Voting under criminal governance: election mobilization by criminal organizations

Voting under criminal governance: election mobilization by criminal organizations | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Voting under criminal governance: election mobilization by criminal organizations Igor Novaes Lins This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9534248/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Why do some armed criminal groups mobilize voters while others depress participation? Existing accounts treat criminal governance as electorally uniform, rarely distinguishing forms that differ in political integration into local electoral networks. I argue that the electoral effects of criminal governance depend on how deeply armed groups are embedded in those networks: durable ties with electoral elites, stable candidate alliances, and accumulated voter mobilization capacity. Politically integrated criminal governance increases participation by organizing territorial brokerage, candidate access, and election-day coordination. Peripheral criminal governance produces weaker, more unstable effects, including suppression. Using geocoded armed-control polygons and polling-place electoral returns across five municipal elections in Rio de Janeiro, I estimate the effects of exposure to different criminal groups on turnout and voter registration within a panel design with neighborhood and year fixed effects. Militia-controlled areas exhibit significantly higher turnout and voter registration, with effects that attenuate under lethal violence but amplify where police forces overlap with paramilitary networks. Drug-trafficking factions generate weaker, heterogeneous, and sometimes negative associations. These findings recast criminal governance as a heterogeneous political order whose electoral effects depend on territorial brokerage and political embeddedness, not merely coercive capacity. Comparative Political Science criminal governance electoral participation militias drug-trafficking factions criminalized electoral politics Figures Figure 1 Figure 2 Introduction Criminal organizations govern elections. Voters stay away from the polls out of fear (Trelles & Carreras, 2012), responding to criminal activity beyond direct victimization (Ley, 2018 ), or facing political violence (Van Baalen, 2024; Lins, 2023 ). This governance goes beyond demobilization. Criminals mediate candidate access to voter groups in urban peripheries (Arias, 2013 ; Trudeau, 2025), condition the preferences of voters living under armed control (Pessoa, 2023 ; Hidalgo & Lessing, 2015), restrict political campaigns (Arias, 2006 ; Bullock, 2019 ), and systematically structure electoral dynamics (Córdova, 2019 ; Accardo; De Feo; De Luca, 2023). I ask why groups with similar territorial control produce different political effects on elections. These effects depend not only on armed control, but on the political governance criminal organizations accumulate over electoral processes over time. Rio de Janeiro offers a singular analytical opportunity. Paramilitaries rooted in police forces and drug-trafficking factions with distinct organizational trajectories coexist within the same municipal electoral system. This allows comparison across modes of criminal governance without the institutional confounders of cross-national designs, isolating political integration as the explanatory dimension. In the city, politicians, police officers, and local administrators maintain longstanding alliances with criminal organizations. These ties regulate who can campaign and channel votes through selective coercion and protection. Electoral strongholds in paramilitary-controlled neighborhoods such as Gardênia Azul and Rio das Pedras show how political families tied to criminals expand their vote share where armed groups govern (Pantaleão & Montini, 2025). They signal voter access and the capacity to mobilize through coercion or clientelist relations (Trudeau, 2025). This pattern extends beyond episodic scandals of violence. Across Latin American cities, armed actors and political elites form durable agreements that intertwine illicit markets, local bureaucracies, and electoral strategies. These configurations allow criminal groups to influence who participates, who runs, and who wins. Democratic representation comes to reflect the degree of political governance exercised by criminal organizations. This produces systematic advantages for embedded actors (Pessoa, 2023 ) and persistent barriers for those who oppose them (Albarracín, 2018 ; Albarracín, Karolczak, & Wolff, 2025 ). After elections, these arrangements shape the distribution of public offices and guide policymaking (Arias, 2013 ; Pessoa, 2023 ). I argue that criminal organizations vary systematically in their degree of political involvement and in the mechanisms of electoral intervention. Some criminal organizations maintain durable ties with political elites, mediate access to voters, and mobilize them. They participate directly in campaigns by promoting or imposing a candidate. Other groups operate with more limited political structures, imposing selective restrictions, blocking rivals, and intervening only when threatened. From this variation, I propose a typology with three levels of political governance. Politically integrated organizations participate systematically in electoral processes. Politically peripheral organizations interfere episodically and reactively. Hybrid configurations identify contingent opportunities for electoral action according to context. I show that political integration, defined by durable ties with electoral elites, stable candidate alliances, and accumulated mobilization capacity, predicts the political effect of territorial control. Where governance is politically integrated, electoral competition is compressed and candidates outside criminal networks face access barriers that convert territorial domination into political exclusion. With approximately 7,000 polling station observations across five municipal elections, the empirical strategy combines regression with two-way fixed effects, interaction models, and a stacked differences-in-differences design. Militia presence is associated with higher participation and voter registration. This effect attenuates as lethal violence increases but is amplified in areas with active Pacifying Police Units (UPPs). Comando Vermelho and Amigos dos Amigos are associated with declines in participation. Terceiro Comando Puro shows unstable effects near zero, consistent with the prediction for the hybrid case. Militia exposure predicts territorial dependence of candidates' electoral bases, and UPP entry shocks disrupt these ties. Qualitative evidence shows how militias organize the vote logistically, from directed registration to election-day transportation. Criminal governance and elections Criminal governance shapes elections through territorial control, voter mobilization, and restricted neighborhood access. In Rio de Janeiro, these practices let criminal groups intervene politically without formal party ties, expanding their independence and influence. Criminal actors interact directly with voters, often outside institutional channels (Trudeau, 2025). Their capacity to impose rules and fill governance gaps reflects a broader pattern in which selective state absence, institutional weakness, or collusion allows illicit groups to exercise political power (Lessing, 2021 ). Recent research shows how these groups regulate community behavior (Feltran, 2019), restrict elected authorities (Córdova, 2019 ), and influence police operations (Misse, 2007 ). Their interactions with formal institutions alter democratic governance (Blattman et al., 2021; Melnikov et al., 2022; Manso, 2020 ). This governance arises through networks connecting criminals, community leaders, elected politicians, and law enforcement agents. These networks