 Automatic ML 3-step 1714.831 C2 Degrading Conditions (44.915) 384.06* 2457.362 C3 DC + Weapons (277.292) -358.471 -742.531*** a Model controls for criminal court region, police investigation took place, and number of other crimes indicted. *p<.1 **p<.05 ***p<.01 For whether the case was convicted for forced labor exploitation, there are similar patterns in class-specific means in the manual and automatic three step methods, with the Degrading Conditions and Debt Servitude class having the highest 163 proportion of cases that receive a conviction and cases in the Degrading Conditions class having the lowest proportion of cases that receive a conviction. About 42-43 percent of cases in the Degrading Conditions and Debt Servitude class receive a conviction in both methods compared to 28-29 percent of cases in the Degrading Conditions and Weapons and Surveillance class and just 18-19 percent of cases in the Degrading Conditions class. There are no significant differences in class specific means in the model using the manual method but in the model using the automatic method, there is a significant difference between the mean for Degrading Conditions and Debt Servitude and Degrading Conditions. For whether the case was acquitted for forced labor exploitation charges, the class specific means are more similar, though a slightly higher proportion of cases in the Degrading Conditions class are acquitted (52.2 percent) relative to Degrading Conditions and Debt Servitude (49.5 percent) or Degrading Conditions and Weapons and Surveillance (41.4 percent). None of the pairwise differences in means are significant for whether the case was acquitted in the models using the manual or automatic methods. The sentence length outcome is available only for those cases that were convicted. In general, defendants will serve a substituted sentence under community supervision for any case less than four years (48months). For the manual and automatic methods, cases in the Degrading Conditions and Weapons and Surveillance class had the highest mean sentence length (123 months and 108 months, respectively), followed by Degrading Conditions and Debt Servitude (86 and 55 months), then Degrading Conditions (75 and 56 months). The difference in sentence 164 length between the Debt Servitude class and the Degrading Conditions and Weapons and Surveillance class and the Degrading Conditions class and Degrading Conditions and Weapons and Surveillance class was significant using both methods. Case length followed a similar pattern as sentence length, with the Degrading Conditions and Weapons and Surveillance class having a longer mean case (1,651 days in manual model, 2,457 days in automatic) length, followed by Degrading Conditions and Debt Servitude (1,511 days in manual model, 2098 in automatic), and then Degrading Conditions (1,222 days in manual model, 1715 in automatic). The differences in class-specific means are significant only for Degrading Conditions and Degrading Conditions and Debt Servitude (289.873*, p<.1) Estimates from the regression models for criminal case outcomes are available in Appendix D Tables 5-6. 165 Chapter 6: Discussion More than 369,000 people in Brazil are currently trapped in a criminally exploitive work environment. Labor inspectors will respond to situations for which complaints are made, but they must prioritize what they deem to be the most severe cases; that is those that happened recently, involve vulnerable groups, or have evidence of imminent danger. Further, complaints are not made in every case and, consequently, a large proportion will remain undetected. An important step toward ending this crime is to hold business owners who exploit laborers accountable; findings from this dissertation call into question the effectiveness and necessity of criminal justice processing to do so. In this chapter, I summarize the major findings from the analyses in chapter five. I then describe the limitations of the study followed by a discussion of how these findings fit within the context of criminological literature more broadly. I also discuss implications for policy and practice. 6.1 Overview of findings To date, there has been limited information on the characteristics of forced labor exploitation; leading to ambiguous conceptualization and measurement of the crime. Further, available information on characteristics of forced labor exploitation is based on small samples and findings are not generalizable for policy and practice. Using administrative data gathered from inspection reports and case processing records nationwide over a ten-year period, I identified three typologies of forced labor exploitation cases in Brazil: degrading conditions and debt servitude, degrading conditions, and degrading conditions and weapons and surveillance. In line with findings from the interviews, the greatest share (83.8 percent) of cases is predicted to 166 fall into the degrading conditions class. While all three classes have a high probability of including degrading conditions, the degrading conditions class is characterized by having a low likelihood of any other forms of control. Prior research on forced labor exploitation in Brazil has found that degrading conditions are the defining feature of the crime. For example, two prior studies cited cramped and unsanitary living quarters and another cited a lack of access to medical care (Patric?a Trindade Maranh?o Costa, 2009; Walk Free Foundation & International Organization for Migration, 2017). In line with findings from the latent class analysis, interview respondents reported that degrading conditions are highly prevalent in forced labor exploitation cases, but that a combination of degrading conditions is needed to characterize forced labor exploitation. Thus, while the administrative data show a high likelihood of only one characteristic, degrading conditions, in reality a combination of dangerous and unhealthy factors are likely present. Participants also described degrading conditions as one of the elements that could be proven through evidence such as photographs and videos. This is important because a primary barrier to prosecution and sentencing, according to interview participants, is that it is difficult to get victim testimony? either because victims do not want to testify or they cannot be located. In the analysis in objective two, I identified a few covariates for which the odds of belonging to the degrading conditions class were lower than for the degrading conditions and weapons and surveillance class (number on inspection team, number of workers rescued during the inspection, report identified other crime: danger to life or health of others, relocating workers under false pretenses, and other crimes). The 167 presence of crimes including danger to life or health of others, relocating workers under false pretenses, and other crimes all were associated with a lower odds of belonging to the degrading conditions class relative to degrading conditions and weapons and surveillance. Cases in rural sectors had higher odds of belonging to the degrading conditions class relative to degrading conditions and weapons and surveillance. These findings are striking because in the absence of other characteristics of forced labor exploitation or other crimes the question that logically arises is how employers manage to keep workers in this situation in degrading conditions cases. Findings from the interviews shed some light on this, offering two explanations, both of which go hand in hand with descriptions of rural forced labor exploitation in general. First, it is possible that other forms of coercion that could not be measured in the data were present such as social and linguistic isolation. Second, it is possible that the workers did not feel they had any other prospects for earning money and thus did not attempt to leave the situation. The latter explanation is one that judges, during the interviews, described mixed understandings of: some would not convict in a situation where a worker voluntary submitted themselves to degrading conditions and others said the worker?s opinion did not matter. This blurry line of consent is one that practitioners struggle with when it comes to all types of trafficking in persons, and is commonly debated in the sex trafficking world in discussions about legalizing sex work?whether it is possible to consent to work in a situation that on the whole is exploitive. In Brazil, ultimately, the law is clear that it is illegal to submit a worker to degrading conditions. That workers ?voluntarily? stay in these situations is a strong indictment of the lack of economic opportunity for those 168 living in poverty, as described by one of the labor auditors, but this cannot be fixed through strengthened court processing of landowners. The second greatest share (13.2 percent) of cases are predicted to belong to the degrading conditions and debt servitude class. Debt servitude is the most common form of control used over workers in situations of forced labor exploitation worldwide, particularly in rural areas (Internationales Arbeitsamt, 2009). There are a few factors that increase vulnerability for debt servitude. Weak labor protections in addition to remote jobsite locations make it difficult for agricultural workers to organize for fair treatment. In some cases, entire families are forced to work in exchange for inadequate food and housing (International Labor Organization, 2014). Families are then unable to generate enough money to pay back imposed debts from recruiters and employers, and this debt is passed down through generations. Employers also take advantage of residents in rural areas who lack employment and educational opportunities (International Labor Organization, 2014). Unlike the degrading conditions class, the debt servitude class includes a high probability of other elements of forced labor exploitation such as exhausting work, restricting transportation or movement, withholding documents, and recruitment fraud. It is not surprising that the Debt Servitude class is also the class with the highest probability of including recruitment fraud (37.7 percent). Prior research has found that employers often withhold personal documents as a form of collateral for the workers? debts (International Labor Office, 2012). The analysis in objective two showed that none of the included covariates were significant predictors of membership in the debt servitude class relative to the weapons and surveillance class. However, both the 169 degrading conditions and debt servitude and degrading conditions and weapons and surveillance classes included a variety of characteristics of forced labor exploitation and were overall more similar to one another than to the degrading conditions class. Only three percent of cases are predicted to belong to the degrading conditions and weapons and surveillance class, which is characterized by a high likelihood of ostensive surveillance and the use of weapons. The use of more physical forms of control to prevent workers from leaving is documented in other studies, though similarly rare. For example, case studies with survivors of forced labor exploitation have documented instances where employers torture and abuse workers who attempt to escape (International Labor Organization, 2009a; United Nations Office on Drugs and Crime, 2008). Similarly a few interview respondents described situations where victims were submitted to torture and physical abuse. Little is known about how different characteristics of forced labor exploitation are associated with sentencing outcomes; in large part because of low overall levels of prosecution for forced labor exploitation(Kappelhoff, 2008; OSCE et al., 2008). In this study, I assessed how typologies of forced labor exploitation are associated with different outcomes in the criminal and civil court systems. There are several interesting findings related to case outcomes. For collective moral damages, there is a stepwise pattern as would be expected where the degrading conditions and debt servitude class is associated with higher collective moral damages than the degrading conditions class, and the degrading conditions and weapons and surveillance class is associated with higher collective moral damages than the debt servitude class. For individual moral damages, the degrading conditions and debt 170 servitude class has the highest individual moral damages, followed by the degrading conditions and weapons and surveillance class, and then the degrading conditions class. Three main findings stand out related to criminal court outcomes. First, while on the whole more extreme cases (such as those that characterize cases in the degrading conditions and weapons and surveillance class) were rare, interview respondents said that many judges and even prosecutors have a hard time convicting or bringing a case forward in the absence of physical control. I did not, however, observe an increased likelihood of conviction for the weapons and surveillance class relative to the other classes in the administrative data as would be expected based on interview findings. Weapons and surveillance cases had significantly longer case lengths relative to degrading conditions cases, however. Factors related to increased case processing time such as inability to locate witnesses, investigators? deteriorating memory of the incident, and even death of defendants or expiration of the statute of limitation may be contributing to insignificant conviction rates in weapons and surveillance cases. It could also be that defendants in weapons and surveillance cases are more likely to hire attorneys that can either have their case dropped or continue to delay the case through the use of appeals. Second, weapons and surveillance cases are associated with a longer sentence length than cases in the other two classes which suggests that when the cases are convicted they are more severely punished. The anti-slave labor and trafficking in persons clinic suggested assuming that any sentence length under four years (48 months) received a substituted sentence, or community supervision. The average 171 sentence length for degrading conditions and debt servitude and degrading conditions cases was just over four years (55 and 50 months), which indicates that the majority of defendants served no or very little prison time. The average sentence length for degrading conditions and weapons and surveillance cases, however, was nine years. Thus, if serving time in prison is used as a proxy for accountability then weapons and surveillance cases are more likely to be held accountable. Third, an in line with interview findings, I found that degrading conditions and debt servitude cases are more likely to receive a conviction relative to degrading conditions cases. Interview respondents said that debt servitude is perhaps the easiest element of forced labor exploitation to prove because employers often keep physical workbooks tracking the debts. The availability of physical evidence that is not reliant on victim testimony is likely a contributing factor to the elevated rates of conviction in debt servitude cases. 