CAF - Criminal Dominance and Campaign Concentration

CRIMINAL DOMINANCE AND CAMPAIGN CONCENTRATION

January 2019 No. 2019/04

Bullock, J.

CRIMINAL DOMINANCE

AND CAMPAIGN

CONCENTRATION2 CRIMINAL DOMINANCE AND CAMPAIGN CONCENTRATION

CRIMINAL DOMINANCE AND CAMPAIGN

CONCENTRATION

Bullock, J.

CAF – Working paper No. 2019/04 January 2019

ABSTRACT

There are many journalistic and anecdotal accounts about the prevalence of electoral corrals in Brazil, geographic areas where brokers, politicians, or community leaders influence residents to vote for a specific candidate. In this paper, I investigate one particular type of suspected electoral corral: the favela, urban slum. This analysis focuses on the 1000+ favelas in the city of Rio de Janeiro, Brazil. I explore whether or not vote share is indeed more concentrated in u rban slums, and then whether or not vote concentration is related to criminal dominance. I contend that politicians in Rio de Janeiro have incentives to work with criminal groups in order to get more votes, and that finding a way to access these electoral corrals may be an election -winning strategy. Using novel, geospatial data and introducing a new text dataset on criminal dominance in Rio de Janeiro, I show that vote concentration is indeed more concentrated in urban slums and, within these slums, even more concentrated in slums that have steady criminal dominance from

one election to the next.

Small sections of text, that are less than two paragraphs, may be quoted without explicit permission as lo ng as this document is stated. Findings, interpretations and conclusions expressed in this publication are the sole re sponsibility of its author(s), and it cannot be, in any way, attributed to CAF, its Executive Directors or the countries they represent. CAF does not guarantee the a ccuracy of the data included in this publication and is not, in any way, responsible for any consequences resulting from its use.

© 2019 Corporación Andina de Fomento3 CRIMINAL DOMINANCE AND CAMPAIGN CONCENTRATION

PREDOMINIO CRIMINAL Y CONCENTRACIÓN

ELECTORAL

Bullock, J.

CAF - Documento de trabajo N° 2019/04 Enero 2019

RESUMEN

Existen muchos relatos periodísticos y anecdóticos sobre la prevalencia de corrales electorales en Brasil; áreas geográficas donde corredores, políticos o líderes comunitarios influyen en los residentes para que voten por un candidato específico. En este t rabajo, investigo un tipo particular de presunto corral electoral: las favelas, o barrios urbanos marginales. Este análisis se centra en las más de 1000 favelas en la ciudad de Río de Janeiro, Brasil. Analizo si la proporción de votos está o no más concentrada en los barrios urbanos marginales y luego si la concentración de votos está relacionada o no con el dominio criminal. Sostengo que los políticos en Río de Janeiro tienen incentivos para trabajar con grupos criminales para obtener más votos y que

de votos está o no más concentrada en los barrios urbanos marginales y luego si la concentración de votos está relacionada o no con el dominio criminal. Sostengo que los políticos en Río de Janeiro tienen incentivos para trabajar con grupos criminales para obtener más votos y que encontrar una manera de acceder a estos corrales electorales puede ser una estrategia electoral exitosa. Utilizando datos geoespaciales novedosos, e introduciendo un nuevo conjunto de datos en formato texto sobre predominio criminal en Río de Janeiro, muestro q ue la concentración de votos está más concentrada en los barrios marginales urbanos y, dentro de estos barrios marginales, aún más concentrada en barrios marginales que tienen un predominio criminal que se mantiene de una elección a la siguiente

Small sections of text, that are less than two paragraphs, may be quoted without explicit permission as lo ng as this document is stated. Findings, interpretations and conclusions expressed in this publication are the sole re sponsibility of its author(s), and it cannot be, in any way, attributed to CAF, its Executive Directors or the countries they represent. CAF does not guarantee the a ccuracy of the data included in this publication and is not, in any way, responsible for any consequences resulting from its use.