protect traffickers and paramilitaries from repression and integrate favelas and urban peripheries into the local political system (Arias, 2006 ). These groups coerce or buy votes and supply services typically associated with the state or the private sector, such as cooking gas and public transportation. According to Lessing (2020), despite continuous repression, these groups seek recognition. They achieve it through practical authority, acquired by regulating violence and social life (Lins & Machado, 2023 ; Lessing & Willis, 2019). From this interaction comes 'violent democracy' (Arias, 2006 , 2013 ). Arias describes systems in which armed criminal actors routinely intervene in elections, negotiate directly with public authorities, and integrate into local political structures. Criminal groups differ from other political mediators in their capacity to coerce voters beyond the limits of partisan institutions. They reach areas other intermediaries cannot, especially zones affected by violence. In these contexts, they press for more votes without facing significant institutional costs (Albarracín, 2018 ; Trudeau, 2022 ). This influence stems from their role in criminal governance. Leaders build social capital through kinship ties (Misse, 2007 ) and create monitoring networks based on extortion and surveillance (Arias, 2013 ). Electoral behavior does not result only from voters' responses to violence or the presence of criminal groups. The active intervention of these groups as vote brokers also conditions it. Criminal governance goes beyond occupying a territory. It requires authority, the regulation of political life, and the delivery of services, including political ones, for residents under its influence (Trudeau, 2022 ; Lins & Machado, 2023 ; Manso, 2022). Criminal organizations use politics strategically. Understanding how illicit activities overlap with formal structures, and how organized crime sustains itself through political contracts, is essential (Pantaleão & Montini, 2025). Politics becomes a commodity in the illicit economy of armed groups (Lins & Machado, 2023 ; Zaluar & Conceição, 2007; Arias, 2013 ). They become efficient political intermediaries because governing marginalized urban territories develops these capacities (Trudeau, 2022 ; Pessoa, 2023 ). Beyond violence, research shows that criminal groups mobilize voters by buying votes or persuading them that certain candidates will help the community while others may cause harm (Manso, 2020 ; Arias, 2013 ). Because these groups meet residents' basic needs, their promises and threats carry credibility. Once established, the alliance between criminal actors and politicians spreads rapidly across the territories under their control (Trudeau, 2022 ). Politicians pursue office through criminally supported electoral structures while criminal actors protect or expand illicit profits, producing mutually beneficial relations (Albarracín, 2018 ). In some cases, the lines between crime and politics blur into a stable arrangement that does not depend on electoral cycles alone, though elections remain a central asset. Criminal groups manage voter access and mobilize participation in exchange for tangible benefits. Politicians, in turn, offer public goods or private advantages through these networks, positioning criminal actors as electoral intermediaries (Albarracín, 2018 ). Both sides calculate the value of intermediation by the results it produces. For candidates, higher participation reduces the cost of bringing voters to the polls. For intermediaries, controlling voter access and blocking rivals' information increases the chances of electing aligned candidates (Trudeau, 2022 ; Manso, 2020 ; Arias, 2013 ). Criminal groups depend on these practices to reinforce authority and expand influence relative to the state (Lins & Machado, 2023 ). They use violence and illicit resources to co-opt community leaders and alter electoral outcomes (Arias, 2006 ; Arias, 2013 ). Albarracín ( 2018 ) calls the situation where organized crime violence reduces candidates and reshapes voter demand 'criminalized electoral politics.' These groups frequently serve as political intermediaries who exchange goods and services for votes (Cano & Duarte, 2012 ; Zaluar & Conceição, 2007; Arias, 2013 ; Manso, 2020 ). Paramilitary groups and factions, my focus here, engage with the political system through distinct forms and levels of governance. Paramilitary groups, typically composed of active and former agents of state repressive forces, maintain formal and stable ties with local politicians (Daly, 2022 ; Manso, 2020 ; Arias, 2013 ). To legitimize their authority and intervene in electoral processes, they offer security services, public goods, and conflict mediation (Arias, 2006 ; Trudeau, 2022 ). Drug-trafficking factions display more fluid and situational political connections. Their actions concentrate on territorial domination and protection of illicit markets (Misse, 2011 ; Duarte, 2019). Paramilitary actors actively promote allies and suppress rivals. Factions prioritize local hegemony and do not always participate directly in electoral contests (Arias, 2013 ; Lins & Machado, 2023 ; Hirata et al., 2022). These groups also differ in their relationships with the communities they control. Drug factions treat the local population as an operational base and source of legitimacy, though their primary markets are often outside the territory they govern (Misse, 2007 , 2011 ). Paramilitary networks, in turn, extract resources directly from residents by charging fees, selling services, and exercising social control. This structure reflects a clientelist logic rooted in police officers who operate in the same neighborhoods (Cano & Duarte, 2012 ). The guiding hypothesis is that these differences produce distinct patterns of electoral participation and registration. Paramilitary control raises participation and voter registration. Factions show more volatile patterns, as previous studies suggest. Electoral violence by armed actors discourages voters by raising the cost of voting, especially when targeting electoral authorities or polling stations (Ley, 2018 ; Van Baalen, 2024). It undermines trust in elections, weakens beliefs in the efficacy of the vote, and reduces participation in violent environments (Trelles & Carreras, 2012; Ley, 2018 ). In Colombia, territories under paramilitary control, analogous in function to Brazil's militias, reduce competition. In those areas, the pressure to vote outweighs the risks. Guerrilla groups, like Brazilian factions, tend to suppress participation by disrupting the electoral process (Gallego, 2018). Coercion and material incentives, such as favors or community benefits, create distinct modes of mobilization (Accardo, De Feo, & De Luca, 2023). Participation helps reinforce local authority and legitimize candidates aligned with criminal structures. The literature on Italian mafias shows that organized crime affects candidate selection and electoral competition (Baraldi et al., 2024 ; Daniele, 2019 ). This article shows that, with multiple types of armed actors, effects on participation and electoral organization vary with political integration, a dimension underplayed by literature focused on single actor types. Shocks to territorial control produce different effects by type. Lethal violence and armed competition between groups destabilize the mobilization infrastructure. They erode the territorial control that sustains electoral brokerage networks. Institutional police presence can coexist with this infrastructure or even complement it when formal forces and paramilitary groups share personnel. The first shock attenuates the militia effect on participation. The second coexists with it and may amplify it under institutional overlap. This leads to my hypotheses: Hypothesis 1 Territories under paramilitary control will exhibit higher electoral participation