6.2 Limitations A primary limitation to this study is that I was able to develop typologies and ask interview respondents about only those forced labor exploitation cases known to the court system. In the hypothetical spectrum of cases based on their severity, I imagine that this study includes information on cases in the middle, lacking both the most and least severe cases. Cases that remain undetected include those for which no complaints are ever made. This is likely to be the case in situations where workers die either because of hazardous working conditions or other job-related injuries or are killed by their employer if they attempt to leave, as was the case for Perreira?s 172 colleagues 25 years ago. Cases that involve immigrants, indigenous people, and children are also likely to remain undetected because these groups are less likely to be familiar with or use the complaint process. It is not only these most severe cases that remain undetected, however. Labor inspectors noted that they prioritize the most severe of their complaints for inspection. Thus, less severe complaints are pushed to the bottom of the pile, so to speak, and may ultimately not be investigated, or may not be investigated in time to identify the crime. While it is impossible to research an undetected incident of a crime, analysis of complaints themselves could yield important information for better understanding forced labor exploitation. In the United States, for example, data from the national human trafficking hotline are a valuable source of information about the nature of human trafficking in the country (Polaris, 2017, 2020). There are also limitations inherent to using administrative data. First, administrative data are not collected with research in mind and in the case of this dataset, it was often difficult to distinguish between missing data and an indicator of ?not present? which rendered many of the potential variables unusable. Further, while labor inspection reports provided rich data on case characteristics, the same information is not available for cases initiated through other means, such as through police investigations or referred directly from the Ministry of Labor. Thus, my analyses are based primarily on cases for which a labor inspection report was available. This means my findings about the typologies of forced labor exploitation are generalizable only to a subset of all known forced labor exploitation cases. As explained above, my analysis likely excludes the least and most severe incidents of 173 forced labor exploitation. Additional research is needed on the types of forced labor exploitation cases that do not become known through labor inspectors. In particular, more information is needed on cases investigated by federal police that are never investigated by labor inspectors. My guess is that these cases make take place alongside other crimes that police are already investigating. They may also be more likely to occur in urban sectors, or areas that draw more public attention leading to police tips. Another limitation of the data in this study is that case outcomes in the criminal system were delivered at the individual defendant level, but defendant information was frequently missing and there was no way to link different case outcomes to specific defendants. An analysis of defendant level characteristics that predict case outcomes, were it possible, would represent a huge contribution to the field. Researchers who have established close partnerships with agencies collecting crime data can work with these agencies to suggest changes to data collection forms and procedures. This way, though the data are not being collected for research purposes, it can be collected in a way that lends well to future research. There are a variety of examples of this type of collaboration through community based participatory research that have demonstrated benefits for researchers and the community-based organization (e.g., Dhungel et al., 2019; Rhodes et al., 2012). Another limitation to this study is that I was not able to interview the individuals directly involved in forced labor exploitation cases, the victims and perpetrators of the crime. In the interviews, respondents?particularly judges-- mentioned that the victims? perception of the case was an important factor in their 174 decision making. Similarly, some respondents mentioned that employers may not realize they were committing a crime by submitting workers to degrading conditions. It is therefore critical to better understand, from victims and perpetrators themselves, why they did or did not consider themselves to be involved a situation of forced labor exploitation. This type of information can inform prevention campaigns, policy making, and even judicial and prosecutorial decision making. There is also a methodological limitation to this study related to the use of latent class analysis. I based latent class membership on the class with the highest probability of assignment for each case. This introduces some degree of uncertainty to the analysis and may have attenuated the relationship between typologies of forced labor exploitation and sentencing outcomes. To reduce the degree of uncertainty, I controlled for measurement error in the analysis using the maximum likelihood manual and automatic three-step approaches. 6.3 Implications and recommendations 6.3.1 Implications for theory In line with other research on decision making among court actors and investigators, respondents in this study were focused on reducing uncertainty in their decision making (Spohn, 2014; D. Steffensmeier et al., 1998). Respondents described several factors that increase uncertainty in forced labor exploitation cases and serve as barriers to investigation, prosecution, and judgment. First, respondents?particularly federal judges and prosecutors-- expressed frustration with the subjectivity of Article 149. Prior research on focal concerns theory has similarly found that a barrier for prosecutors and law enforcement officials is uncertainty over the elements needed to 175 prove evidence of human trafficking (A. Farrell et al., 2014b). In the United States, however, much of the uncertainty lies in whether all three elements of the law (force, fraud, coercion) must be present. In this study, Brazilian practitioners seemed clear that all elements in their law did not need to be present but expressed uncertainty about the definition of the most common elements of forced labor exploitation? degrading conditions and the exhausting workday. As discussed earlier, most cases were characterized only by degrading conditions; if this element is unclear then a majority of cases are at risk for not being prosecuted or receiving a conviction. Federal judges further mentioned that they never receive any training on handling forced labor exploitation cases. In fact, labor inspectors are the only group that reported receiving specialized training related to forced labor exploitation. Prosecutors in the federal and criminal systems mentioned that they may have opportunities for training made available to them, but it is not mandatory. Comprehensive, regular training, for court actors in both the labor and criminal systems can help to reduce uncertainty about what constitutes a forced labor exploitation case. Additionally, the development of a centralized case study database may help prosecutors and judges better operationalize the concepts with real life examples, thus reducing uncertainty in decision making. These types of databases exist in the United States and are often used in conjunction with practitioner training (Motivans & Snyder, 2015; United States Department of State, n.d.). Improved training and availability of a case law database is unlikely to reduce all uncertainty in forced labor exploitation cases, however. If it could, researchers would not have identified issues related to uncertainty in the U.S.-based studies mentioned above. 176 Prior research on focal concerns theory finds that prosecutors are likely to file charges when they believe they have the best chance of securing a conviction. Since prosecutors have infrequent experience with forced labor exploitation cases, they may still be hesitant to bring cases forward, especially if they have a high case load and other cases they perceive to be more prosecutable (Albonetti, 1986; Steffensmeier et al., 1998). In line with prior research on focal concerns theory in human trafficking cases, respondents in this study said that a major consideration in prosecutors? and judges? decision making is the testimony of the victim. Reliability of the victim and the ability to obtain a strong victim statement may be particularly important when prosecutors and judges are uncertain about other elements of the case (Hawkins, 1981). Research looking at prosecutorial and judicial decision making in U.S. cases finds that in the absence of certainty about the legal aspects of a case, prosecutors and judges may use perceptions about victim respectability to determine whether the victim testimony is an honest assessment of the situation. Practitioners may also rely on irrelevant factors like race, class, or gender to assess victim statements, especially when victim credibility is low (Spohn, 2014). Interestingly, conversations about victim credibility did not come up in the interviews for this study. Discussions about the need for victim testimony focused more on the barriers to obtaining this testimony in the first place, such as barriers to locating victims and victims? reluctance to identify themselves as victims. Issues around credibility may be less important in the context of forced labor exploitation because there is less societal stigma around the types of work involved in 177 forced labor exploitation relative to sex work, which is the predominant form of exploitation in US human trafficking cases. Additionally, most victims of forced labor exploitation are men. Men engaged in societally acceptable forms of work may be perceived as more credible than women engaged in sex work. Thus, many of the factors which are detrimental to victim credibility in U.S. human trafficking cases are potentially not as great of a concern for Brazilian forced labor exploitation cases. Alternatively, the inability to obtain victim testimony in the first place may preclude the need for any informal judgments about victim credibility. If this is the case, more research is needed to understand the role of perceptions around victim credibility in cases where testimony is available. Interview respondents noted other difficulties related to victim testimony, however, such as victims being afraid to testify and losing contact with victims before the case is tried. One potentially easy way to rectify this issue is to draw on lessons learned from longitudinal survey research (Lynn, 2009). Investigators should immediately request victims? contact information as well as the contact information of close friends and family members. Court staff should follow up with victims or their contacts on a regular basis to update the information and to give them a status update on their case. This way, if the case is not heard for years, the courts still have up to date contact information on the victim. Another option to encourage victim testimony that is a practice in the United States is to tie federal benefits for survivors to their participation in the trial against the individual who exploited them; however, victim advocates do not condone this policy as it can have damaging mental health repercussions for survivors. 178 Another barrier to obtaining victim testimony is that many workers do not perceive themselves to be victims. While some respondents said a victim?s opinion about whether they were exploited is irrelevant, others noted this is an important consideration. If investigators or prosecutors are not certain about the case, victims being unwilling to say they were in an exploitive situation may prevent them from bringing the case forward. Further, nearly all federal judges agreed that victim testimony was critical evidence to be presented in the case. With a few exceptions, such as the presence of a debt workbook, federal judges said a case could not be decided without victim testimony. An additional way that respondents sought to reduce uncertainty in forced labor exploitation cases was to only investigate, prosecute, or convict the most extreme cases?such as those that included physical violence or cases in which workers were physically prohibited from leaving the worksite. As demonstrated in the analysis of administrative data, however, cases with these elements are rare. Further, there was no evidence that cases in the weapons and surveillance class- which were those cases that included physical violence and ostensive surveillance- were more likely to receive a conviction relative to cases in other classes. This suggests that even when these more ?objective? elements are present, judges have a hard time convicting. Thus, while prosecutors may think that the more severe elements reduce uncertainty in receiving a conviction, it may not actually be the case. This finding on its own poses a challenge to focal concerns theory, which suggests that safety of the community is a primary factor in decision making for judges and prosecutors. If this were the case, elements that post an objective and severe threat, like the presence of 179 weapons, should be associated with greater likelihood of conviction. It may be the case that judges are not confident the individual defendant(s) are the individuals ultimately accountable (discussed further below) or that proving these elements is more difficult and evidence is insufficient. When convictions are made, cases in the weapons and surveillance class receive the longest sentences. In particular, the average sentence length for cases in the weapons and surveillance class receive sentences of longer than four years, on average. A sentence length of four years or longer guarantees that a substituted sentence was not offered, thus this can be a proxy, albeit an imperfect one, for whether the defendants served actual prison time. If that is the case, it appears that when judges are confident that more severe elements of the crime are present, they hand down longer sentences, in line with what focal concerns theory would suggest. This has two implications for theory. First, focal concerns theory should take into account not only decision makers? considerations about factors that will reduce uncertainty, but also what a given decisionmaker perceives to be important for the other decisionmakers in the process. For example, prosecutors perceive that cases with objective characteristics like violence are more likely to receive a conviction because they think judges assign disproportionate weight to these factors. However, many judges stated that these elements are not needed to characterize forced labor exploitation. Thus, focal concerns for prosecutors may include their assessment of the relevant focal concerns of judges. Second, the sentencing decision may be a greater reflection on the weight of legal evidence while the associated sentencing length may be a greater reflection of judges? focal concerns. 180 A final theoretical implication from this study relates to the conceptualization around blameworthiness of the offender. Respondents disagreed about whether a local farm manager, or the recruiter (gato) should be held accountable for forced labor exploitation if they were only a middleman in a larger organizational scheme. It was interesting to see the dichotomy between participants responses to questions about the characteristics of forced labor exploitation in which they spoke ardently about the inhuman working conditions coupled with their empathy for potential perpetrators of this crime whom they referred to not as criminals but as farmers or bosses. All respondent types overwhelmingly agreed that the dirty list was an effective and appropriate punishment for forced labor exploitation, but there was more disagreement around criminal convictions. This suggests that respondents considered the business or the company as a whole to be worthy of sanctions but they did not necessarily view the employees of that business, operating for the company?s benefit, to be guilty. As Fisse and Braithwaite (1983) note, ?corporateness obscures blameworthiness.? To my knowledge, applications of focal concerns theory to businesses (rather than individuals) is limited (Homer & Higgins, 2020); however this is an area merits additional research and theoretical development. In a situation where the manager found in charge of the operation is not considered responsible for worker exploitation, there may not be anyone to charge with the crimes. In sex trafficking cases in the United States, prosecutors may make a conviction even when they cannot identify a third party such as a brothel owner who is in charge of the operation. In those cases, convictions focus on the consumers of exploited sex work, which is a viable option because sex work is illegal in the United 181 States. It is not feasible to hold consumers of exploited labor responsible, however, because nearly every member of society purchases goods produced with exploited labor. Respondents unanimously agreed, however, that better intelligence work is needed to investigate the full supply chain in these cases, and at least hold larger corporations accountable for contracting out for labor at prices that can only be attained through worker exploitation. A shift toward this type of investigation strategy would require treating forced labor exploitation cases more similar to white collar crime cases, as discussed in the section below. 