© 2019 Corporación Andina de FomentoCriminal Dominance and Campaign Concentration∗ Jessie Bullock† Abstract There are many journalistic and anecdotal account about the prevalence of electoral

corrals in Brazil, geographic areas where brokers, politicians, or community leaders influence residents to vote for a specific candidate. In this paper, I investigate one particular type of suspected electoral corral: the favela, urban slum. This analysis focuses on the 1000+ favelas in the city of Rio de Janeiro, Brazil. I explore whether or not vote share is indeed more concentrated in urban slums, and then whether or not vote concentration is related to criminal dominance. I contend that politicians in Rio de Janeiro have incentives to work with criminal groups in order to get more votes, and that finding a way to access these electoral corrals may be an electionwinning strategy. Using novel, geospatial data and introducing a new text dataset on criminal dominance in Rio de Janeiro, I show that vote concentration is indeed more concentrated in urban slums and, within these slums, even more concentrated in slums that have steady criminal dominance from one election to the next. ∗Prepared for CAF Working Paper Series. †PhD Candidate, Harvard University. jbullock@g.harvard.edu 11 Introduction In Complexo de Alem˜ ao, one of Rio de Janeiro’s largest urban slums and the headquarters of the Comando Vermelho (CV), Rio’s largest drug gang, select candidates are allowed by the drug traffickers to enter the neighborhood to run their campaigns. Once the selected candidates purchase entry, either through bribes directly to the criminal group (money, guns, other contraband, or favors) or through promises of goods and services to the community, they are allowed to run their campaign inside the slum. In the recent 2018 elections, journalists uncovered evidence of candidates’ staff offering to pay for doctor’s visits, government

bills, and even cataract surgery for slum residents if they promised to vote for the candidate (Corrˆ ea & Ará ujo 2018). When asked if politicians make good on these ex ante campaign promises, slum residents responded that “As soon as they win the election, the disappear. They throw out the chip in their cell phone, get a new number, and we never hear from them again”1. In 2018, approximately 1.7 million voters in the state of Rio de Janeiro (14.6% of the entire state) live in communities dominated by criminal groups and may be subject to this type of electoral propaganda by the select few candidates that are allowed to enter their neighborhood (Otavio & Araujo 2018). Journalistic accounts show that there is considerable variation in persuasive strategies in slums during the campaign: some candidates may offer promises to residents and simply not deliver as in the example above, some may be the “chosen candidate” by the criminal group and the only one that the residents of the slum are exposed to in their daily life, and some others may be directly endorsed by the criminal groups, who use force or the threat of force to convince voters to vote for their preferred candidate (Ará ujo & Corrˆ ea 2018). Despite the increasing frequency of journalistic accounts of corruption linked to organized 1Interview Transcript, August 8, 2017 2criminal groups, there is scant scholarly work on this phenomenon, especially quantitative work. It is difficult enough to measure corrupt vote buying, not to mention vote buying that involves criminal groups that conduct their entire business outside of the licit sector. This paper begins to try and measure this phenomenon by tackling one part of the criminalpolitician relationship: criminals as gatekeepers that control access to voters. Two research questions motivate this study and my larger research agenda. Is there evidence that voting is more concentrated in criminally dominated areas? How does consistent criminal dominance affect electoral outcomes in their neighborhoods? Though a corrupt exchange between politicians and criminal groups may be impossible to observe, the consequences of this exchange may be observable at the ballot box. I first look at the concentration of votes for a particular candidate as an observable indicator of criminal groups acting as gatekeepers for the voters in their community or, at a minimum, reducing the number of candidates that voters are exposed to a select few that purchase access to the community from the criminal group. I then look at changes over time in voting concentration in criminally dominated territories. In this paper, I analyze the vote concentration in slums exposed to criminal groups during the federal and state elections in Rio de Janeiro in 2014 and the municipal elections in 2016. Preliminary findings show that electoral returns are 6.7 percentage points more concentrated around the most voted candidate in urban slums than non-slums. Further examining the sample of polling stations where slum residents vote, the vote share for the most voted candidate is 8.2 percentage points higher in areas occupied by criminal groups and the vote share for the most voted candidate increases by 6.2 percentage points from election to election following a period of stable criminal dominance. This provides supporting evidence to the theory that criminal groups, either indirectly or directly, help restrict electoral competition and are the gatekeepers for electoral corrals. This paper is structured as follows. In the second section, I review the literature on traditional vote buying, and focus on situations where criminal groups play a role in the 3vote buying or campaign process. In the third section, I argue that the criminal landscape in Rio de Janeiro and the electoral institutions have created an opening for criminal to be involved in electoral campaigns. In the fourth section, I introduce the data and present estimates of voting concentration levels in slums versus non-slums. Section five presents