than those without criminal governance or under drug faction control. Hypothesis 2 Territories under paramilitary control will exhibit higher voter registration than those without criminal governance or under drug faction control. The type of criminal organization, its political engagement, and local context shape the outcome. Clear evidence on how these differences affect voting behavior remains scarce, and the literature lacks consensus. Building on this distinction, I expect results consistent with my hypotheses. Criminal governance and modes of electoral incorporation Why do criminal actors support politicians? Political alliances help stabilize illicit markets against state regulation, secure protection (often through local police), and grant access to drug-trafficking routes. Why do politicians seek criminal groups? Among other reasons, to reach voters under their direct political control where electoral competition is candidate-centered rather than party-centered (Albarracín, forthcoming). Through criminal control as a mechanism of political mediation, candidates benefit from concentrated voter mobilization capacities. Even individual coercion targets enough vote shifts to alter local results. Criminal governance conditions patterns of electoral participation. This effect depends on how criminal governance interacts with pre-existing political arrangements, which structure opportunities for electoral engagement. Paramilitary networks, more dependent on the state, and drug-trafficking factions operate under distinct political arrangements, producing divergent effects on participation. The degree of electoral governance varies across contexts, as Fig. 1 illustrates. Source: author's elaboration. I identify three types of political governance structures that produce distinct forms of electoral intervention. The least common, politically integrated criminal governance, is frequently linked to paramilitary groups that systematically control electoral processes over time, from voter registration through election-day mobilization. Their role goes beyond coercion. They allocate economic revenues to local leaders, who connect voters and candidates through neighborhood associations and clientelist ties. The second configuration, politically peripheral governance, describes organizations that remain outside integrated political arrangements. These groups intervene only in response to specific threats, using vetoes to block access to politicians aligned with rivals or perceived adversaries. Without the incentives or networks for sustained engagement, they raise the costs of political participation and tend to suppress turnout. From tactical agreements between peripheral and integrated criminal organizations, a hybrid configuration emerges. Peripheral groups operate under integrated electoral arrangements, generally mediated by politically integrated actors. These groups lack stable candidate alliances and rarely develop autonomous mobilization structures. Their electoral involvement is contingent, emerging through occasional alignment with integrated actors who provide candidate access and co-governance, as Fig. 2 shows. Source: author's elaboration. I classify the groups a priori, drawing on qualitative literature about their organizational structures and political ties. Militias represent politically integrated governance. CV and ADA represent peripheral cases. TCP represents the hybrid case, given the historical overlap of its networks with militias in the West Zone. For the hybrid type, I predict inconsistent effects across specifications, with low magnitude and unstable direction. Effects therefore depend on which mode of political governance dominates in each area: integrated groups raise participation, peripheral groups depress it, and hybrid arrangements cannot sustain mobilization over time. The process unfolds in two stages. The first establishes local political control. The second mobilizes voters and sets the incentives they face. In some areas, criminal actors exercise uncontested authority. In others, they contend with ongoing disputes involving the state or rival organizations. When control is stable, criminal groups can engage systematically in elections. In areas marked by instability or conflict, political arrangements are often fragile or altogether absent. Groups under politically integrated governance are directly present in the political arena. They control local economic flows, charge for services, regulate movement, and determine who can participate in community networks. During elections, they suppress competition by blocking rival campaigns and directing votes toward allied candidates (Trudeau, 2022 ). Voters are incentivized through clientelist ties or coerced into voting, and coercion is sustained over time. These groups support voter registration, monitor who voted, and intimidate residents. They apply sanctions when necessary or suspend access to public goods they control. Control involves material incentives such as benefits for merchants, infrastructure promises, and service delivery, often mediated through residents' associations. In some cases, they receive payments to authorize campaigns or impose electoral discipline. Together, these practices form an electoral governance with consistent mobilization capacity. Groups under politically peripheral governance, by contrast, intervene in politics only reactively. The absence of stable political structures and the focus on self-protection condition their logic of interference in the electoral process. They impose vetoes, restrict movements, obstruct campaigns, and limit the number of candidates. Criminal presence reduces political options, increases perceived risks, and tends to suppress electoral participation. When mobilization occurs, it responds to specific disputes and lacks continuity. The control strategy goes beyond managing illicit markets and monopolizing state presence. Community leaders treat threats to criminal authority as dangerous, reinforcing a clientelist system where residents trade loyalty for basic services. For politically integrated groups, selling services and cultivating political ties follow a market-preservation logic. Violence is explicit and instrumental, used to expand political influence and secure aligned candidates without undermining the conditions needed for voting. Non-electoral criminal organizations rely on more indirect control, prioritizing territorial defense and neutralizing threats. Some areas feature unstable political governance, where disputes between groups generate volatility and fuel recurrent violence. This instability raises voting costs and undermines campaign and voting conditions. In these contexts, participation remains persistently low, shaped by violence. The second stage concerns mobilization and incentives. Under politically integrated governance, coercive and clientelist networks maintain continuous relations with candidates, pressure voters to vote, and monitor election-day presence. Voters may be rewarded or punished. By limiting competition, the networks reduce cognitive costs and make support for the endorsed candidate more predictable. In contexts with hybrid political configurations, electoral interference is selective and contingent. The effects on participation vary with the degree of alignment with integrated actors, the nature of ties with candidates, and residents' perceived risks, which form through their understanding of governance over time. These groups limit available options and occasionally mobilize to block rivals. Interventions are typically episodic. Organized crime and elections in Rio de Janeiro The typology serves as a schema of expectations for interpreting the heterogeneity of results. I treat militias as the closest case to politically integrated governance, CV and ADA as politically peripheral, and the TCP, whose territorial networks historically overlap