6.3.2 Implications for research, policy, and practice Overall, this study yields recommendations in several areas to further research and improve policy and practice. First, while most research to date has focused on the ways to improve criminal court processing for forced labor exploitation and related cases, this study suggests criminal court processing may not achieve the goals of holding perpetrators accountable, providing reparations to victims, and preventing reoffending. Perhaps one of the most striking findings from the interviews related to case outcomes is that participants, including federal judges and prosecutors, generally agreed that processing cases in the criminal court system does little to prevent or stop forced labor exploitation. Outcomes that hurt employers? bottom line such as inclusion on the dirty list or large fines were described as the most effective instruments to combat forced labor exploitation. The dirty list can hold larger corporations accountable which is critical because these larger companies can often avoid criminal punishment or afford to pay fines handed down in labor court. Respondents also described the criminal justice system as slow, and unlikely to result 182 in a serious penalty; these claims are supported by the analyses of administrative data. Interestingly, while prior research on forced labor exploitation or human trafficking yields few solutions in terms of combating these challenges, there are substantial lessons that can learned from the study of white-collar crime. The similarities between forced labor exploitation and white-collar crime are abundant. Although the definition of white collar crime can be ambiguous, many criminologists cite Sutherland?s definition ?a violation of criminal law by a person of the upper socioeconomic class in the course of his occupational activities? (Rorie et al., 2018; Sutherland, 1945). Indeed, most forced labor exploitation cases are committed, ultimately, by a person of upper socioeconomic class in order to gain a profit. Both crime types face similar challenges: they are considered among the most detrimental to society, lack sufficient research, there is limited data available to study the crimes, they are plagued by uncertainty about what constitutes criminal behavior, and both may be processed in civil as well as criminal courts (Rorie et al., 2018). In this sense, forced labor exploitation can easily be conceptualized as a subtype of white-collar crime. This conceptualization confers several interesting policy implications: whether criminal court processing of forced labor exploitation cases be a priority; conceptualization and profile of offenders in forced labor exploitation cases; and re-conceptualizing lower-level employees as victims rather than holding them accountable for worker exploitation. First, some scholars, most notably John Braithwaite, have argued that white collar crime should be handled through a restorative justice framework (J. Braithwaite, 2009; John Braithwaite, 2018). Restorative justice is a model that 183 suggests that criminal justice system outcomes are not always the best ways to respond to crimes, in terms of efficiency, victim preference, or ability to hold perpetrators accountable. Instead, restorative justice models focus on non-punitive outcomes. In the case of forced labor exploitation this could include acknowledging their culpability in worker exploitation, allowing victims to confront the perpetrator to make a statement (if desired), and payment of reparations. Studies in the United States find that many survivors of human trafficking, particularly labor trafficking survivors, would prefer the use of restorative justice to criminal court processing in their cases as it would provide a relatively more direct way for them to receive backpay and monetary reparations for their work (Hussemann et al., 2018; Yu et al., 2018). One study in United States found that victims of labor trafficking were more likely to say they would want reparations than were victims of sex trafficking (Hussemann et al., 2018). Brazil has a strong foundation for a model of restorative justice for forced labor exploitation cases because the payment of reparations is already a practice in the labor court system via individual and collective moral damages. The use of a restorative justice model in lieu of traditional criminal justice system processing could create many efficiencies, including freeing time of investigators, federal police, and prosecutors who may otherwise spend months or years trying to build a case. The courts could also use restorative justice as an opportunity to provide thorough education to employers and survivors around worker?s rights and acceptable working conditions. Restorative justice could be used to hold companies responsible, rather than individuals, who respondents were hesitant to criminally convict. One drawback to restorative justice is that victims are a key 184 component of the mediation process. If victims are difficult to find or unwilling to participate, there would likely be the same difficulties clearing case loads with a restorative justice approach as are currently seen in the criminal justice system processing. Thus, a second approach from corporate crime that could be used as an alternative to restorative justice is to focus exclusively on the name and shame punishment approach. Naming and shaming is an approach commonly use in corporate crime to increase transparency and accountability (John Braithwaite & Drahos, 2002; Simpson et al., 2013). Given respondents? support for the use of the dirty list, which is effective in part because of the naming and shaming of businesses involved in forced labor exploitation, this type of approach could guide how the dirty list is implemented and the implementation of similar naming and shaming approaches in other countries. Braithwaite and Drahos (2002) described naming and shaming as a two-step process in which the corporation is named and then the corporation fixes the issue that led to the incident through their internal processes. This approach is a form of reintegrative shaming, in which the behavior or crime is pointed out and described as bad, but the individuals within the company work with the accountable person in a respectful, reintegrative way to solve the problem. Naming and shaming is an effective deterrent of future crime for a few reasons. First, publicizing sanctions can have a general deterrent effect because other corporations realize the direct consequences of the crime. Second, publicizing sanctions can damage a company?s reputation; scholars point out that fear of reputational damage is a stronger incentive than fear of financial sanctions (May, 185 2005). Reputational damage has been documented as an effect of Brazil?s dirty list: nearly all companies experience a negative market reaction after being published. For example, shares for the clothing company Zara dropped by about four percent after they were added to the list (Kelly, 2013). Reputational damage after public sanctions have also been documented in other forms of corporate crime (e.g., Armour et al., 2017; Gunningham et al., 2004). Given that forced labor exploitation is nearly universally acknowledged to be an unacceptable practice, the public and advocacy groups may improve the effectiveness of naming and shaming approaches by pressuring companies to comply. This approach may not work as well for smaller businesses, or businesses that do not sell directly to consumers, such as charcoal producers or exporters of wood, for example. In these cases, the addition of their name to the dirty list would result in minimal, if any, reputational damage because the products are sold to suppliers prior to being marketed and sold to consumers. If naming and shaming approaches could hold all companies responsible for the behaviors of their entire supply chains, these approaches would likely be the most impactful. A third takeaway from corporate crime that can be applied to forced labor exploitation is to reframe our understanding of perpetrators of forced labor exploitation. Respondents said that they would prefer that cases were investigated more extensively, including identifying the top of the supply chain rather than holding the middlemen or lower levels employees accountable. A similar debate about whether to hold lower level employees accountable is cited in the white collar crime literature (Rorie et al., 2018). Accountability in both forced labor exploitation 186 and white-collar crime cases is an elusive concept. One way to improve investigations in forced labor exploitation may be to clearly define parameters for who is considered culpable and then to develop a robust profile of the people who tend to be perpetrators of forced labor exploitation. White collar crime literature finds that offenders have different criminal careers than individuals who engage in street crime- they may begin offending later in life and continue to reoffend, with only a slight drop off. There is incredibly limited research on perpetrators of forced labor exploitation, so beginning with a framework for white collar perpetrators of crime offers a useful starting point for future research. Part of improving investigations and holding the right people accountable would require more extensive investigations of the supply chain in forced labor exploitation cases. Focusing on the top of the supply chain rather than individual landowners or managers can alleviate some of the ambiguity associated with accountability while conferring other benefits. As labor inspectors pointed out, many of the individuals they identify as the responsible party during investigations are mid-level employees, not the business owners. Further, these individuals are often victims themselves and processing them in the criminal justice system only serves to punish them for their own victimization. The United Nations has even established the non-punishment principle as part of the Palermo Protocol, which says that victims of forced labor exploitation (and other forms of human trafficking) should not be punished for crimes they commit during their exploitation (The Interagency Coordination Group Against Trafficking in Persons, 2020; Working group on trafficking in persons, 2010). In line with recommendations from the United Nations, Brazil and other nations or can 187 implement the non-punishment principle by adding in victim protections to existing laws, providing training for investigators to identify victims, and providing training to court actors so that the principle is applied equitably for all individuals. Many countries struggle to implement the non-punishment principle, including the United States where no states were identified as implementing all recommended aspects of this principle (Marsh et al., 2019). Further, improved information sharing and dedication of resources to investigating the companies ultimately benefitting from contracting out exploited labor could have more significant impact on reducing the crime than simply arresting a lower-level farm owner. One example of this practice is in place in Brazil via the state-wide labor inspection group for Sao Paolo. Unlike labor inspectors in the national group, the state group is free to extend their investigations as long as they needed, even outside the two-week timeline. The labor inspector from this group said tracking the supply chain is how they were able to find out that large corporates such as Nestle and Danon were connected to local work exploitation. Other respondents pointed out that investigating white collar crime would be an easy starting point for more advanced intelligence work. As several respondents noted, forced labor exploitation almost always includes some form of white-collar crime to embezzle or launder money out of the country, to set up layers of contracting organization to distance larger companies from the exploitation of workers. Improving investigations of white-collar crimes can vastly improve investigations of forced labor exploitation. To conclude this dissertation, I provide key takeaways for select stakeholder groups. Key takeaways for labor inspectors, federal police, prosecutors, and judges 188 This study has important takeaways for practitioners that work on forced labor exploitation cases. 1. All practitioners, including inspectors, police, prosecutors, and judges in the labor and criminal systems, would benefit from regular training around forced labor exploitation. Judges and prosecutors noted that they do not have formal training around handling forced labor exploitation cases, and importantly, that they would like to receive formal training. They are instead required to rely on their experience; however, since forced labor exploitation cases are rare, they have little professional experience handling them. Further, the lived experiences of judges and prosecutors is often distinct from that of potential victims and perpetrators of forced labor exploitation. This means that practitioners may have to rely on stereotypes and other non-legal factors affecting individuals involved in forced labor exploitation. Labor inspectors receive thorough training specific to forced labor exploitation. In a few isolated cases, prosecutors receive special training for working with labor inspectors. This model of combined training should be expanded so that judges, prosecutors, labor inspectors, and even federal police are in the same training and sharing experiences. Trainings should emphasize that cases that include elements like weapons, physical violence, restriction of freedom, and ostensive surveillance are rare- making up just three percent of cases overall. This means prosecutors and judges should not be afraid to conceptualize cases as forced labor exploitation if they do not contain these elements. 2. Different practitioner types should debrief after a forced labor exploitation is closed, and at regular intervals each year to discuss the case outcome and contributing factors to the case outcome. Currently, there is no feedback loop about case processing for forced labor exploitation cases. This contributes to issues in improving investigation, evidence gathering, and prosecution strategies. Debriefs about specific cases or about a given set of cases processed during a given period can improve the performance and collaboration of all actors. Debriefs can also improve the knowledge base for all practitioners as they will hear feedback about cases they may not have been involved with. Meeting notes from these debriefs can also be archived for practitioners to read in the future who want more background on forced labor exploitation cases. 3. Labor inspectors and federal police should work with community partners to develop a greater understanding of urban forced labor exploitation and consider whether it is worthwhile to devote personnel to investigating urban forced labor. Despite the growth in urban forced labor exploitation, most investigative attention is given to rural cases. Labor inspectors, understandably, prioritize the most severe cases and those that include the most vulnerable victims. However, having a greater 189 knowledge of victim profiles and labor exploitation in the urban sectors can help inspectors to make informed decisions when prioritizing cases. Findings from this study show that urban forced labor exploitation cases look different, and often include more diverse forms of control and coercion, while rural cases are more likely to be characterized by degrading conditions only. This means that the types of evidence needed for urban cases may also be distinct. It may be beneficial for additional labor inspectors to be dedicated to a group that specializes in urban forced labor exploitation. These inspectors should work closely with federal police, as respondents noted that often inspecting urban cases can be dangerous, as forced labor exploitation may co-occur with street crimes like drug trafficking, posing a greater threat to the investigation teams. 