estimates of the effect of criminal dominance on electoral outcomes. There is a discussion and a concluding section. 2 Vote Buying and Criminality 2.1 Vote Buying with Traditional Brokers The is widespread evidence on the brokerage networks in Brazilian slums and how political machines are able to effectively demand neighborhood public goods from politicians in exchange for the votes in the community. Generally, the partisan brokers in Latin America provide positive inducements to voters who vote for their preferred candidate as a form of reciprocity (Gingerich 2013, Finan & Schechter 2012, Stokes, Dunning, Nazareno & Valeria 2013). In the Brazilian slums, there is already documented evidence of elevated levels of this behavior. Some community leaders that acted as brokers have been able to get politicians to pay for new neighborhood pools, new buildings, or other political favors before the elections even happened (Gay 1999), while other politicians engaged in targeted vote or even voter buying through positive inducements to the individual, and received incredibly high return on investment (Hidalgo & Nichter 2016, Gingerich 2013). 2.2 Vote Buying with Criminal Groups as Brokers On the contrary, very little recent literature has focused on the role of criminal organizations and militias in elections. Acemoglu et al (2013) show that former AUC criminals in Colombia 4were rewarded by politicians for delivering votes with an amnesty deal. This provides some evidence that the AUC were criminal brokers and used their coercive tactics to marshal people to the polls. They argue that the politicians had reduced incentives to fight or eliminate these criminal organizations because of their ability to be electoral brokers. A key part of Acemoglu et al’s (2013) theory, however, is the historic linkage between the AUC

paramilitaries and President Uribe’s party, as well as implicit or explicit linkages between new third parties and the AUC. In Brazil, Hidalgo and Lessing (2015) argue that strategic electioneering helped Rio de Janeiro’s militias rise to power. There is evidence the militias would choose one of their own to run for a particular office, extort residents in a slum or other unoccupied territory, and corral them to the polls. In this case, criminal groups were both the brokers and the candidates. What both of these examples miss, however, is explaining how these criminal groups became close to or partners with the politicians in the first place. In the case in Colombia, the relationship is path dependent and historic: the historic linkage between the AUC and the President’s party explains why both the government and the criminals were willing to enter into this exchange. In Brazil, the candidates were actually former criminals or affiliated with the criminals in one way or another, so they were “homegrown” candidates willing and able to cooperate with the criminals. In Brazil, we also hear about a third kind of politiciancriminal relationship that neither has a long history nor is one where the politician is already affiliated with the criminals. These are the gatekeeper, protection racket style relationships between criminal groups and politicians. 2.3 Vote Buying with Criminal Groups as Gatekeepers By referring to criminal groups as gatekeepers, I specify that they are not brokers or even necessarily politically motivated actors. When organized criminal groups are gatekeepers, I 5define them as having control or the ability to control the behavior of a group of people where they operate their criminal business. Just as Charles Tilly talks about the state and statebuilding in the legal sector as a protection racket, the same concept applies to gatekeeper