with militias in parts of the West Zone, as the hybrid case. Between 2002 and 2003, the Rio das Pedras Residents' Association organized voter registration campaigns in a neighborhood populated predominantly by migrants from northeastern Brazil. The initiative aimed to elect a 'favela candidate,' a label adopted by paramilitary leader Nadinho. Backed by local leadership, Nadinho had already run for the state legislature in 1998, framing his campaign around the need for political representation to address local demands. The strategy proved effective. With 3,624 votes, he became an alternate member of the PTdoB (Zaluar & Conceição, 2007) and was later elected city councilman with strong community support. Members of the residents' association went door to door to convince neighbors of the importance of electing a representative from the periphery. Once convinced, residents were directed to designated meeting points, where vans waited to transport them to the city's electoral registry offices (Zaluar & Conceição, 2007). The message centered on basic needs: lack of sanitation, infrastructure, and public services. Community members trusted the association and its intermediaries despite their known ties to armed groups. These campaigns successfully mobilized a significant number of new voter registrations. Two decades later, the data show the scale of criminal governance over Rio's electorate. Of the more than 4.9 million registered voters in 2020, approximately 1.4 million, or 29.38% of the electorate, lived in areas under paramilitary control. Another 15.25% resided in territories controlled by Comando Vermelho (CV), while 5.50% lived in areas dominated by Terceiro Comando Puro (TCP). Another 5.11% were in zones disputed by rival groups. Only 41.68% of voters lived in neighborhoods without reported criminal control. In total, nearly 60% of the city's voters were exposed to environments where coercive electoral strategies could be applied or some form of criminal governance operates. The main criminal organizations operating in Rio de Janeiro are Comando Vermelho (CV), Amigos dos Amigos (ADA), and Terceiro Comando Puro (TCP). Paramilitary groups, commonly called militias in Brazil, compete for territory and control local markets through extortion schemes targeting essential services such as water, electricity, internet, and informal urban transportation (Arias & Barnes, 2016). Unlike drug factions, which focus on retail drug and firearms trafficking, militias are composed largely of active or former state agents, including police officers and firefighters. They exploit these institutional ties to expand both political and economic influence (Zaluar & Conceição, 2007; Manso, 2020 ). These groups follow distinct strategies of territorial control and political engagement. Factions rely on direct coercion and maintain relations with security forces through corrupt agreements involving extortion (Misse, 2011 ; Hirata; Grillo, 2017). Militias benefit from institutional protection by state actors who enable their integration into formal politics and shield their illicit operations (Hirata, Cardoso, & Grillo, 2022). This control lets militias convert neighborhoods into electoral strongholds and secure seats for allied candidates (Motta, 2024). Factions frequently negotiate with candidates to secure campaign access in the territories they dominate (Arias & Barnes, 2016). They also regulate illegal markets through informal coercion and negotiated order (Hirata & Grillo, 2017). Militias, for their part, maintain organic ties with politicians at different levels of government, pursuing their interests through legislative influence and electoral success (Hirata, Grillo, & Telles, 2022). Militiamen are also known for their connections within the judicial and legislative branches (Arias, 2013 ). Coercion is central to the electoral strategies of both types of groups. Paramilitary actors impose threats, restrict campaign activity, and apply targeted pressure to ensure residents vote for their preferred candidates. Armed presence on the streets is a form of intimidation. Research has documented threats of eviction for residents who fail to comply with voting instructions, as well as efforts to block rival campaigns in contested areas (Hidalgo & Lessing, 2015). The effects of this control are visible in voting patterns. Neighborhoods under militia domination show higher support for candidates affiliated with security forces (Hidalgo & Lessing, 2015), reinforcing the notion that these territories operate as controlled electoral districts. Weak law enforcement and clientelist urban networks, often supported by political allies, sustain these arrangements and strengthen paramilitary power (Arias & Barnes, 2016). Rio de Janeiro reflects a broader pattern of state capture by organized crime. Criminal groups convert territories into coercive electoral bases, secure political representation, and protect their interests through sustained influence over formal institutions. Map 1 - Territorial control and electoral participation Source: author's elaboration based on data from TSE, TRE and Fogo Cruzado. The map shows electoral participation rates overlaid on areas under criminal control in Rio de Janeiro. The West Zone (shown on the western, left-hand portion of the map), which concentrates the main militias, especially in Campo Grande and Santa Cruz, shows a concentration of blue circles indicating high participation. In the North Zone, participation is mixed, as is the presence of competing factions. Empirical strategy The explanatory variable is exposure to criminal governance. The main measure is a binary variable indicating whether a polling station falls within 500 meters of a documented territorial control polygon. As a dynamic complementary measure, I use georeferenced Fogo Cruzado events classified by criminal group in the same radius, capturing temporal variation that static polygons miss. The two types of measures enter the analyses separately. Polygons feed the base specification for mobilization and registration. Fogo Cruzado events feed the robustness tests and the candidate panel. The base specification regresses each station's electoral outcome on group exposure, with standardized demographic controls (proportion of elderly voters, low-education voters, women, and biometrically registered voters), neighborhood and year fixed effects, and standard errors clustered by electoral zone. The sample includes approximately 7,033 polling-station observations from the 2008 to 2024 municipal elections. Beyond the base model, I use four complementary identification strategies. I vary the exposure radius and use continuous log distance to test whether results are sensitive to the window. I estimate interaction models to test whether the militia effect is conditional on lethal violence and on Pacifying Police Unit presence. I use a stacked differences-in-differences design comparing stations moving from low to high militia exposure between consecutive elections with stations that did not change. Finally, in the electoral organization analysis, I shift the unit to examine whether candidates with a higher proportion of votes in militia areas develop a distinct pattern of territorial dependence. Data The analysis combines official electoral records from the Superior Electoral Court (TSE) with georeferenced data from territorial control polygons in Rio de Janeiro. Electoral data cover the municipal elections of 2008, 2012, 2016, 2020, and 2024. For each year, I use voter profiles and ballot-box vote counts aggregated to the polling-station level, yielding the number of registered voters and the turnout rate. The panel includes approximately 7,033 polling station observations per year, after harmonizing identifiers across elections and keeping only stations with complete information. I operationalize exposure to criminal governance using georeferenced