4. Develop an infrastructure to maintain contact with victims of forced labor exploitation after they are rescued from the exploitive situation. This may need to include incentives for victims to maintain contact with courts. The current system for rescuing victims and providing emergency shelter and linkages to social supports like unemployment insurance is incredibly effective. However, victim services and follow up should not stop after victims are signed up for these services. Victim follow up could confer two primary benefits prevent revictimization and increase the likelihood of obtaining victim testimony during trials. Labor inspectors can help to prevent revictimization by connecting victims with free job training services and linkages to well paying jobs. Job training should include education about forced labor exploitation so that individuals are empowered to avoid exploitive situations in the future. In order to maintain contact with victims until the trial for their case, courts should request contact information of victims as well as individuals they live with or maintain contact with. Twice per year, courts should call to confirm the contact information, and provide updates about the status of the case. To encourage participation, courts should consider offering small financial incentives for confirming or updating their contact information, if there are resources to do so. These check ins can also serve to identify whether the victims need job assistance and link to services if appropriate, to prevent revictimization. 5. Forced labor exploitation cases have several elements of white-collar crime, and investigations into the white-collar crimes can help in identifying the supply chain responsible for worker exploitation. Several interview respondents noted that forced labor exploitation cases almost always contain elements we think of as white-collar crime such as money laundering, embezzlement, and setting up shell corporations to contract out to farmers so that exploitive labor cannot be traced to a large corporation. Labor inspectors typically only have two weeks to inspect all of the cases in a given geographic area, thus they do not have the time or resources to follow the trail and link exploited labor to the top of the supply chain, but changes to the investigative process such as training police to do so 190 could be useful. Additionally, viewing forced labor exploitation as white-collar crime may reframe who is considered accountable in these cases (i.e. corporations rather than the individual landowner or manager who works on site). Key takeaways for policy makers This study has important takeaways for policymakers that shape policy that affects the inspection and processing of forced labor exploitation cases. 1. Fund and implement more state-level labor inspection teams that follow the model in place in Sao Paolo that allows inspectors to investigate cases after the two-week initial inspection. This will allow inspectors to investigate the full supply chain of forced labor exploitation cases. Two weeks is not sufficient to gather evidence about the full supply chain that results in forced labor exploitation in a given case. This results in insufficient evidence for criminal processing in some cases. Further, it may lead to farm managers and other individuals being held accountable for a crime when they are victims themselves. Allowing labor inspectors the resource to gather intelligence or to work with other investigative authorities, like federal police, to investigate the root cause of forced labor exploitation would be a more effective way to combat and prevent forced labor exploitation. 2. Work with community partners and survivors of forced labor exploitation to identify and implement alternatives to criminal justice processing, when appropriate. In some cases, criminal justice system processing may not be the ideal way to process cases, or victims may benefit from a complementary approach. For example, if there is no clear accountable party, the victim is unwilling to participate in the trial, or prosecutors think the case will be substantially delayed. Community partners and survivors of forced labor exploitation should be involved in developing specific alternatives, but additional responses may include restorative justice or restitution approaches. 3. Policy makers outside of Brazil should consider the feasibility of implementing an approach similar to the dirty list. All respondents spoke highly of the dirty list, mentioning it is an effective way to combat forced labor exploitation. Further, corporations included on the list have demonstrated significant financial losses after being added. Other countries can implement similar public lists of employers who are known to engage in forced labor exploitation, and accompany inclusion on this list with financial disincentives, as is done in Brazil. Implementation of this list should not be a one size fits all approach, however. Small businesses should have less severe consequences if they are added to 191 such a list to promote reintegration, allowing the business to bring their standards up to compliance with labor regulations. Large disincentives or bad public fallout resulting from inclusion on the list could completely bankrupt a small business, leaving more workers without a job and vulnerable to future exploitation. 4. More resources should be dedicated to investigating forced labor exploitation, and the number of labor inspectors should remain constant or increase each year. A key barrier to combating forced labor exploitation is that funding for labor inspectors is dependent on political priorities. Although forced labor exploitation cases are rare, and thus may be considered a lower priority than other public safety threats, the consequences of this crime are severe. Further, other crimes are commonly identified alongside forced labor exploitation and, in some cases, tied to a larger network of criminal activity. Labor inspectors are critical to identifying this crime and rescuing workers for exploitation and the resources available to them should not waiver with changing political leaders. 5. Strengthening labor protections for domestic workers is essential to combat forced labor exploitation. Respondents noted that forced labor exploitation in domestic work is common, yet this is one of the hardest sectors to inspect. Policy makers in Brazil and in other countries can protect these workers by implementing laws that protect workers from discrimination and harassment, build in breaks during the day, and guarantee the right to time off and sick leave. Key takeaways for researchers This study yields several important takeaways for researchers, including directions for future research. 1. In studies of decision making that use a focal concerns approach, there may be a discrepancy between an individual?s focal concerns and what others think are the focal concerns of that individual. When identifying the different factors that affect their own decision making, federal police and prosecutors brought up factors that they thought were important to decision makers down the line (i.e. prosecutors and judges). For example, prosecutors think that judges are more likely to convict in the presence of physical violence, restriction of freedom, or ostensive surveillance, but the analysis of the administrative data finds that cases with those elements are actually no more likely to receive a conviction. when multiple people are involved in making a decision during a case, such as police, then prosecutors, then judges, focal concerns theory should take into account an individual?s assessment of the focal concerns of the practitioner down the line. 192 2. Researchers should identify whether the categories of forced labor exploitation identified in Brazil apply to other countries. Developing a more nuanced understanding of forced labor exploitation can help to improve prevention, investigation, and court processing. Further, better understanding how forced labor exploitation is carried out can help to identify weaknesses in labor protections for workers. 3. Evaluate alternatives to criminal justice system processing. As noted above, a key takeaway from this study is that practitioners were unsatisfied with outcomes from criminal justice system processing. However, philosophies of punishment dictate that formal criminal justice system processing is essential to deter crime and rehabilitate individuals engaged in crime. Any alternative approach, such as restorative justice, restitution, and naming and shaming recommended above, must be rigorously evaluated to determine whether these goals can be accomplished outside of formal criminal court processing. 193 Appendices Appendix A. Data Cleaning Documentation This appendix provides detailed information about the administrative data cleaning procedures. Data Source I used a dataset developed by the Anti-Slave Labor and Trafficking in Persons clinic (the clinic) in the law school at the Federal University of Minas Gerais in 2018- 2019. This dataset links information from labor inspection reports with case processing and outcome data. The dataset includes data on the sentences and judgments handed down within the scope of the five Federal (criminal) Regional Courts and the 24 Regional Labor (civil) Courts, in addition to the Superior Court of Justice, Superior Labor Court and the Supreme Federal Court, that is, encompassing the entire Brazilian territory and all judicial bodies. To develop the dataset, the clinic team requested a list of all criminal actions and public civil actions from the Federal Public Ministry and the Public Labor Ministry, respectively. The team then reviewed the courts? websites to identify any other lawsuits that were missing from the lists and then requested data from those cases specifically. Upon review of these lawsuits, the clinic team discovered that most criminal actions were initiated in 2008 and most civil lawsuits began in 2012. Thus, they limited their sample to cases initiated in the criminal court system in 2008 or later and in the civil court system in 2012 or later. The team also thought it would be analytically advantageous to limit the sample to criminal cases that were initiated in 2008 or later because all cases would be reviewed against the revised version of article 149 (revised in 2003). Establishing 2008 as the initial time frame also facilitated access to data on case outcomes, given that not all sentences and judgments handed down in the early 2000s are available on the institutional websites of the regional federal and labor courts. The clinic team searched for procedural changes (e.g. if the case was rejected or dropped for any reason), as well as sentencing decisions for each case on the courts? websites. Some cases were confidential and, therefore, it was not possible access them or to see their procedural progress. Additionally, some cases, despite containing information on all the procedural changes, did not present the full detail about sentencing. In cases where there was not complete data for a case, the team searched on sites like JusBrasil, which is a privately run website that makes court data publicly available. When the team could not find the missing case information on JusBrasil, they made direct contact with the court via email and telephone. When the team could still not fill in details after these attempts, they excluded the case from the dataset. I do not have information on precisely how many cases were excluded for this reason. 194 For cases in the labor court system, the team excluded all cases that were not public civil actions (i.e. cases brought forward by the government). That is, any individual lawsuit from a single employee brought against their employer is not included in the data. The team also excluded any cases that did not include charges under Article 149 of the penal code (forced labor exploitation). During the initial review of the data, the clinic team identified a few public civil actions with incomplete data (for example, data on appeals only and nothing prior). Upon further exploration of the Public Labor Ministry website, the team identified an additional 23 public civil actions that took place between 2008 and 2011 that were related to incomplete cases that were included in the original sample frame of cases from 2012 onward. These 23 cases are included in the final dataset despite having a start date prior to 2012. In some instances, it was not clear whether a case should be included in the dataset. For example, in criminal cases that did not end up moving to sentencing, the court?s decision to receive the complaint was not available. In this situation, there was no data on the crimes reported in the complaint. In these cases, the team chose to trust the Public Federal Ministry that these contained charges for article 149 and missing data was listed for other crimes in this complaint. A similar situation happened with public civil actions; when information on the initial complaint was missing the team trusted the list provided by the Public Ministry of Labor and indicated that the complaint was for article 149. The team matched this court data with labor inspection reports for each case, when available. Inspection reports provide robust detail about the situation at the jobsite itself. Labor inspection reports are frequently the basis for complaints made in court. The team was able to match court data to an inspection report for 67.6% of all cases. The Clinic team was limited to information from criminal proceedings to determine whether a police investigation took place. The team did not have access to the Federal Police database, but they were able to use a publication titled "Action by the Federal Police in combating crimes that violate Human Rights." The data on police investigations is limited to whether an investigation took place. In terms of investigation length, the team assumed the entire period after the crime and before the filing of the action was spent on investigation procedures. Data Entry The clinic team entered three categories of data about each case: 1) procedural movement, 2) time lapses, and 3) content of decisions. Procedural movement of a case refers to its progression through the court system from the court receiving the complaint from the prosecutor?s office, their details of the complaint, the sentencing outcome, the details of the sentence, and the punishment, if any, associated with the sentence. When entering data related to the procedural movement of the case, the 195 team was able to flag cases that should have been excluded and to drop unneeded measures and flag missing data for key variables. The clinic team used Google Forms to enter data from inspections and court processing in an objective and standardized way. The full version of the form is available as an attachment at the end of this work. The form was divided into four sectors: data from inspection, data from labor court proceedings, data from criminal proceedings and, finally, time frames. Each analyzed process received a specific name. It was adopted as a criterion the use of the acronyms "AP" (criminal action) and "ACP" (public civil action), followed by the acronym of the State where it is being processed and the number of the case. Thus, examples of names given to the processes are: AP.MG.1508- 38.2008.4.01.3810 and ACP.SP.1405-90.5.02.0071. Part of the form's fields was filled with information extracted from the reading of the judicial decisions. Then, it was necessary to use the inspection reports prepared by the labor inspectors for each judicial process, to extract some of the information required in the form. Next, the databases from the ministry of labor and data from public civil actions were consulted to obtain information related to labor agreements (TACs). Finally, files from the Labor Inspection Secretariat were consulted to identify the names of individuals and companies in the dirty list. When filling out the forms, the clinic team linked legal proceedings to the inspection reports. This was done by matching the name of the