criminal groups that earn loyalty and protect residents within the legally sanctioned state (Tilly 1985). Classic examples of vote buying with criminal groups as gatekeepers are where criminal groups organize protection rackets and trade favors with political groups. This has been noticed in some of the oldest criminal groups alive today: the Italian mafia (Alesina, Piccolo & Pinotti 2017, Gambetta 1993, Pinotti 2015). Though the relationship between the mafia and certain political parties in Italy has grown considerably since its origin, the genesis of the relationship happened because politicians needed votes, the mafia had the loyalty of the people that they protected from rivals or from the predatory state, and the two parties gradually built trust, one delivering votes and the other turning a blind eye to illicit activity or even aiding it. Tajima identifies conditions under which modern day protection rackets form in democracies through an analysis of state strength in Indonesia (Tajima 2018). He argues that criminal groups will seek political affiliation in order to get the benefits of being a political ally, offering protection and coercive services in exchange for law enforcement to look the other way. He characterizes these settings as places where the state has the resources to constrain criminal groups, but lacks the institutional features to constrain politicians from engaging with them. In Brazil, the situation is slightly different. The state has but squanders the resources to constrain criminal groups, and lacks the institutional features to constrain politicians from engaging with them. To be fair, the constraints on politicians have improved in recent years with the high-profile Lava Jato trials and new laws like Ficha Limpa and tighter campaign finance regulations2. As I show in the next section, I believe the aforementioned inability 2The Lava Jatotrials put a number of high profile politicians, including former President Lula In acio de

6to control the criminal groups and lax constraints on politicians, all under an electoral system that creates perverse incentives, creates a new dynamic between criminal groups and politicians where the politician will be the one seeking criminal affiliation for electoral gain. I argue these politicians will seek criminals not to serve as traditional brokers, but rather to serve more as gatekeepers or protection rackets to keep other politicians out of their desired territory and away from potential voters. 3 Setting: Rio de Janeiro In this section, I discuss two factors that complicate our understanding of the role of criminal groups in electoral politics in Rio de Janeiro. The first is the landscape of criminality in the city and state of Rio de Janeiro and how territorial conflict relates to electoral politics. The second is the institutional features of electoral politics in Brazil, and how the single district, open list system creates incentives that I argue make politicians willing and open to establishing protection rackets with criminal groups. I describe the institutional context below and present my hypotheses. 3.1 Criminality in Brazil An analysis of organized criminal activity in Rio de Janeiro would be incomplete without understanding the unit of analysis: the urban slum, the favela. It is important for understanding the evolution of Rio’s criminal groups, but also for understanding the territories they fight over, dominate, and influence voters in the present. Rio’s first urban slum, Morro da Providˆ encia,was constructed in 1897 on a steep hill near the city center and still stands today (Meade 1996). As waves of Brazil’s rural poor moved to Providˆ enciaand as new slums were built on Rio’s steep hillsides, the word favela became synonymous with these informal Silva, in prison for money laundering or other white collar charges. Ficha Limpais a regulation barring any

candidate from running for office if they have a criminal record of corruption or electoral-related crimes. 7settlements high on the hills. During the first wave of urban migration in the 1920s, rural Brazilians and immigrants moved en masse to the densely populated favelas, characterized by few paved roads, citizen insecurity, extreme poverty, and, importantly, steep, hilly terrain. The public sector response to the growth of favelas in the early 20 th century was social exclusion. Today, this is still apparent: the infrastructure is still decades behind in many favelas, and in some areas nearly everything is still operated entirely in the informal sector. Today, there are 700+ favelas in the municipality of Rio de Janeiro and favela residents make up nearly a quarter of the city’s 6.3 million people. The favelas are scattered across the entire city, present in both the wealthy, touristic parts of the city and the more homogenous, lower-income areas. The Brazilian Institute of Geography and Statistics (IBGE) keeps records of all favelas in census data, and collects additional measures in areas they define as favelas. IBGE defines a favelas as “an area that is predominantly residential, characterized by informal employment and low income, precarious urban infrastructure and public services, unpaved roads and irregular buildings, the absence of formal property rights and construction conducted within the formal sector, which does not comply with current legal norms.” (Instituto Brasiliero de Geografia e Estadisticas 2010). Figure 1 shows the location of favelas recognized by IBGE inside the municipality of Rio de Janeiro. All of these conditions make favelas ideal places for criminal factions to conduct business. The proliferation of organized crime and violence in Rio’s favelas during the late 1970s