territorial-control polygons compiled from the Instituto de Segurança Pública (ISP) and Pista News, complemented by armed-incident records from Fogo Cruzado. These sources allow me to classify each polling-station catchment as either under the control or under the influence of one of the criminal organizations under study. For the candidate-level analysis, I build a candidate × year panel restricted to candidates with at least two contests. Territorial dependence is the proportion of votes obtained at stations under militia control. Spatial concentration is the HHI index over the distribution of votes across stations. UPP deployment provides the exogenous variation for examining how territorial control shocks affect candidates' electoral bases. Results Criminal groups differ systematically in their modes of governance and therefore in their electoral effects. I expect militias, as politically integrated actors, to be associated with higher participation and registration, conditional on territorial stability and attenuated where lethal violence is more intense. For CV and ADA, as peripheral actors, I expect negative or null effects. Given the TCP's ambiguous network character, I do not formulate a directional expectation. These contrasts inform the interpretation of the models below. I present the results in three blocks. The first documents the effects on mobilization (participation and registration) in the main panel model. The second examines how lethal violence and UPP presence moderate militia mobilization capacity. The third addresses electoral organization, examining whether and when militia presence affects vote concentration and candidates' territorial bases. Evidence is more limited at the aggregate level but more consistent when focused on intra-candidate dynamics. Findings The main panel model shows that proximity to a militia control polygon is associated with a significant increase in participation (Table 1 ). The result is robust to different exposure radii. The coefficient grows monotonically as the window expands, suggesting that the mobilizing effect extends beyond the immediate vicinity of the station. When all groups enter simultaneously (Table 2 ), the result holds. CV and ADA are associated with declines in participation. TCP shows effects close to zero and unstable across specifications, consistent with the prediction for the hybrid case. Results remain stable when territories with pre-established militia control are excluded (β = 1.42 vs 1.45 in the full model). Estimators robust to cohort heterogeneity confirm the direction and magnitude of the effects. Patterns by alternative exposure radius are detailed in the Appendix. Table 1 Militia and electoral mobilization Variable (1) Turnout (2) Log Electorate Militia (500 m) 1.4546** 0.0354*** (0.4297) (0.0055) N 7,033 7,033 Dem. controls Yes Yes FE neighborhood + year neighborhood + year Cluster electoral zone electoral zone Notes: Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60–69), low education, women, and biometric registration. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Results hold with Conley standard errors at radii of 1 to 10 km (Appendix A.2) and with spatial SAR and SEM models (Appendix A.4). Table 2 Comparison across Criminal Groups (Joint Model) Criminal group (1) Turnout (2) Log Electorate Militia 1.3117** 0.0329*** (0.4155) (0.0055) CV −0.8132* −0.0144** (0.3351) (0.0047) TCP −0.0421 −0.0024 (0.3510) (0.0069) ADA −0.8645* −0.0019 (0.3792) (0.0056) N 7,033 7,033 Dem. controls Yes Yes FE neighborhood + year neighborhood + year Cluster electoral zone electoral zone Notes: Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. All criminal groups enter simultaneously in the same model. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60–69), low education, women, and biometric registration. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Results hold with Conley standard errors at radii of 1 to 10 km (Appendix A.2) and with spatial SAR and SEM models (Appendix A.4). Table 3 Robustness: Alternative Measures of Militia Proximity Specification (1) Turnout (2) Log Electorate 250 m 0.6198 0.0240*** (0.3969) (0.0056) 500 m 1.3117** 0.0329*** (0.4155) (0.0055) 1,000 m 1.9827*** 0.0450*** (0.5542) (0.0071) Dist. (log) −0.2879*** −0.0065*** (0.0742) (0.0010) N 7,033 7,033 Dem. controls Yes Yes FE neighborhood + year neighborhood + year Cluster electoral zone electoral zone Notes: Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Each row is an alternative specification of the militia treatment variable. Criminal controls for other groups included in all specifications. Voter registration follows the same pattern. Militia areas have more registered voters. Unlike the turnout result, this effect is already significant at the smallest exposure radius. CV shows a negative and significant effect on registration, reinforcing the interpretation that these territories have smaller electorates. TCP and ADA show no effect distinguishable from zero. Controls follow the expected pattern. Biometric registration is associated with a smaller electorate due to the institutional cost of re-registration. The Appendix details robustness by radius. The pattern is monotonic, with the coefficient growing as the exposure window expands. The effect radiates beyond the immediate vicinity of the polling station. Using log distance as a continuous variable, the sign reverses as expected, reinforcing the spatial interpretation. To check whether the effect depends on comparing consolidated militia territories with uncontrolled ones, I re-estimated the main model excluding stations already militia-controlled in the first panel period. The coefficient is stable (β = 1.42 vs 1.45 in the full model). Identification does not rest on structural differences between old and new militia territories. Effect conditions The mobilizing effect of militias is not constant. I examine two conditions that may modify it. The first is the intensity of lethal violence around the polling station. The second is the presence of Pacifying Police Units. The first test rests on the idea that mobilization capacity depends on territorial control, which violence undermines. The second examines whether organized state presence disrupts or coexists with militia mobilization networks. The UPP result is counterintuitive and deserves particular attention. The argument behind the interaction with lethal violence is straightforward. If militia mobilization depends on stable territorial control, then areas of high violence, where control is contested, should exhibit a smaller effect. The interaction with UPPs is less obvious. The conventional expectation is that organized police presence would reduce militia mobilization capacity. But militia clientelist networks may survive and adapt to police presence, especially when UPP officers themselves have paramilitary ties. Table 4 Conditional Effect of Militia. Interaction with Lethal Violence Coefficient (1) Turnout (2) Log Electorate Militia 1.5186*** 0.0308*** (0.3389) (0.0053) Lethal violence (z) 1.8399*** 0.0343*** (0.2851) (0.0049) Militia × Violence −1.0801*** −0.0147** (0.3030) (0.0050) N 7,011 7,011 Dem. controls Yes Yes FE neighborhood + year neighborhood + year Cluster electoral zone electoral zone Notes: Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60–69), low education, women, and biometric registration. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Moderator (lethal violence = intentional homicides ISP/CISP) standardized to mean zero and standard deviation one within year. Other criminal groups (CV, TCP, ADA) included as controls. Table 4 a. Militia effect by level of lethal violence Lethal violence percentile (1) Turnout (2) Log Electorate p10 (low violence) 2.2798*** 0.0412*** (0.4827) (0.0080) p25 1.9932*** 0.0372*** (0.4217) (0.0069) p50 (median) 1.4126*** 0.0293*** (0.3250) (0.0051) p75 0.4963 0.0169*** (0.3156) (0.0045) p90 (high violence) −0.1820 0.0076 (0.4246) (0.0063) Notes: Marginal effect = β(militia) + β(militia × violence) × z_intentional_homicides, with standard errors by delta method. Moderator: intentional homicides (ISP/CISP), standardized within year. *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Table 4 a details the marginal effect of militia presence by level of lethal violence. In low-violence contexts, the positive effect on participation is substantial. As violence intensifies, the effect diminishes. At the 90th percentile of lethal violence, the effect loses significance and turns negative. Militia control ceases to operate as a mobilization structure where armed conflict is intense. This gradient shows that militia mobilization efficacy depends on territorial stability. Where other armed groups contest control, the incentives sustaining electoral brokerage weaken. The result on voter registration follows the same direction, though less pronounced. Table 5 Conditional Effect of Militia. Interaction with UPP Presence Coefficient (1) Turnout (2) Log Electorate Militia 1.1417*** 0.0293*** (0.3177) (0.0054) Active UPP −2.6416*** −0.0347*** (0.4821) (0.0058) Militia × UPP 1.9272** 0.0090 (0.7188) (0.0097) N 7,012 7,012 Dem. controls Yes Yes FE neighborhood + year neighborhood + year Cluster electoral zone electoral zone Notes: Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60–69), low education, women, and biometric registration. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Active UPP = dummy for presence of active Pacifying Police Unit in the electoral period. Other criminal groups (CV, TCP, ADA) included as controls. Table 5 a. Marginal Effect of Militia by UPP Presence Institutional context (1) Turnout (2) Log Electorate No UPP 1.1417*** 0.0293*** (0.3177) (0.0054) With Active UPP 3.0688*** 0.0382*** (0.7345) (0.0102) Marginal effect calculated at upp_active = 0 and upp_active = 1. Standard errors by delta method. *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Notes: Each interaction column corresponds to a separate model including the other criminal exposures, controls and fixed effects. The interaction with UPPs reveals a counterintuitive result. As Table 5 shows, stations near militia polygons with an active UPP exhibit even higher participation than those without. On voter registration, the interaction is not significant. This suggests that militia mobilization networks are stable enough to persist, and possibly intensify, when coexisting with formal police presence. This is unsurprising because, like UPP officers, militiamen are themselves drawn from police forces. Electoral organization and base territorialization The evidence on electoral organization, vote concentration and candidate territorialization, is more heterogeneous. In the aggregate model with neighborhood and year fixed effects, militia presence has no significant effect on the most common concentration indicators. The Herfindahl-Hirschman Index and top candidate's vote share are close to zero and insignificant. The most consistent result is for the effective number of parties. Stations near militias have a slightly more restricted competitive field, though the effect is modest. Results appear in Table 6 in the Appendix and serve as complementary evidence. The static specification with 2019 polygons limits what can be inferred about dynamics. I cannot separate militia presence effects from pre-existing neighborhood characteristics that the polygons map. Table 6 Electoral Organization. Criminal Groups (Joint Model) Criminal group (1) HHI (2) ENP (3) share_top1 Militia 0.0002 −1.6576 0.0062 (0.0017) (1.1212) (0.0038) CV −0.0031* 0.6039 −0.0068* (0.0013) (0.9264) (0.0034) TCP 0.0058† −3.3851* 0.0149* (0.0033) (1.3150) (0.0068) ADA 0.0024 −2.5110† 0.0059 (0.0029) (1.4303) (0.0069) N 5,660 5,660 5,660 Dem. controls Yes Yes Yes FE neighborhood + year neighborhood + year neighborhood + year Cluster electoral zone electoral zone electoral zone Notes: All criminal groups enter simultaneously in the same model. Treatment variable: territorial dominance (polygon Pista News / ISP, ~ 2019). Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. Demographic controls: proportion of elderly (60–69), low education, women, and biometric registration. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Table 7 Electoral Organization. Robustness by Specification Specification (1) HHI (2) ENP (3) share_top1 Militia dom. (polygon) 0.0001 −1.6398 0.0062† (0.0017) (1.1073) (0.0037) FC militia (dummy, 500 m) 0.0007 −1.9921* 0.0059 (0.0016) (0.8146) (0.0036) FC militia (intensity, 500 m) 0.0001 −1.8179 0.0010 (0.0047) (1.9608) (0.0094) N 5,660 5,660 5,660 Dem. controls Yes Yes Yes FE neighborhood + year neighborhood + year neighborhood + year Cluster electoral zone electoral zone electoral zone Notes: Each row is an alternative specification of the militia treatment variable. Unit: polling station × year. FE: neighborhood + year. Standard errors clustered by electoral zone (G = 97), in parentheses. Demographic controls: proportion of elderly (60–69), low education, women, and biometric registration. Significance: *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. The design that captures temporal variation produces more consistent evidence of competitive reorganization. The stacked differences-in-differences compares stations whose militia exposure radius changes between consecutive elections to stable stations in the same period. Table 8 shows that when a station moves from low to high militia exposure, the effective number of parties falls significantly. The HHI and the top candidate's share point in the expected direction. Together, these results suggest that expanding militia presence narrows the competitive field, even if the effect does not appear uniformly across the three measures. Table 8 Electoral Organization. Stacked DiD (FC Exposure Switchers) Coefficient (1) HHI (2) ENP (3) share_top1 Militia × Post (DiD) 0.0045† −5.7631** 0.0107† (0.0026) (1.8054) (0.0064) N 4,122 4,122 4,122 Dem. controls Yes Yes Yes FE station + year station + year station + year Cluster electoral zone electoral zone electoral zone Stacked Differences-in-Differences (DiD) design. Treatment = station moving from low to high militia exposure (Fogo Cruzado, 500 m) between two consecutive elections. Control = stations without exposure change in the same period. Each electoral pair constitutes an independent stack; FE station + year. HHI = electoral Herfindahl-Hirschman index; ENP = effective number of parties; share_top1 = top candidate's vote share. *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. The strongest evidence of territorialization appears at the candidate level. Table 9 shows that candidates closer to militia polygons get a larger share of their votes from those territories, regardless of total vote count. A UPP entry shock significantly reduces this territorial dependence, suggesting militias structure specific candidates' electoral bases and police presence disrupts this tie. Table 11 completes the picture. Candidates with more spatially concentrated bases depend more on militia territories, reinforcing the link between electoral territorialization and militia control. This pattern is hard to explain by demographic characteristics alone. It points to a structural link between vote organization and territorial control. Table 9 Candidates' territorial base Specification (1) Militia territorial dep. Militia intensity (FC) 2.7221*** (0.1676) % militia stations 1.2284*** (0.0707) UPP shock (% stations) −0.9764*** (0.1474) N 1,945 Dem. controls Yes FE candidate + year Cluster electoral zone Unit: candidate × year (candidates with ≥ 2 elections, N = 827). Territorial dependence = proportion of votes obtained at stations under militia control. FE: candidate + year. Control: log(total votes). Militia intensity (FC) = mean intensity