defendant registered in the assessment of the case with the name of the employer in the inspection. However, some lawsuits did not have the employer as a liability, but a representative, a third party involved in the case, or a related company, which were not necessarily mentioned in the inspection report. In these situations, additional information was sought in the decisions of the process, such as the name of the inspected establishment and the date of inspection. As many judges do not mention this data in decisions, in some cases it was impossible to establish the connection. The situation was more complex in relation to the cases for which court proceedings were still in progress, and therefore had not been sentenced, nor had a decision made available. After data collection on the form was completed, they were converted into an Excel spreadsheet. The final sample includes 1464 criminal actions and 432 public civil actions. I renamed all variables with English names for my own benefit. In the documentation below I include original variable names in Portuguese for reference. Variables Labor Case The variable laborcase (Deslindetrabalhista0sen?o) indicates whether a case was processed in the labor court. Cases that ended with a plea agreement (TAC) are coded 196 as 0, not processed in the labor court. Roughly 25% (n=432) of cases were processed in the labor court system. Another variable laborcase2(Houvedeslindetrabalhista0s) also indicates whether a case was processed in the labor court. In this variable, the same number of cases are identified as being processed in the labor system (n=432), one case is coded as missing, and 1,362 are coded as not being processed in the labor system. However, when I cross tabulated the two different labor case variables, I saw that there were three cases that were coded ?yes? in the first variable and ?no? in the second, and vice versa. I followed up with the Clinic team to learn which of these cases was accurate and they identified a data coding error which they subsequently corrected. A cross tab of both labor case variables now shows identical number of cases in the yes and no response options. Criminal Case The variable crimcase (Deslindepenal0sen?o) indicates whether a case was processed in criminal court. Roughly 83% (n=1,464) cases were processed in the criminal court system. As with the labor case variable, a second variable crimcase2(Houvedeslindepenal) also indicates whether a case was processed in criminal court. In this variable, the same number of cases are identified as being processed in the criminal system (n=1,464), and similar to the first variable, no cases are missing. However, when I cross tabulated the two criminal case variables I saw that there was one case coded ?yes? in the first variable and ?no? in the second, and vice versa. Again, I followed up with the clinic and they identified a data coding error which was corrected. A cross tabulation of both criminal case variables now shows identical number of cases in the yes and no response options. Comparison of labor and criminal processing All cases that are processed in the criminal system should be processed in the labor system as well, per legislative policy in Brazil. I did a cross tab to see if this was the case. I found, however, that only 132 cases were processed in both court systems and that 1,332 cases were processed only in the criminal court system. I reached out to the Clinic to inquire about this and they noted that cases that show up in the data as ?0? for labor case and ?1? for criminal case ended with a TAC (plea agreement) in the labor system. Public collective action in labor court One option for case processing in labor court is a public civil action. This is a situation where a public prosecutor brings forward a case (rather than an individual victim?s lawyer). The other option is for a case to end in a TAC. Given that TACs will show up as ?0? in the labor case variable, I expected to find that all 432 rows labeled labor case were also coded as ?1? for the variable labor_civaction, which indicates whether there was a civil action. This is the case. 197 Public collective action received a sentence This variable, labor_civaction_sent, indicates whether the case was convicted and sentenced in labor court. About 87% (n=377) of the 432 labor cases received a sentence. Labor inspection report data linked The variable audited indicates whether a labor inspection report was identified to link to the court data. There are four response options for this variable, No (n=418); Yes (n=1,121), no information (n=99), and N/A- duplicate (n=126). N/A duplicate is used in cases where multiple cases arise from the same inspection report. This ensures that one inspection is not double counted for each case it generated. I did not use this variable in my analysis, but I used it for data checks during the cleaning and analysis process. Economic Sector Sector is a string variable indicating the primary economic activities that workers took part in. Data in this variable contained spelling errors, inconsistent capitalization, and inconsistent phrasing. I used a series of string match commands to group the same activities together, show in Table 1 below. I made these decisions in conjunction with advice from Dr. Haddad and his team. Table 1. Economic Sector Categories Category Economic activities included (translated) Agriculture, sugar cane, rubber ? Agriculture ? Fruticulture ? Production of citrus ? Mill and grain ? Harvesting ? Sugar and alcohol production ? Rubber Livestock and fishing ? Fish raising ? Raising poultry for meat ? Raising cows for meet ? Pork and poultry Charcoal ? Charcoal production Forestry/logging ? Wood ? felling of the native forest for later formation of the farm ? Illegal allotment of preserved area 198 ? Forest management ? Reforestation ? Forestry ? Deforestation ? Commercial pine reforestation ? Felling trees ? Eucalyptus extraction ? Cutting Carnauba Transportation, domestic work, hotel- ? Hotel related, sales ? Cleaning ? Restaurant ? Hospitality ? Diner ? Food ? Domestic services ? Domestic work ? Retailing products and utilities for vehicles and personnel ? Locksmiths ? Sales ? Services ? Parking ? Transport ? Transportation and custody of asset surveillance ? Transport and delivery ? Coastal shipping ? Railroad ? Fast delivery services ? Transportation of valuables Construction and energy ? Construction ? Civil construction ? Industry ? Wholesale of building materials in general ? Rental of stages, roofs, and other structures for temporary use, except scaffolding ? Industrial assembly ? Steel workers ? Energy ? Providing services for an energy company ? Electricity 199 Manufacturing ? Manufacturing ? Pottery ? Nets ? Ceramics ? Beverages Mining/Extraction ? Mining ? Stone extraction and shredding ? Mineral extraction ? Extraction ? Mining ? Vegetable extraction ? Extraction and processing of passava fibers ? Resin ? Metallurgy Rural other ? Collection of non-timber products ? Construction of a fence Urban Other ? Recycling ? Refrigerator ? Sorting recyclable garbage ? Victims exercised the function of municipal guards ? Surveillance and security ? Trade Rural Economic Sector In the interviews, I learned that the distinction between urban and rural activities was more important than the distinction between activities within urban and rural subgroups. Accordingly, I created a variable that indicated whether the sector was rural. I then divided the economic sectors as indicated in table 2 below. Table 2. Categorization of activities into urban vs. rural Rural ? Agriculture, sugar cane, rubber ? Livestock and fishing ? Charcoal ? Forestry/logging ? Mining and extraction ? Other rural Urban ? Transportation, domestic work, hotel-related, sales 200 ? Construction and energy ? Manufacturing ? Other urban State and City State2 and city are variables that indicate the state and city in which alleged forced labor exploitation took place. I removed non-numeric characters and converted these to numeric variables. Region I created a variable, region, to group states by the geopolitical areas of Brazil: The North, Northeast, Midwest, and Southeast. ? North: Acre, Amap?, Amazonas, Par?, Rond?nia, Roraima, Tocantins ? Northeast: Alagoas, Bahia, Cear?, Maranh?o, Para?ba, Pernambuco, Piau?, Rio Grande do Norte, Sergipe ? Midwest: Goi?s, Mato Grosso, Mato Grosso do Sul, Distrito Federal (Federal District) ? Southeast: Esp?rito Santo, Minas Gerais, Rio de Janeiro, S?o Paulo ? South: Paran?, Rio Grande do Sul, Santa Catarina Source: World Atlas Number of workers reached and rescued 201 The variable n_reached indicates how many workers were reached during the inspection and n_rescued indicates how may workers were rescued. I probed on this distinction in the interviews and learned that not all workers in a business may be subject to forced labor exploitation. Further, not all workers will agree to be removed from the jobsite so they can receive assistance. Based on discussions with clinic staff, I determined that number of workers rescued, however, is the best proxy for the true number of workers victimized. There are 703 missing values for number reached and 796 missing values for number rescued. Number on the inspection team The variable n_team refers to the number of individuals involved in an inspection team. Values for this variable range from 1-91. Larger teams are typically deployed for suspected larger and more complex operations, however, inspection team size may also reflect the availability of resources in an area. There are 704 cases with missing data for the number of individuals on the inspection team. Members of the inspection team The variable insp_parts is a string variable that lists all agencies involved in the labor inspection for a given case. I created a series of dummy variables to indicate whether specific agencies were involved. These are summarized in Table 3 below. Table 3. Proportion of inspections that include specific agencies (N=1,074) Inspection team member Mean SD Federal Police 0.491 0.500 Federal labor prosecutor 0.711 0.453 Federal road police 0.256 0.437 Public defender's office 0.0493 0.217 Military police 0.0680 0.252 Civil police 0.0531 0.224 Environmental/forest police 0.0279 0.165 State labor prosecutor 0.715 0.452 Local antitrafficking agency 0.00559 0.0746 Construction workers union 0.000931 0.0305 Brazilian Army 0.00279 0.0528 Regional labor office 0.00279 0.0528 Brazilian EPA federal and local 0.0158 0.125 Public criminal prosecutor?s office 0.00186 0.0431 Other inspection participant 0.00652 0.0805 202 Number of workers who received unemployment and unemployment amount For the variable number of workers who received unemployment n_unempl I renamed the variable and made no other changes. For the amount of unemployment unemp_amt I removed non-string characters ($,R) and recoded all values to replace commas with a decimal and remove the decimal from the comma position. Crimes Identified in the Labor Inspection Report A series of variables indicates whether specific crimes were identified in the labor inspection report. Specifically, human trafficking (article 149-A), endangering the health of livelihood of an individual (article 132), violation of worker?s labor rights (article 203), recruitment fraud (article 206), and enticing workers (article 207). Articles 206 and 207 are similar; 206 is typically applied to international recruitment fraud and 207 is typically applies to domestic recruitment fraud. There is also a variable indicating whether another crime type was identified. Missing data was originally coded as ?2,? I recoded as missing ?.? Crimes identified in the labor inspection report are summarized in Table 4. Table 4. Proportion of cases which identified other crimes Crimes N N missing Mean SD Human trafficking (149-A) 1,072 692 0.0103 0.101 Endanger someone?s health or life (132) 1,075 689 0.0577 0.233 Violation of worker?s rights (203) 1,075 689 0.110 0.313 Recruitment fraud/ enticing workers 1,075 689 0.000930 0.0305 across international borders (207) Recruitment fraud/ enticing workers 1,075 689 0.0558 0.230 domestically (206) Other crime identified 1,072 692 0.0858 0.280 Characteristics of forced labor exploitation A series of dummy variables (char_forcedlabor char_degcond char_jornada char_debt char_transport char_guards char_guns char_documents char_recruitmentfraud) indicate whether the labor inspector indicated in their report that a given characteristic of forced labor exploitation was present. For each of these variables I recoded the value ?2? to missing as indicated in the value label. Dirty list 203 The variable conseq_dirtylist indicates whether the business was added to the dirty list following their inspection. I recoded values ?2? and ?3? to missing, as indicated in the codebook. These values indicated ?no information? and ?duplicate?, respectively. Request for individual moral damages For the variable moral_ind_req, I recoded the value ?2? to missing and made no other changes to this variable. Individual moral damages For the variable moral_ind_amt I removed non-string characters ($,R) and recoded all values to replace commas with a decimal and remove the decimal from the comma position. Collective moral damages For the variable morald_coll_amt, I removed non-string characters ($,R) and recoded all values to replace commas with a decimal and remove the decimal from the comma position. Judicial declaration of forced labor exploitation For the variable labor_jud_decl, I renamed this variable and made no other changes. Sentencing outcomes for public civil actions The variable labor_outcome represents the final case outcome after all appeals, if there were any. I recoded string values as numeric, with seven values including convicted, partial conviction, no conviction, agreement, partial agreement, agreement in process, partial conviction and agreement, case dropped, and in progress. I combined these values into four unique values, as follows: ? Full or partial conviction includes convicted, partial conviction, and partial conviction and agreement. ? Full or partial agreement includes agreement, partial agreement, and agreement in process. ? Not convicted includes not convicted and case dropped. ? In progress includes in progress. Case processing length for public civil actions I used a series of timeline variables to generate the length_labcase variable. Using guidance from the clinic, the start date for labor court processing was the civil action was received by the labor court. This variable, tl_lab_civaxn was formatted as string. 204 I replaced the value ?X? to be missing (? ?) then formatted the variable as MDY. One case was missing data on the start date, I left this as missing. The end of the process was the date of public collective action sentencing. There were no non-numeric characters in this variable, tl_lab_civaxn_sent, so I formatted the variable as a date in MDY format. I then generated a new variable that subtracted the end date from the beginning date. Initially this generated one negative value, so I substituted the judgment date in for this case. Judgment date is earlier than the sentencing date for this value and any value where sentencing date was not available and the case was not listed as still in progress. I then browsed the data and did a few manual calculations to confirm the subtraction worked correctly. Sentencing outcomes for criminal actions To identify criminal sentencing outcomes I had to undertake a multistep process. First, the outcomes for all defendants in a case were listed in one cell separated by commas or semi colons. I first split the variable by comma and then each of those split variables I split at semicolons. Using a loop across all of the different outcome variables, I replaced the individual outcome variables with a numeric value depending on whether the contained different combinations of string values. I created 16 categories: ? Convicted ? Statute of limitations ? Pardon ? In progress ? Total acquittal ? Partial acquittal ? Archived ? Extinct ? Complaint rejected ? Suspended ? Death ? Court not competent to receive case ? Lis Pendens ? In review ? Locked ? No information available/missing I then generated individual variables identifying whether a given outcome occurred for any defendant in a case. For the analysis I combined partial and total acquittal into one acquittal category. 