and early 1980s has been attributed to three key societal events: the most brutal years of state violence during the military dictatorship, emergence of the Brazilian cocaine trade and narcotrafficking, and deepening social and economic exclusion of favela residents (Misse 2009, Penglase 2008). The brutal state repression and increased incarceration rates – both of common criminals and political opposition – during the military dictatorship led to the birth of Brazil’s first criminal organization. The Comando Vermelho (CV), or Red Command, developed as 8Figure 1: Favelas in the Rio de Janeiro Municipality All white polygons on the map are the favelas in the municipality of Rio de Janeiro, which is subdivided by administrative boundaries. Source: Instituto Pereira Passos, IPP. political prisoners, usually well educated, taught ideological and organizational lessons to common thieves and prisoners while organizing a mass prison break (Penglase 2008). The leaders emphasized a united opposition against the prison guards, a common enemy, and an end on prisoner-against-prisoner crime. After the fall of the dictatorship in 1985 and as common prisoners were released or escaped, many took the central tenets of the CV back to the favelas where they were from. The CV remained an organization in the Rio de Janeiro state prison where it originated, but membership spread to the favelas and nearby prisons (Penglase 2008, Misse 2009, Imbusch, Misse & Carrí on 2011). Two rival criminal groups grew in the 1990s to compete with the CV: the Third Pure Command (TCP) and the Friends of Friends (ADA). The TCP grew out of a leadership dispute between leaders of the CV and a dissatisfied leader formed his own faction. ADA grew as non-CV favelas residents were witnessing the lifestyles of CV traffickers and the

destruction that crack cocaine caused in their communities. ADA, much smaller than the 9other two, quickly grew to control one of the city’s most desirable favelas, Rocinha, through collusion and corruption with politicians and city law enforcement, specific strategies towards protecting women and children in the favela, and distributing goods to the favela residents where they dominated. One of the most infamous traffickers for ADA, “Nemesis,” orchestrated these policies and attempted to signal to law enforcement that ADA was not as violent as the CV, did not engage in the sale of crack cocaine, and was more willing to bargain with the police (Glenny 2015). The 1990s and early 2000s also saw the growth and expansion of civilian militias across the city. These were composed primarily of disillusioned police, military, and firemen that were frustrated with the security situation and wanted to take justice into their own hands (Hidalgo & Lessing 2015, Imbusch, Misse & Carrí on 2011, Glenny 2015). They began behaving more like criminal bandits themselves, extorting citizens for electric, cable television, and utility payments under the auspices of providing protection from traffickers. Reporters have discovered a militia presence in some public housing complexes where they extort the poor for utilities and sometimes even rent payments. Although militias have only started engaging in the drug trade recently, they are no longer condoned by the government and are regarded as a second type of criminal group present in the city, whose criminal behavior is quite different from that of the drug trafficking gangs. Today, the militias, CV, TCP, and ADA all compete for territory among the many favelas in Rio de Janeiro. Nemesis is in jail, as are many of the stronger leaders for the

other trafficking factions. The three drug gangs are similarly characterized by high turnover of leadership, and high levels of violence with rival gangs or militias and the police. The militias have grown considerably in the last two decades, and they too are violent when fighting drug gangs for territory, but have fewer noted conflicts with the police. The birth of the CV under the repressive military dictatorship, the growth of the drug trade and subsequent violence once the CV monopolized Rio’s cocaine trade, and the forbearance (or 10lack of state capacity) of state actors in favelas towards all criminal groups are all argued to be contributing factors towards the high rates of organized crime in favelas today and how they influence and interact with their communities. 3.2 Electoral Institutions in Rio de Janeiro Brazil is famous for weak party system institutionalization, which implies that parties may not have easily identifiable ideologies, they may have short birth and death cycles, it is easy for politicians to party switch, and parties can win elections when they are particularistic rather than programmatic (Mainwaring 1998). All of the above is true in Brazil. Executive elections operate on the two round runoff system, and legislative elections are single district, open list, proportional representation. This means that citizens vote for a specific candidate, but the number of candidates that are elected depends on the entire proportion of votes the party gets. The single district legislative elections means that all representatives of the state of Rio de Janeiro, either in their state legislature or that represent the state in the national legislature are “at large,” for the approximately 13 million residents of the state, or do not represent any one particular district. These electoral rules create confusion and opportunism for a variety of reasons. First, there are simply too many parties. Just this year in the 2018 lower house elections, deputies