of militia Fogo Cruzado events at candidate's polling stations. *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Table 11 Territorialization of the electoral base Variable (1) Militia dep. (2) Spatial HHI (3) Militia dep. Log(votes) 0.0032 −0.0066*** 0.0077 (0.0030) (0.0005) (0.0033) Spatial HHI 0.6798*** — — (0.1077) N 6,203 6,203 6,203 FE year year year Cluster electoral zone electoral zone electoral zone Unit: candidate × year (candidates with ≥ 10 votes, N = 6,203). Year FE. Territorial dependence = proportion of votes obtained at stations ≤ 500 m from militia polygon. Spatial HHI = Herfindahl-Hirschman index calculated over the distribution of candidate votes across stations. *** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10. Qualitative evidence To describe the mechanism tested by the statistical models, I added a qualitative component explaining the political processes underlying the estimated effects. The qualitative corpus consists of triangulated secondary sources. I consulted official documents produced by the Parliamentary Commission of Inquiry on Militias in the Rio de Janeiro State Legislative Assembly. Through a survey of 105 news articles in four local newspapers (Jornal do Brasil, O Dia, O Globo, Extra) covering 2006 to 2020, I built a qualitative database. It documents voter co-optation and electoral capture by organized crime. I selected articles using keywords related to militias, elections, voter registration, and residents' associations. I included only concrete records, excluding non-investigative articles. The analytical unit is the event. Reports function as documentary sources. Mechanisms In Rio de Janeiro's West Zone, and often in the North Zone, criminals go beyond direct coercion in their mobilization strategies on and before election day. Criminal organizations govern the electoral process. Residents of Campo Grande, Santa Cruz, and Jacarepaguá, to name only a few neighborhoods, report that militiamen map individual families through residents' associations. These organizations conduct community population censuses and monitor local political behavior. First, militiamen summon residents to meetings in social centers they control, usually the association headquarters. There, criminals demand that residents bring their identity documents and voter registration cards for entry into the militia's database, which identifies each voter's polling station. In Santa Cruz, for example, militiamen from the Babu clan collected voter card numbers under the promise of improvements. They conditioned basic urban services on proof of electoral regularity and support for the 'house' candidate, former state deputy Jorge Babu. Voters know that non-compliance will be discovered and punished, and they believe it. In more extreme cases, as reported in Muzema, the militia physically collected residents' voter cards weeks before the election. The documents were returned on the eve of voting, accompanied by the local candidate's flyer and instructions for attendance on election day. When a voter is not registered—common given the migrant population—militiamen take them to electoral registry offices to regularize their status. Militias also register voters from other cities or neighborhoods through targeted strategies. Rio's electoral court itself reported cases of anomalous voter registration and the use of public vehicles, such as kombis, to transport them to registry offices. Militiamen also handle the bureaucratic side, sometimes providing false proof of residency for submission to the local electoral authority. On election day, the militia sets up real-time mobilization logistics. Alternative transportation, the militiamen's initial income source since the late 1990s, is mobilized in a kind of 'vote caravan.' The clandestine company 'Viação Coringa,' run by militiamen in Santa Cruz and Paciência, displayed a clown smile as a group identifier. It transported residents to their polling stations free of charge. Militiamen position themselves at the doors of schools and polling stations as parallel election monitors. According to residents' accounts, police officers assigned to election security are seen conversing with militiamen or deliberately ignoring their activity. This symbiosis discourages resistance and undermines ballot secrecy, even though the vote inside the booth cannot be identified. To validate who voted, militiamen demand the delivery of the voting receipt issued by the electoral justice. In other cases, there is even photo confirmation of votes sent from the ballot machines. After the election, failure brings consequences. Reports range from suspended access to public goods like internet service to eviction of residents who refuse to cooperate. Victims report late-night approaches with 24 to 48-hour deadlines to leave under threat of death for all family members. Outcomes can also be positive for the community when militia brokerage succeeds. Militiamen offer, in exchange for votes, the suspension of extortionate security fees and internet access for the months following the election, November and December. Militiamen create an atmosphere of pressure so that no one stays home. Concluding remarks I show that the political effect of territorial control on elections varies with the degree of political integration of criminal organizations, measured by the depth of ties to electoral elites, the stability of candidate alliances, and systematic mobilization capacity. Paramilitary groups, under politically integrated governance, mobilize voters and produce higher participation through coercion and clientelist incentives. Drug-trafficking factions, under politically peripheral governance, intervene more selectively, without systematic mobilization. Their influence on participation varies with context, ranging from suppression to no distinguishable effect. The TCP, whose territorial networks partly overlap with militias in the West Zone, shows inconsistent effects across specifications, with low magnitude and unstable direction. This pattern is consistent with the prediction for the hybrid case. Across multiple identification strategies, politically integrated groups exercise structured political control, act as intermediaries between voters and candidates, and produce higher participation in their areas of influence. This mobilization works through direct coercion, suppression of competition, and economic benefits, consolidating their role as local political actors. The militia coefficient is stable when territories with pre-established control are excluded and persists under estimators robust to cohort heterogeneity. Politically peripheral actors do not systematically mobilize voters. Their actions depend on territorial stability and the political interests at stake. The central contribution is the distinction between forms of political engagement by criminal organizations. Electoral behavior is predicted both by armed presence and by how crime articulates with institutional politics. This approach broadens analytical perspectives on Brazilian politics and enables comparisons with other contexts of territorial capture. My results challenge the assumption that higher electoral participation corresponds to greater democratic quality. Criminal groups reorganize voting incentives, distort competition, and depoliticize electoral choice. The inference on coerced voting rests on descriptive evidence such as individual family mapping, physical collection of voter registration cards, organized transportation on election day, attendance monitoring, and post-electoral sanctions documented in the qualitative corpus. Under these conditions, voting ceases to express the voter's will. It becomes a mechanism that legitimizes domination through power structures that permeate the state. Understanding voting behavior under criminal governance requires abandoning linear conceptions of democratic participation. In