205 Reasons for acquittal I renamed this variable and made no other changes. Court recognition of elements of forced labor exploitation A series of dummy variables indicated whether the judge recognized specific elements of forced labor exploitation in the sentence. For each of the variables, I recoded values ?2? (not enough information) and ?3? (not mentioned) as missing and added value labels but made no other changes. Case processing length for criminal cases The start date for criminal case processing length was the date the court received the complaint. For this variable, tl_crim_received I recoded ?NTI? as missing. For most values, the date was formatted as follows DD/MM/YYYY, but in some cases it was formatted, for example, 02jan2021 (nine characters). I first split the variable at the slash in the date so that there were three separate variables, month received, day received, and year received. When there were no slashes in the value, the month received variable became identical to the original, unparsed variable. I then created a day variable equal to the parsed day column if the first parsed month column was one or two characters long. I replaced the value of the day variable with the first two characters of the month variable if the month variable was nine characters long. I then generated a month variable equal to the month column that was parsed from the original variable, if the parsed month column had a value that was one or two characters long. I replaced the month variable as equal to the third and fourth values of the parsed month variable if the parsed month variable was nine characters long. Finally, I created a year variable equal to the parsed year column if the parsed month variable was one or two characters long. I then replaced the year variable with the last four values of the parsed month variable, if the parsed month variable was nine characters long. I then converted the month values from string to numeric, e.g. ?jan? to ?1?. Then, I converted the string variable for month, day, and year, to numeric and generated one combined date variable using the three, separate month, day, and year variables. I then browsed the original date variable and the one I created to confirm the transformation happened correctly. For the end date of the criminal court processing I used sentencing date in federal court. This variable was already in date format so no changes were needed. 206 I subtracted the start date from the end date to come up with the final case length. If the final sentence date was missing, and the case was not listed as in progress, I substituted in the judgment date (which is before sentence date). I then browsed the data and did a few manual calculations to confirm the subtraction worked correctly. 207 Appendix B. Interview Protocols Below are the interview protocols I used. I began with English language protocols, then translated into Portuguese. I worked with Clinic staff to make edits in Portuguese, and then back translated the protocols provided here. 208 Federal and Labor Judge Interview Protocol RESPONDENT ID: _________________ RECORD DATE: _________________ RECORD START TIME: _________________ R HAS COPY OF CONSENT FORM: YES NO VERBAL CONSENT OBTAINED: YES NO PERMISSION TO RECORD: YES NO PERMISSION TO QUOTE: YES NO Do you have any questions before we begin? During the interview, let me know if at any moment you would like me to repeat a question or if you don?t understand what I am asking. You can also let me know if you would like to skip a question. 209 Section 1: Explanation of Study and Introduction Thank you for taking the time to talk with me today. First, I would like to tell you a little about the study and why I wanted to talk with you. The objective of our conversation today is to learn more about forced labor exploitation cases, including things that make it easier or more difficult to judge these cases. To start, I would like to ask a few questions about you and your experience working with forced labor exploitation cases. 1. What is your job title? 2. How long have you been in this position? 3. How are you involved in forced labor exploitation cases in your regular work? 4. For how long have you worked with forced labor exploitation cases? 5. Can you tell me about what kind of training, if any, you received related to processing forced labor exploitation cases? 210 Section 2: Experience with Slave Labor Cases Now I would like to talk about forced labor exploitation in general and then I would like to discuss the types of cases you saw in your courtroom. 1. What are the most common kinds of forced labor exploitation that you see in your court? a. If not mentioned: Based on your experience, slave labor is most common in which economic activities? 2. What are the crimes that occur most often alongside slave labor (art. 149 do C?digo Penal)? 3. In your opinion, what characterizes (ask about each element) [Questions 4 and 5 added after first three interviews] 4. Is physical violence needed to characterize the crime of slave labor? 5. Is restriction of liberty or transportation needed to characterize the crime of slave labor? 6. What kinds of slave labor cases should be judged in [labor/criminal court] (versus [labor/criminal] court)? a. If not mentioned: In what circumstances, if any, should a case only be judged in the [labor/criminal] court? 7. What kinds of things make it more likely that a slave labor case will receive a conviction? 8. What kinds of things make it less likely that a slave labor case will receive a conviction? [stopped asking the below probe after first few interviews] If not mentioned: if a case has already received a judgment in the labor court, does that make it more or less likely that it will receive a sentence in criminal court? 9. Is there anything about the penal code that makes it easier or more difficult to judge slave labor cases? 10. What is the minimum [sentence/fine] you would give for defendants who are found guilty of slave labor? 211 11. Which instruments are most effective for preventing or punishing slave labor (if needed, list: fines implemented by labor inspectors, dirty list, TAC, criminal sentence, labor sentence) Section 3: Collaboration Thank you. Now I would like to talk about the different people who are involved in the process of investigating and prosecuting slave labor cases. 1. How are police involved in the process? a. Is it common for the police to arrest an individual suspected of slave labor before their trial? 2. What is the role of labor inspectors involved in investigating and prosecuting slave labor cases? 3. How much weight or importance do you give to the reports that labor inspectors prepare? (If needed: do you trust that their reports are accurate?) 4. Federal only: In your opinion, when is a police investigation needed? (If not mentioned: in what situation is the labor inspector?s investigation sufficient on its own)? 5. How has the [Public Prosecutor's Office/ Labor Prosecutor?s Office] been in working on slave labor processes? Can anything be done to improve their performance? 6. What groups are not usually involved in slave labor cases that you think should be involved? [Stopped asking question 18 a few interviews in] 7. How, if at all, do you collaborate with judges, prosecutors, or other court staff from the labor court about a case you are all working on? 19. In your opinion, is the processing of slave labor processes, from inspection to judgement, slow, normal or fast? Why? 20. What do you consider as the ideal period to complete the process, from inspection to judgement? What kinds of things need to be changed to achieve this ideal timeframe to complete the process? 212 21. What kind of research would you like to have on slave labor in Brazil or in the world? Section 4: Closing and Final Thoughts Before we end, I have one final question. 1. Is there anything we did not talk about that you think we should have covered? If yes, what would you like to add? Thank you for your time and contribution to this study. RECORD END TIME: ________ 213 Labor and Federal Prosecutor Interview Protocol RESPONDENT ID: _________________ RECORD DATE: _________________ RECORD START TIME: _________________ R HAS COPY OF CONSENT FORM: YES NO VERBAL CONSENT OBTAINED: YES NO PERMISSION TO RECORD: YES NO PERMISSION TO QUOTE: YES NO Do you have any questions before we begin? During the interview, let me know if at any moment you would like me to repeat a question or if you don?t understand what I am asking. You can also let me know if you would like to skip a question. 214 Section 1: Explanation of Study and Introduction Thank you for taking the time to talk with me today. First, I would like to tell you a little about the study and why I wanted to talk with you. The objective of our conversation today is to learn more about forced labor exploitation cases, including things that make it easier or more difficult to prosecute these cases. To start, I would like to ask a few questions about you and your experience working with forced labor exploitation cases. 6. What is your job title? 7. How long have you been in this position? 8. How are you involved in forced labor exploitation cases in your regular work? 9. For how long have you worked with forced labor exploitation cases? 10. Can you tell me about what kind of training, if any, you received related to processing forced labor exploitation cases? 215 Section 2: Experience with Slave Labor Cases Now I would like to talk about forced labor exploitation in general and then I would like to discuss the types of cases you have prosecuted. 1. What are the most common kinds of forced labor exploitation that you see in your work? a. If not mentioned: Based on your experience, slave labor is most common in which economic activities? 2. Who refers cases to you? a. Which of these individuals or agencies most commonly refers cases to you? 3. After a case is referred, how do you decide whether to move forward with it? b. Is there a specific protocol you use to decide whether to bring forward charges for a slave labor case? 4. Of all the cases you receive each year, about what proportion are slave labor cases? 5. What are the crimes that occur most often alongside slave labor (art. 149 do C?digo Penal)? 6. In your opinion, what characterizes (ask about each element) [Questions 7 and 8 added after first three interviews] 7. Is physical violence needed to characterize the crime of slave labor? 8. Is restriction of liberty or transportation needed to characterize the crime of slave labor? 9. What kinds of slave labor cases should be judged in [labor/criminal court] (versus [labor/criminal] court)? c. If not mentioned: In what circumstances, if any, should a case only be judged in the [labor/criminal] court? 10. What types of evidence do you need to prosecute a slave labor case? b. Do these factors change depending on the type of slave labor case you are processing? 216 11. What factors, if absent, would prevent you from processing a case as slave labor (for example, victim testimony)? 12. What factors make it more likely that the defendant will be convicted (when processing a case of slave labor)? 13. What challenges do you face during the prosecution of slave labor cases? c. What additional support would you like to have to overcome these challenges? 14. In your opinion, what would help, overall, to improve the prosecution of cases of slave labor? Section 3: Collaboration Thank you. Now I would like to talk about the different people who are involved in the process of investigating and prosecuting slave labor cases. 1. How are police involved in the process? b. Is it common for the police to arrest an individual suspected of slave labor before their trial? 2. What is the role of labor inspectors involved in investigating and prosecuting slave labor cases? 3. How much weight or importance do you give to the reports that labor inspectors prepare? (If needed: do you trust that their reports are accurate?) 4. How has the [Public Prosecutor's Office/ Labor Prosecutor?s Office] been in working on slave labor processes? Can anything be done to improve their performance? 5. What groups are not usually involved in slave labor cases that you think should be involved? [Stopped asking question 18 a few interviews in] 6. How, if at all, do you collaborate with judges, prosecutors, or other court staff from the labor court about a case you are all working on? 7. In your opinion, is the processing of slave labor processes, from inspection to judgement, slow, normal or fast? Why? 217 8. What do you consider as the ideal period to complete the process, from inspection to judgement? What kinds of things need to be changed to achieve this ideal timeframe to complete the process? 9. What kind of research would you like to have on slave labor in Brazil or in the world? [section on COVID-19 added April: questions asked only if respondent mentioned COVID-19 during their interview] Responding to the Pandemic Now I would like to talk a little about how the pandemic is affecting your work with respect to slave labor cases. 1. What, if anything, are you doing to adapt your work for the pandemic? 2. Considering that labor inspection is dynamic, with the need to carry out inspections in the establishments, long trips in vehicles, contact with many people, what precautions are being taken to avoid COVID-19? 3. In your opinion, could the pandemic affect the prevalence of any of the characteristics of slave labor? 4. In your opinion, how will the pandemic affect the prosecution of cases of slave labor? 5. Was there a reduction in complaints about slave labor in this period? Section 4: Closing and Final Thoughts Before we end, I have one final question. 1. Is there anything we did not talk about that you think we should have covered? If yes, what would you like to add? Thank you for your time and contribution to this study. RECORD END TIME: ________ 218 Interview Protocol for Labor Inspectors RESPONDENT ID: _________________ RECORD DATE: _________________ RECORD START TIME: _________________ R HAS COPY OF CONSENT FORM: YES NO VERBAL CONSENT OBTAINED: YES NO PERMISSION TO RECORD: YES NO PERMISSION TO QUOTE: YES NO Do you have any questions before we begin? During the interview, let me know if at any moment you would like me to repeat a question or if you don?t understand what I am asking. You can also let me know if you would like to skip a question. 219 Section 1: Explanation of Study and Introduction Thank you for taking the time to talk with me today. First, I would like to tell you a little about the study and why I wanted to talk with you. The objective of our conversation today is to learn more about forced labor exploitation cases, including things that make it easier or more difficult to investigate these cases. To start, I would like to ask a few questions about you and your experience working with forced labor exploitation cases. 11. What is your job title? 12. How long have you been in this position? 13. Could you describe a typical work day, if there is such thing as a typical work day for you? 14. Can you tell me about what kind of training, if any, you received related to processing forced labor exploitation cases? a. Who provides this training? b. How often do you undergo continuing training related to investigating slave labor? c. What additional training, if any, do you think would be useful? Section 2: Identifying Cases Now I would like to talk about the types of cases you see in your work. 1. How do you identify new cases of slave labor? a. Are there any other ways you would learn about slave labor cases? b. Who provides tips or complaints that you investigate? (General public, employees, employees' family members, anyone else?) c. What makes it more likely that you will investigate a complaint? What makes it less likely? Is there a protocol or policy that you use to decide whether to investigate a complaint? 2. How do you prioritize which companies to inspect? 3. What are the most common kinds of forced labor exploitation that you investigate in your region? 220 d. If not mentioned: Based on your experience, slave labor is most common in which economic activities? 