from 30 political parties won seats across the country. The system is inefficient and makes it difficult for parties to clearly distinguish themselves from each other. Second, because party systems have never consolidated in Brazil and there are so many parties that have similar ideologically vague platforms, voters claim to vote based on the individual candidate rather than follow a party line (Cohen, Lupu & Zechmeister 2017). This means that politicians can misbehave within their party, change their ideological stance once they are elected, or even switch parties entirely – as long as they do not anger their voters and maintain a cult of personality, candidates can (and have!) enjoy high levels of support across a range of 11party affiliations and ideological stances. Lastly, because of the freedom candidates have, the plethora of parties to affiliate with, and the (illegal) returns to office that many are aware of, far too many people decide to run for office. In the last ALERJ (the State Legislative Assembly of Rio de Janeiro) election in 2014, 2000+ candidates competed for 70 seats in the state. Four candidates won by a large margin, and the 66 remaining winning candidates won by a few thousand or even hundred votes compared to those that barely lost. This means that the margin of victory is razor thin. For those candidates near the margin, just one favela’s votes could make the difference between winning and losing. Strategically, the candidates would reach more voters targeting campaigns towards the densely populated favelas than any other geographic area. These areas may also be desirable because the voters are poor, and as we know from traditional clientalist literature, poor voters are cheaper to buy than middle class voters. This is where candidates may face the dilemma of engaging with criminal groups: is it worth it to engage with criminal actors if they can

secure the vote of a community dominated by criminals? Or if not secure it, ate least be one of a small number allowed to campaign inside? I argue that this tradeoff is what Rio de Janeiro’s political candidates are considering, and one way we can observe if there actually is a higher return for campaigning in favelas or not is by observing concentration of votes in the ballot box. Without making any assumptions about criminal groups buying votes, forcing candidates on the voters, extorting them, or anything in this vein, we don’t actually know how aggregate voting patterns in slums compares to voting patterns in similarly poor, non-slum neighborhoods. If there are differences in how concentrated vote levels are in favelas versus non-favelas, this could lead to follow up questions about why favela residents vote differently, how criminality may play a role, and how campaigns are run in favelas. 123.3 Why Rio de Janeiro? Given the information laid out above about electoral institutional design and the current state of criminality in Rio de Janeiro, I believe that, despite the risks, there are incentives on the part of the politician to attempt to enter urban slums in Rio de Janeiro to buy votes, and to use the criminal groups as gatekeepers in doing so. I contend that: H1: Voters are more likely to converge around a single candidate when they live in an urban slum. H2: Slum voters that live under clear criminal presence are more likely to converge around a single candidate than slum voters where criminal presence is unclear. H2a: Slum voters that have experienced continued dominance in their criminal group will continue to consolidate their vote around a single candidate compared to their past electoral behavior. I limit this analysis to the city of Rio de Janeiro initially for a few reasons. First, Rio de Janeiro, unlike other parts of Brazil where vote-buying may be present, has poverty and

criminality essentially randomly distributed across the city due to the geographic dispersion of the slums. The urban slums are scattered across the city, unlike S˜ ao Paulo or some other large cities in Latin America where the urban poor mainly resides in the periphery. This is useful for comparing voters in the sample and reducing confounding. For instance, a member of the urban poor in a slum in the North Zone of the city may be very similar to a member of the urban poor that lives in a lower class neighborhood (but not a slum) in the North Zone of the city. It is possible they have similar household incomes and they take the same bus line in to the center of the city. Leveraging the random placement of slums across the city provides a degree of control over the sample especially with respect to other issues areas that voters care about, such as public services. Second, there is considerable variation in which candidates get the most votes from slum residents in Rio de Janeiro. In other words