4. What are the crimes that occur most often alongside slave labor (art. 149 do C?digo Penal)? 5. In your opinion, what characterizes (ask about each element)? a. For each element, ask: what kind of evidence would you collect to prove the existence of [element]? 6. Can you walk me through what happens if you identify that slave labor is present during an inspection? a. Who would you notify? b. What kind of follow-up do you do? How long does follow-up last? c. In general, are the business owners aware that you are coming? If so, do they take efforts to hide evidence of slave labor? d. Is the business owner arrested? e. In cases where the business owner is arrested, will they still be able to pay fines and backpay to the workers? 7. Based on your experience, what are the best practices for inspecting slave labor cases? 8. What challenges do you face when investigating cases of slave labor - from the moment you start the inspection to filing your report? a. How, if at all, do you overcome these challenges? Section 3: Prosecution 1. What is your role in the prosecution of slave labor cases? (you need to testify in court, provide evidence, provide a written statement, etc.) 2. Are you called to testify as a witness to the cases you inspected in federal and labor courts? 221 a. What, if anything, can be improved in collecting your testimony in federal or labor courts? 3. Do you know if the cases you inspect are heard in federal and labor courts? Are you aware of the outcome of the trials? 4. In your experience, what kinds of things make it more likely that a case of slave labor will be convicted? What kinds of things make it less likely? [section on COVID-19 added April: questions asked only if respondent mentioned COVID-19 during their interview] Responding to the Pandemic Now I would like to talk a little about how the pandemic is affecting your work with respect to slave labor cases. 2. What, if anything, are you doing to adapt your work for the pandemic? 2. Considering that labor inspection is dynamic, with the need to carry out inspections in the establishments, long trips in vehicles, contact with many people, what precautions are being taken to avoid COVID-19? 3. In your opinion, could the pandemic affect the prevalence of any of the characteristics of slave labor? 4. In your opinion, how will the pandemic affect the prosecution of cases of slave labor? 5. Was there a reduction in complaints about slave labor in this period? Section 3: Closing and Final Thoughts Before we end, I have two final questions. 2. What kind of research would you like to have on slave labor in Brazil or in the world? 3. Is there anything we did not talk about that you think we should have covered? If yes, what would you like to add? Thank you for your time and contribution to this study. RECORD END TIME: ________ 222 Interview protocol for federal police RESPONDENT ID: _________________ RECORD DATE: _________________ RECORD START TIME: _________________ R HAS COPY OF CONSENT FORM: YES NO VERBAL CONSENT OBTAINED: YES NO PERMISSION TO RECORD: YES NO PERMISSION TO QUOTE: YES NO Do you have any questions before we begin? During the interview, let me know if at any moment you would like me to repeat a question or if you don?t understand what I am asking. You can also let me know if you would like to skip a question. 223 Section 1: Explanation of Study and Introduction Thank you for taking the time to talk with me today. First, I would like to tell you a little about the study and why I wanted to talk with you. The objective of our conversation today is to learn more about forced labor exploitation cases, including things that make it easier or more difficult to investigate these cases. To start, I would like to ask a few questions about you and your experience working with forced labor exploitation cases. 1. What is your job title? 2. How long have you been in this position? 3. How are you involved in slave labor cases in your daily work? a. Have you ever participated in inspections with labor inspectors? Voc? participa de investiga??es lideradas por auditores do trabalho? Se sim, qual ? o seu papel nesta investiga??o? 4. What special task forces, if any, related to slave labor operate in your district? a. Are you a member of this task force? 5. Can you tell me about what kind of training, if any, you received related to processing forced labor exploitation cases? a. Who provides this training? b. How often do you undergo continuing training related to investigating slave labor? c. What additional training, if any, do you think would be useful? Section 2: Experience with Forced Labor Exploitation Cases Now I would like to talk about the types of cases you see in your work. 1. Voc? acha que o trabalho escravo ? um problema s?rio em sua comunidade? Por que ou por que n?o? 2. Que tipos de casos de trabalho escravo s?o investigados em sua regi?o com mais frequ?ncia? e. SE N?O MENCIONAR: Baseado na sua experi?ncia, o trabalho escravo ocorre principalmente em quais atividades econ?micas? 224 3. What are the most common kinds of forced labor exploitation that you investigate in your region? f. If not mentioned: Based on your experience, slave labor is most common in which economic activities? 4. What are the crimes that occur most often alongside slave labor (art. 149 do C?digo Penal)? 5. In your opinion, what characterizes (ask about each element)? b. For each element, ask: what kind of evidence would you collect to prove the existence of [element]? 6. Is physical violence needed to characterize the crime of slave labor? 7. Is restriction of liberty or transportation needed to characterize the crime of slave labor? 8. How do identify possible cases of slave labor? a. Once you are aware of a possible incident of slave labor, how do you decide whether to move forward with an investigation? 9. Based on your experience, what are the best practices for inspecting slave labor cases? 10. What challenges do you face when investigating cases of slave labor - from the moment you start the inspection to filing your report? b. How, if at all, do you overcome these challenges? 11. Thinking about all of the cases you?ve investigated where you found evidence of slave labor, about what proportion were charged in the criminal court system? a. What do you think would increase the chances that a prosecutor would bring charges on a case? What would decrease the chances? 12. Based on your experience, what kinds of proof does a federal prosecutor need to bring charges forward for slave labor? 13. In your opinion, what would improve prosecution of slave labor cases? 225 Section 3: Prosecution 1. What is your role in the prosecution of slave labor cases? (you need to testify in court, provide evidence, provide a written statement, etc.) 2. What, if anything, can be improved in collecting your testimony in federal or labor courts? 3. Do you know if the cases you inspect are heard in federal and labor courts? Are you aware of the outcome of the trials? [section on COVID-19 added April: questions asked only if respondent mentioned COVID-19 during their interview] Responding to the Pandemic Now I would like to talk a little about how the pandemic is affecting your work with respect to slave labor cases. 1. What, if anything, are you doing to adapt your work for the pandemic? 2. Considering that labor inspection is dynamic, with the need to carry out inspections in the establishments, long trips in vehicles, contact with many people, what precautions are being taken to avoid COVID-19? 3. In your opinion, could the pandemic affect the prevalence of any of the characteristics of slave labor? 4. In your opinion, how will the pandemic affect the investigation or prosecution of cases of slave labor? 5. Was there a reduction in your investigations related to slave labor in this period? Section 3: Closing and Final Thoughts Before we end, I have two final questions. 4. What kind of research would you like to have on slave labor in Brazil or in the world? 5. Is there anything we did not talk about that you think we should have covered? If yes, what would you like to add? 226 Thank you for your time and contribution to this study. RECORD END TIME: ________ 227 Appendix C. Descriptive Information This appendix lists the tables and figures that provide additional descriptive information to complement the findings in Chapter 5. 228 Table 1 Interview respondent characteristics Experience Experience Respondent Current Respondent with urban with rural Number Sex Tenure Location type cases cases federal FJ001 Female 15+ years South East judge Yes Yes federal FJ003 Male <5 years Northeast judge Yes No Center- federal FJ004 Male 15+ years West judge No Yes Center- federal FJ005 Male 15+ years West judge Yes Yes 10-<15 federal FJ006 Male years North judge No Yes 5->10 federal FJ008 Female years Northeast judge No Yes 10-<15 Center- federal FP001 Male years West prosecutor No No federal FP002 Male <5 years Northeast prosecutor Yes Yes 10-<15 federal FP003 Male years North prosecutor Yes Yes federal FP006 Female 15+ years Northeast prosecutor No Yes 10-<15 federal FP007 Male years North prosecutor No Yes Center- federal FP009 Female 15+ years West prosecutor No Yes JT001 Male 15+ years South East labor judge No Yes JT002 Male 15+ years Northeast labor judge Yes Yes 5->10 Labor LA001 Female years Northeast inspector Yes Yes 5->10 Center- Labor LA002 Female years West inspector Yes Yes 10-<15 Labor LA003 Female years North inspector Yes Yes Labor LA004 Female <5 years Northeast inspector Yes Yes 5->10 Labor LA005 Female years South East inspector Yes Yes Labor LA006 Male 15+ years South East inspector Yes Yes 5->10 Not Labor LA008 Male years Applicable inspector Yes Yes 10-<15 Federal PO001 Male years North police Yes Yes Center- Federal PO002 Male 15+ years West police No Yes Center- Federal PO003 Male 15+ years West police No Yes labor TP001 Male 15+ years South East prosecutor Yes Yes 229 5->10 labor TP003 Male years Northeast prosecutor No Yes 10-<15 Center- labor TP004 Male years West prosecutor Yes Yes 5->10 Not labor TP005 Female years Applicable prosecutor Yes Yes 230 Table 2 Top cities where forced labor exploitation identified City State No. % Cum Sao Paolo Sao Paolo 48 2.9 2.9 Sao Felix do Xingu Para 34 2.1 5 Itupiranga Para 17 1 6 Acailandia Maranhao 16 1 7 Maraba Para 16 1 8 Rio de Janeiro Rio de Janeiro 16 1 9 Rondon do Para Para 14 0.9 9.9 Belo Horizonte Minas Gerais 11 0.7 10.6 Novo Repartimento Para 11 0.7 11.3 Porto Velho Rondonia 11 0.7 12 Toma-Acu Para 11 0.7 12.7 Boca do Acre Amazonas 10 0.6 13.3 Goaianesia do Para Para 10 0.6 13.9 Juara Mato Grosso 9 0.5 14.4 Paragominas Para 9 0.5 14.9 Altamira Para 8 0.5 15.4 Labrea Amazonas 8 0.5 15.9 Medicilandia Para 8 0.5 16.4 Salvador Bahia 8 0.5 16.9 Sao Desiderio Bahia 8 0.5 17.4 Other Other 1,355 82.7 100.1 Total 1,638 100 231 Table 3 Regional Labor Court Jurisdictions Region Location Jurisdiction 1 Rio de Janeiro Rio de Janeiro 2 S?o Paulo Grande S?o Paulo 3 Belo Horizonte Minas Gerais 4 Porto Alegre Rio Grande do Sul 5 Salvador Bahia 6 Recife Pernambuco 7 Fortaleza Cear? 8 Bel?m Par? e Amap? 9 Curitiba Paran? 10 Bras?lia Distrito Federal e Tocantins 11 Manaus Amazonas e Roraima 12 Florian?polis Santa Catarina 13 Jo?o Pessoa Para?ba 14 Porto Velho Acre e Rond?nia 15 Campinas Munic?pios do estado de S?o Paulo[1] 16 S?o Lu?s Maranh?o 17 Vit?ria Esp?rito Santo 18 Goi?nia Goi?s 19 Macei? Alagoas 20 Aracaju Sergipe 21 Natal Rio Grande do Norte 22 Teresina Piau? 23 Cuiab? Mato Grosso 24 Campo Grande Mato Grosso do Sul 232 Figure 24 Map of Labor Court Jurisdictions 233 Table 4 Federal Court Districts in Brazil and their Jurisdictions Region Location Jurisdiction 1st Acre, Amap?, Amazonas, Bahia, Distrito Federal, Goi?s, Maranh?o, Bras?lia Mato Grosso, Minas Gerais, Par?, Piau?, Rond?nia, Roraima, Tocantins 2nd Rio de Janeiro Esp?rito Santo e Rio de Janeiro 3rd S?o Paulo Mato Grosso do Sul e S?o Paulo 4th Porto Alegre Paran?, Rio Grande do Sul e Santa Catarina 5th Recife Alagoas, Cear?, Para?ba, Pernambuco, Rio Grande do Norte, Sergipe Figure 25. Map of the federal court districts in Brazil 234 Table 5 Crimes indicted and convicted with forced labor exploitation criminal actions* Indicted N=1,464 Sentenced, N=308 Variable Mean SD Freq. Mean SD Freq. Article 125- nonconsensual abortion 0.75% 8.64% 11 0 Article 147-threatening, by word, writing, gesture, or any other symbolic means, to cause unjust & serious harm 0.48% 6.90% 7 0.97% 9.84% 3 Article 149a- human trafficking 0.55% 7.37% 8 0 Article 150- trespassing 0.07% 2.61% 1 0 Article 159- abduction with ransom 0.07% 2.61% 1 0 Article 132- exposing the life or health of others to imminent danger 2.19% 14.63% 32 0 Article 171-crime against property 0.41% 6.39% 6 0.32% 5.70% 1 Article 203- not paying minimum wage 28.55% 45.18% 418 3.57% 18.59% 11 Article 206- recruit workers through fraud to take them out of the country 0.07% 2.61% 1 0 Article 207- recruit workers through fraud to take them to another location within country 14.62% 35.34% 214 4.87% 21.56% 15 Article 228- crime was committed with violence, or serious threat or fraud 0.07% 2.61% 1 0 Article 229-maintain, on their own account or as a third party, an establishment in which sexual exploitation occurs, whether or not there is an intention for profit or direct mediation 0.34% 5.84% 5 0.32% 5.70% 1 Article 230-Take advantage of the prostitution of others, directly participating in their profits or making themselves supported, in whole or in part, by those who exercise it 0.14% 3.69% 2 0 Article 231- International human trafficking for the purpose of sexual exploitation 0.41% 6.39% 6 0 Article 236- marriage fraud 0.07% 2.61% 1 0 Article 288-criminal association or gang participation (must be three or more people) 0.96% 9.74% 14 0 Article 293- mail fraud 0.07% 2.61% 1 0 Article 297- Falsify, in whole or in part, a public document, or change a true public document 25.20% 43.43% 369 11.36% 31.79% 35 Article 298-Falsify, in whole or in part, a private document, or change a true private document 0.20% 4.52% 3 0 235 Article 299-To omit, in a public or private document, a declaration that should be included in it, or to insert or cause to insert a false or different declaration 0.89% 9.38% 13 0.65% 8.05% 2 Article 304-use any of the forged or altered papers, referred to in arts. 297 to 302: Penalty - combined with falsification or alteration 0.07% 2.61% 1 0 Article 334-smuggling and embezzlement 0.07% 2.61% 1 0 Article 337a-Subtract, or render useless, totally or partially, an official book, process or document entrusted to the custody of an official, due to an official letter, or from a private person in public service 1.02% 10.07% 15 0 Article 342- perjury 0.07% 2.61% 1 0 Article 343- witness tampering/bribery 0.14% 3.69% 2 0 *Bolded crimes are those which were indicted and sentenced 236 Appendix D. Supplemental Analyses Table 1. Multinomial logistic regression of predicted latent class on case type, N=1,075 Labor Case Criminal Case Class 1 (versus 2) Coefficient -0.1 -0.18 SE 0.23 0.3 P-value 0.66 0.55 Class 3 (versus 2) Coefficient -0.07 -0.05 SE 0.46 0.62 P-value 0.88 0.94 Table 2. Multinomial logistic regression public civil action outcomes on latent class categories, N=185 Agreement vs. Not convicted vs. conviction RRR(SE) conviction RRR(SE) c = 2 (vs 1) 2.10 1.61 (1.01) (1.88) c = 3 (vs 1) 4.77 5.06e-06 (5.85) (0.01) Region= Northeast 1.11 508,147 (0.589) (4.230e+08) Region=North 2.850** 420,867 (1.404) (3.504e+08) Region-Southeast 1.070 1.404e+06 (0.561) (1.169e+09) Region=South 0.723 1.390e+06 (0.553) (1.157e+09) Federal labor prosecutor was part of inspection team 1.030 2.277 (0.367) (2.594) Constant 0.457 4.09e-08 (0.297) (3.41e-05) *** p<0.01, ** p<0.05, * p<0.1 237 Table . Final class counts and proportions for the latent classes based on their most likely latent class membership C1 Debt Servitude C2 Degrading Conditions C3 Weapons Class Proportio Class Proportio Class Proportio Count n Count n Count n Overall (Step 1) 142 13.21% 901 83.81% 32 2.98% Manual Criminal 135 13.83% 811 83.09% 30 3.07% Manual Labor 192 19.90% 745 77.20% 28 2.90% 238 Table 4. Regression of labor outcomes on latent classes of forced labor exploitation 1 2 Collective Individual Moral moral Damages in damages in $RS $RS N=171 N=131 C1 Debt Servitude (vs. C3 Weapons and surveillance) -94,099* -22,158 (48,160) (29,528) C2 Degrading Conditions (vs. C3 Weapons and surveillance) -105,690** -30,840 (44,591) (28,201) Region= Northeast -27,258 9,711 (22,235) (13,370) Region=North -7,525 2,533 (19,096) (11,328) Region-Southeast -29,509 19,545 (24,773) (13,485) Region=South -22,717 -13,675 (24,174) (13,457) Federal labor prosecutor was part of inspection team -6,804 -31,131** (17,105) (13,067) number of workers rescued during inspection 2,245*** 1,680*** (347.1) (210.1) Constant 138,647*** 61,745** (48,453) (30,449) R-squared 0.244 0.435 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 239 Table 5 Linear regression case length (days) on latent classes of forced labor exploitation, N=731 Case Length C1 Debt Servitude (vs. C3 Weapons and surveillance) -227.8 (186.6) C2 Degrading Conditions (vs. C3 Weapons and surveillance) -444.0*** (170.4) Police investigation took place -306.1*** (65.33) Number of other crimes indicted 72.33** (33.03) Criminal court region = 2 -551.2*** (191.6) Criminal court region= 3 -393.4*** (150.3) Criminal court region = 4 -358.9*** (112.3) Criminal court region = 5 -745.7*** (166.3) Constant 2,336*** (186.1) R-squared 0.116 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 240 Table 6 Logistic regression criminal outcomes on latent classes and covariates, odds ratios displayed 1 2 Convicted, Acquitted, N=976 N=976 C1 Debt Servitude (vs. C3 Weapons and surveillance) 1.169 1.059 (0.542) (0.453) C2 Degrading Conditions (vs. C3 Weapons and surveillance) 0.770 1.046 (0.331) (0.412) Police investigation took place 1.396** 1.404** (0.220) (0.189) Number of other crimes indicted 1.179** 0.942 (0.0887) (0.0628) Court Region = 2 1.051 1.060 (0.443) (0.385) Court Region = 3 1.486 1.082 (0.476) (0.318) Court Region = 4 1.425 3.508*** (0.387) (0.952) Court Region = 5 1.398 1.728 (0.543) (0.598) Constant 1.169 1.059 (0.542) (0.453) SE in parentheses *** p<0.01, ** p<0.05, * p<0.1 241 242 References Albonetti, C. (1986). CRIMINALITY, PROSECUTORIAL SCREENING, AND UNCERTAINTY: TOWARD A THEORY OF DISCRETIONARY DECISION MAKING IN FELONY CASE PROCESSINGS. Criminology, 24(4), 623. Alderman, J. (2017). Does auditor gender influence auditor liability? Exploring the impact of the crime congruency effect on jurors? perceptions of auditor negligence. Advances in Accounting, 38, 75?87. https://doi.org/10.1016/j.adiac.2017.07.006 Armour, J., Mayer, C., & Polo, A. (2017). Regulatory Sanctions and Reputational Damage in Financial Markets. Journal of Financial and Quantitative Analysis, 52(4), 1429?1448. https://doi.org/10.1017/S0022109017000461 Bakk, Z., Tekle, F. B., & Vermunt, J. K. (2013). Estimating the Association between Latent Class Membership and External Variables Using Bias-adjusted Three- step Approaches. Sociological Methodology, 43(1), 272?311. https://doi.org/10.1177/0081175012470644 Bouch?, V. (2017). An Empirical Analysis of the Intersection of Organized Crime and Human Trafficking In the United States. 128. Braithwaite, J. (2009). Restorative Justice for Banks Through Negative Licensing. British Journal of Criminology, 49(4), 439?450. https://doi.org/10.1093/bjc/azp038 243 Braithwaite, John. (2018). Restorative Justice and a Better Future 1. In John Braithwaite, Regulation, Crime, Freedom (1st ed., pp. 317?339). Routledge. https://doi.org/10.4324/9781315200361-15 Braithwaite, John, & Drahos, P. (2002). Zero Tolerance, Naming and Shaming: Is There a Case for it with Crimes of the Powerful? Australian & New Zealand Journal of Criminology, 35(3), 20. Brewer, D. (2009). Globalization and Human Trafficking. HUMAN RIGHTS, 11. Cria o Grupo Executivo de Repress?o ao Trabalho For?ado e d? outras provid?ncias, no. 1.538 (1995). http://www.planalto.gov.br/ccivil_03/decreto/Antigos/D1538.htm Costa, Patricia Trindade Maranhao. (2009). Fighting forced labour: The example of Brazil. International Labour Office Special Action Programme to Combat Forced Labour. https://www.ilo.org/wcmsp5/groups/public/---ed_norm/--- declaration/documents/publication/wcms_111297.pdf Costa, Patric?a Trindade Maranh?o. (2009). Fighting forced labour: The example of Brazil. Internat. Labour Off. DW. (2014). Brazil?s fight against slave labor. DW.COM. https://www.dw.com/en/brazils-fight-against-slave-labor/a-17664728 Eggleston, E. P., Laub, J. H., & Sampson, R. J. (2004). Methodological Sensitivities to Latent Class Analysis of Long-Term Criminal Trajectories. Journal of Quantitative Criminology, 20(1), 1?26. JSTOR. Farrell, A., Mcdevitt, J., & Fahy, S. (2008). Understanding and Improving Law Enforcement Responses to Human Trafficking, Final Report, 2008, June. 244 Human Trafficking Data Collection and Reporting Center Research and Technical Reports. Farrell, A., Owens, C., & McDevitt, J. (2014a). New laws but few cases: Understanding the challenges to the investigation and prosecution of human trafficking cases. Crime, Law and Social Change, 61(2), 139?168. https://doi.org/10.1007/s10611-013-9442-1 Farrell, A., Owens, C., & McDevitt, J. (2014b). New laws but few cases: Understanding the challenges to the investigation and prosecution of human trafficking cases. Crime, Law and Social Change, 61(2), 139?168. https://doi.org/10.1007/s10611-013-9442-1 Farrell, R. A., & Holmes, M. D. (1991). The Social and Cognitive Structure of Legal Decision-Making. The Sociological Quarterly, 32(4), 529?542. Portaria MIN no 1.150 de 18/11/2003, no. 1.150 de 18/11/2003, Departamento de Gest?o de Fundos de Desenvolvimento Regional da Secretaria de Pol?ticas de Desenvolvimento Regional do Minist?rio da Integra??o Nacional (2003). https://www.normasbrasil.com.br/norma/portaria-1150-2003_184483.html Go?dziak, E. M., & Bump, M. (2008). Data and Research on Human Trafficking: Bibliography of Research-Based Literature (No. 224392; p. 59). NCJRS. Guedes, P., Rodrigues, W., Paes, N., & Meressi, F. (2019). Plano PluriAnual 2016- 2019. Ministry of the Economy. https://www.gov.br/mma/pt-br/centrais-de- conteudo/relatrio-20final-20de-20avaliao-20-20volume-20i-pdf 245 Gunningham, N., Kagan, R. A., & Thornton, D. (2004). Social License and Environmental Protection: Why Businesses Go beyond Compliance. Law & Social Inquiry, 29(2), 307?341. Haddad, C. (2017). The Definition of Slave Labor for Criminal Enforcement and the Experience of Adjudication: The Case of Brazi. Michigan Journal of International Law, 38(3). Haddad, C., & Miraglia, L. (2018a). Trabalho Escravo. Haddad, C., & Miraglia, L. (2018b). Trabalho Escravo. Tribo da Ilha. Hawkins, D. F. (1981). Causal attribution and punishment for crime. Deviant Behavior, 2(3), 207?230. https://doi.org/10.1080/01639625.1981.9967554 Higgins, G. E., Vito, G. F., & Grossi, E. L. (2012). The Impact of Race on the Police Decision to Search During a Traffic Stop: A Focal Concerns Theory Perspective. Journal of Contemporary Criminal Justice, 28(2), 166?183. https://doi.org/10.1177/1043986211425725 Homer, E. M., & Higgins, G. E. (2020). Corporation blameworthiness and federal criminal fines. Journal of Financial Crime, 27(2), 413?422. https://doi.org/10.1108/JFC-01-2019-0010 Hussemann, J., Owens, C., Love, H., McCoy, E., Flynn, A., & Woods, K. (2018). Bending Towards Justice: Perceptions of Justice among Human Trafficking Survivors. National Criminal Justice Reference Service, 46. Inglis, S. (2001). Expanding International and National Protections against Trafficking for Forced Labor Using a Human Rights Framework. Buffalo Human Rights Law Review, 7(3). 246 https://digitalcommons.law.buffalo.edu/cgi/viewcontent.cgi?article=1141&co ntext=bhrlr International Labour Office (Ed.). (2005). A global alliance against forced labour: Global report under the follow-up to the ILO Declaration on Fundamental Principles and Rights at Work, 2005. International Labour Office. International Labour Office. (2012). ILO global estimate of forced labour: Results and methodology. International Labour Office. http://site.ebrary.com/id/10622497 International Labour Office. (2017a). Global estimates of child labour: Results and trends, 2012-2016. https://www.ilo.org/wcmsp5/groups/public/---dgreports/--- dcomm/documents/publication/wcms_575499.pdf International Labour Office. (2017b). Global Estimates of Modern Slavery: Forced Labour and Forced Marriage. International Labour Office. https://www.ilo.org/wcmsp5/groups/public/---dgreports/--- dcomm/documents/publication/wcms_575479.pdf International Labour Organization. (2001). Ratifications of ILO Conventions by Convention. NORMLEX Information System on International Labour Standards. https://www.ilo.org/dyn/normlex/en/f?p=NORMLEXPUB:11300:0::NO:1130 0:P11300_INSTRUMENT_ID:312327:NO Internationales Arbeitsamt (Ed.). (2009). The cost of coercion: Global report under the follow-up to the ILO Declaration on Fundamental Principles and Rights 247 at Work?; International Labour Conference, 98th Session 2009, Report I (B). Internat. Labour Off. Issa, D. (2017). Reification and the Human Commodity: Theorizing Modern Slavery in Brazil. Latin American Perspectives, 44(6), 90?106. https://doi.org/10.1177/0094582X17727480 Kappelhoff, M. (2008). Federal Prosecutions of Human Trafficking Cases: Striking a Blow Against Modern Day Slavery. University of St. Thomas Law Journal, 6(1), 9. Kelly, A. (2013, July 24). Brazil?s ?dirty list? names and shames companies involved in slave labour. Reporter Brasil. https://reporterbrasil.org.br/2013/07/brazils- dirty-list-names-and-shames-companies-involved-in-slave-labour/ Laan, P. van der, Smit, M., Busschers, I., & Aarten, P. (2011). Cross-border Trafficking In Human Beings: Prevention and Intervention Strategies for Reducing Sexual Exploitation. Campbell Systematic Reviews, 7(1), 1?50. https://doi.org/10.4073/csr.2011.9 Lynn, P. (2009). Methodology of longitudinal surveys. John Wiley & Sons. Marsh, E., Anthony, B., Emerson, J., & Mogulescu, K. (2019). Grading criminal record relief laws for survivors of human trafficking. Polaris. https://polarisproject.org/wp-content/uploads/2019/03/Grading-Criminal- Record-Relief-Laws-for-Survivors-of-Human-Trafficking.pdf May, P. J. (2005). Regulation and Compliance Motivations: Examining Different Approaches. Public Administration Review, 65(1), 31?44. https://doi.org/10.1111/j.1540-6210.2005.00428.x 248 Motivans, M., & Snyder, H. (2015). Federal Prosecution of Human-Trafficking Cases, 2015. 15. Nacoes unidas no Brasil. (2013). Position-paper-trabalho-escravo.pdf. https://brasil.un.org/sites/default/files/2020-07/position-paper-trabalho- escravo.pdf Nylund-Gibson, K., Grimm, R. P., & Masyn, K. E. (2019). Prediction from Latent Classes: A Demonstration of Different Approaches to Include Distal Outcomes in Mixture Models. Structural Equation Modeling: A Multidisciplinary Journal, 26(6), 967?985. https://doi.org/10.1080/10705511.2019.1590146 OSCE, Alliance against Trafficking in Persons Conference on Human Trafficking for Labour Exploitation, Forced and Bonded Labour, & Alliance against Trafficking in Persons Conference on Human Trafficking for Labour Exploitation, Forced and Bonded Labour (Eds.). (2008). Human trafficking for labour exploitation/forced and bonded labour: Identification - prevention - prosecution. OSCE. Owens, C. (2015). Understanding the organization, operation, and victimization process of labor trafficking in the United States. Trends in Organized Crime, 18(4), 348?354. https://doi.org/10.1007/s12117-015-9257-9 Ray, B., & Dollar, C. B. (2013). Examining Mental Health Court Completion: A Focal Concerns Perspective. The Sociological Quarterly, 54(4), 647?669. https://doi.org/10.1111/tsq.12032 249 Roberts, B. M. (2016). The Impact of Gender and Focal Concerns Theory on the Treatment of White-Collar Defendants by Federal Judges. Walden University Scholar Works, 148. Rorie, M., Alper, M., Schell-Busey, N., & Simpson, S. S. (2018). Using meta- analysis under conditions of definitional ambiguity: The case of corporate crime. Criminal Justice Studies, 31(1), 38?61. https://doi.org/10.1080/1478601X.2017.1412960 Santos, E. (2019, September 15). Grupo M?vel completa 24 anos como refer?ncia no combate ao trabalho escravo. Ministerio Da Economia. https://www.gov.br/economia/pt-br/assuntos/noticias/2019/05/grupo-movel- completa-24-anos-como-referencia-no-combate-ao-trabalho-escravo Regula o Programa do Seguro-Desemprego, o Abono Salarial, institui o Fundo de Amparo ao Trabalhador (FAT), e d? outras provid?ncias, no. 7998 (1990). http://www.planalto.gov.br/ccivil_03/leis/l7998.htm Shively, M., Smith, K., Jalbert, S., & Drucker, O. (2017). Human Trafficking Organizations and Facilitators: A Detailed Profile and Interviews with Convicted Traffickers in the United States. 23. Simpson, S. S., Gibbs, C., Rorie, M., Slocum, L. A., COHEN, M. A., & VANDENBERGH, M. (2013). AN EMPIRICAL ASSESSMENT OF CORPORATE ENVIRONMENTAL CRIME-CONTROL STRATEGIES. The Journal of Criminal Law and Criminology (1973-), 103(1), 231?278. 250 Spohn, C. (2014). The non-prosecution of human trafficking cases: An illustration of the challenges of implementing legal reforms. Crime, Law and Social Change, 61(2), 169?178. https://doi.org/10.1007/s10611-013-9508-0 Steffensmeier, D. J. (1980). Assessing the Impact of the Women?s Movement on Sex- Based Differences in the Handling of Adult Criminal Defendants. Crime and Delinquency, 26(3), 344?357. Steffensmeier, D., Kramer, J., & Streifel, C. (1993). Gender and Imprisonment Decisions. Criminology, 31(3), 411?446. Steffensmeier, D., Ulmer, J., & Kramer, J. (1998). THE INTERACTION OF RACE, GENDER, AND AGE IN CRIMINAL SENTENCING: THE PUNISHMENT COST OF BEING YOUNG, BLACK, AND MALE - STEFFENSMEIER - 1998?Criminology?Wiley Online Library. Criminology, 36(4), 763?798. Sutherland, E. H. (1945). Is ?White Collar Crime? Crime? American Sociological Review, 10(2), 132?139. https://doi.org/10.2307/2085628 The Interagency Coordination Group Against Trafficking in Persons. (2020). Non- punishment of victims of trafficking (Issue Brief). United Nations. https://www.un.org/sexualviolenceinconflict/wp-content/uploads/2020/04/19- 10800_ICAT-Issue-Brief-8_Ebook_final.pdf Ulmer, J. T. (2012). Recent Developments and New Directions in Sentencing Research. Justice Quarterly, 29(1), 1?40. https://doi.org/10.1080/07418825.2011.624115 United Nations Office on Drugs and Crime. (2009). Global Report on Trafficking in Persons. https://www.unodc.org/documents/Global_Report_on_TIP.pdf 251 United Nations Office on Drugs and Crime. (2018). Global report on trafficking in persons: In the context of armed conflict, 2018. United States Department of State. (n.d.). Countering Labor Trafficking: Two Case Studies. United States Department of State. Retrieved March 1, 2021, from https://www.state.gov/countering-labortrafficking-two-case-studies/ Vermunt, J. K. (2010). Latent Class Modeling with Covariates: Two Improved Three- Step Approaches. Political Analysis, 18(4), 450?469. https://doi.org/10.1093/pan/mpq025 Vermunt, J. K., & Magidson, J. (2020). How to perform three-step latent class analysis in the presence of measurement non-invariance or differential item functioning. Structural Equation Modeling. https://doi.org/10.1080/10705511.2020.1818084 Walk Free Foundation & International Organization for Migration. (2017). Global estimates of modern slavery: Forced labour and forced marriage. Warren, J. R., Luo, L., Halpern-Manners, A., Raymo, J. M., & Palloni, A. (2015). Do Different Methods for Modeling Age-Graded Trajectories Yield Consistent and Valid Results? American Journal of Sociology, 120(6), 1809?1856. https://doi.org/10.1086/681962 Working group on trafficking in persons. (2010). Non-punishment and non- prosecution of victims of trafficking in persons: Administrative and judicial approaches to offences committed in the process of such trafficking. United Nations. 252 https://www.unodc.org/documents/treaties/organized_crime/2010_CTOC_CO P_WG4/WG4_2010_4_E.pdf Yang, C.-C. (2006). Evaluating latent class analysis models in qualitative phenotype identification. Computational Statistics & Data Analysis, 50, 1090?1104. https://doi.org/10.1016/j.csda.2004.11.004 Yu, L., Hussemann, J., Love, H., McCoy, E., & Owens, C. (2018). Alternative Forms of Justice for Human Trafficking Survivors (p. 16). Urban Institute. 253