many territories, voting is not a free choice but an imposition. Politics becomes an instrument of domination managed by a hybrid state where formal and informal authorities intertwine to shape democratic conditions. Measuring electoral participation without accounting for this context distorts the true state of democracy. I contribute to the comparative politics of criminal governance, drawing on the case of Rio de Janeiro. Militias, composed largely of active and former agents of repressive forces, and drug-trafficking factions coexist in the same political system with opposite ties to the state and electoral machines. The distinction between politically integrated and politically peripheral governance, combined with the nature of state relations, helps explain the varied effects of criminal governance on electoral incorporation. The framework applies to contexts like Colombia, Mexico, or Italy. This lens applies to other Latin American countries where armed groups interact with public institutions and shape electoral processes. Electoral behavior under criminal governance emerges from structured relations between the state, criminal actors, and voters. Higher participation in territories dominated by politically integrated actors reflects durable arrangements of coercion, intermediation, and institutional ties rooted in historical struggles over territory and authority. Declarations Funding This research was supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, Brazil) through a doctoral sandwich fellowship (Programa de Doutorado Sanduíche no Exterior, PDSE, 2025–2026) at the University of Illinois Chicago, awarded to Igor Novaes Lins. Declaration of Competing Interests The author declares no competing financial or personal interests that could have influenced the work reported in this article. References Albarracín J, Karolczak M, Rodrigo, Wolff J (2025) Violence against civil society actors in democracies: Territorialization of criminal economies and the assassination of social activists in Brazil. J Peace Res 62(5):1411–1427. https://doi.org/10.1177/00223433251347784 Albarracín J (2018) Criminalized Electoral Politics in Brazilian Urban Peripheries. Crime Law Social Change 69(4):553–575 Albarracín J Forthcoming. Undermining Democracy from the Peripheries: Criminalized Electoral Politics in Brazil. Book Manuscript Alesina A, Piccolo S, Pinotti P (2019) Organized crime, violence, and politics. Rev Econ Stud 86(2):457–499 Anduiza E, and Guillem Rico (2022) Sexism and the Far-Right Vote: The Individual Dynamics of Gender Backlash. American Journal of Political Science Arias ED (2006) The Dynamics of Criminal Governance: Networks and Social Order in Rio de Janeiro. J Latin Am Stud 38(2):293–325 Arias ED (2013) The Impacts of Differential Armed Dominance of Politics in Rio de Janeiro, Brazil. Stud Comp Int Dev 48:263–284 Baraldi A, Laura E, Papagni, and Marco Stimolo (2024) Neutralizing the Tentacles of Organized Crime. Assessment of the Impact of an Anti-Crime Measure on Barnes N (2017) Criminal Politics: An Integrated Approach to the Study of Organized Crime, Politics, and Violence. Perspect Politics 15(4):967–987 Blattman C et al (2024) Gang Rule: Understanding and Countering Criminal Governance. Rev Econ Stud : rdae079 Boas T, Smith AE (2015) Religion and the Latin American Voter. In: Ryan E, Carlin MM, Singer, Zechmeister EJ (eds) The Latin American Voter: Pursuing Representation and Accountability in Challenging Contexts. University of Michigan Press, Ann Arbor, pp 112–134 Bohn SR (2004) Evangélicos no Brasil: Perfil Socioeconômico, Afinidades Ideológicas e Determinantes do Comportamento Eleitoral. Opinião Pública 10:288–338 Bullock J (2019) Criminal Dominance and Campaign Concentration Cano I, Duarte (2012) Thais. No sapatinho: a evolução das milícias no Rio de Janeiro (2008–2011). LAV, Laboratório de Análise da Violência. LAV-UERJ) Carreirão Sérgio et al (2022) Polarização, Fragmentação e Participação Eleitoral no Brasil. Revista Brasileira de Ciência Política 29(1):55–80 Carreras M, Néstor Castañeda-Angarita (2014) Who Votes in Latin America? A Test of Three Theoretical Perspectives. Comp Polit Stud 47(8):1079–1104 Cepaluni G, Daniel Hidalgo F (2016) Compulsory Voting Can Increase Political Inequality: Evidence from Brazil. Political Anal 24(2):273–280 Cigtar V (2018) Party System Fragmentation and Electoral Volatility in Eastern Europe. East Eur Politics Soc 32(3):621–648 Córdova A (2019) Living in Gang-Controlled Neighborhoods: Impacts on Electoral and Nonelectoral Participation in El Salvador. Latin Am Res Rev 54(1):201–221 Corrales J (2020) The Expansion of LGBT Rights in Latin America and the Backlash. In The Oxford Handbook of Global LGBT and Sexual Diversity Politics, 185–200 CPI das Milícias (2008) Relatório Final da Comissão Parlamentar de Inquérito Destinada a Investigar a Ação de Milícias no Âmbito do Estado do Rio de Janeiro. Assembleia Legislativa do Estado do Rio de Janeiro (Alerj), Rio de Janeiro Dahl RA (1998) On Democracy. Yale University Press, New Haven Daly SZ (2022) How do violent politicians govern? The case of paramilitary-tied mayors in Colombia. Br J Polit Sci 52(4):1852–1875 Daniele G (2019) Strike One to Educate One Hundred: Organized Crime, Political Selection and Politicians’ Ability. J Econ Behav Organ 159:650–662 Dettrey BJ, Leslie A, Schwindt-Bayer (2009) Voter Turnout in Presidential Democracies. Comp Polit Stud 42(10):1317–1338 Feldmann AE, Juan PL (2022) Gobernanza Criminal y la Crisis de los Estados Latinoamericanos Contemporáneos. Ann Rev Sociol 48(1):S–1 Gallego A (2015) Unequal Political Participation Worldwide. Cambridge University Press, Cambridge Geys B (2006) Explaining Voter Turnout: A Review of Aggregate-Level Research. Electoral Stud 25(4):637–663 Grupo de Estudos dos Novos Ilegalismos (GENI) and Fogo Cruzado (2024) Grande Rio sob Disputa: Mapeamento dos Confrontos por Territórios. Niterói: GENI/UFF and Fogo Cruzado-RJ. Available at: https://geni.uff.br/2024/06/05/grande-rio-sob-disputa-mapeamento-dos-confrontos-por-territorios/ . Accessed June 14, 2025 Healy A, and Neil Malhotra (2013) Retrospective Voting Reconsidered Annual Rev Political Sci 16:285–306 Lessing B (2021) Conceptualizing Criminal Governance. Perspect Politics 19(3):854–873 Ley S (2018) To vote or not to vote: how criminal violence shapes electoral participation. J Conflict Resolut 62(9):1963–1990 Lijphart A (1997) Unequal Participation: Democracy’s Unresolved Dilemma. Am Polit Sci Rev 91:1–14 Lins IN, Machado CAM (2023) O Crime É Político: Elementos Teóricos para uma Análise Neoinstitucionalista das Milícias no Rio de Janeiro. Revista Brasileira de Ciência Política. 42:e271780 Lins I, Novaes, and Carlos Machado (2024) A Geografia do Voto das Milícias na Cidade do Rio de Janeiro. Revista de Ciência Política, Teoria & Pesquisa, p e024008 Lins IN (2023) Da Baixada à Zona Sul: Caminhos da Violência Política de Raça no Rio de Janeiro. Revista Brasileira de Segurança Pública 17(1):188–207 Madrid RaúlL (2008) The Rise of Ethnopopulism in Latin America. World Polit 60(3):475–508 Mafia Violence in Italy J Econ Behav Organ 223: 57–85 Mainwaring S et al (2015) The Left and the Mobilization of Class Voting in Latin America. In The Latin American Voter: Pursuing Representation and Accountability in Challenging Contexts, eds. Ryan E. Carlin, Matthew M. Singer, and Elizabeth J. Zechmeister. Ann Arbor: University of Michigan Press, 99–105 Manso BP (2020) A República das Milícias: Dos Esquadrões da Morte à Era Bolsonaro. Todavia, São Paulo Misse M (2011) Crime organizado e crime comum no Rio de Janeiro: diferenças e afinidades. Revista de sociologia e política 19:13–25 Misse M (2007) Mercados Ilegais, Redes de Proteção e Organização Local do Crime no Rio de Janeiro. Estudos Avançados 21(61):139–157 Norris P (2011) Democratic Deficit: Critical Citizens Revisited. Cambridge University Press, Cambridge Pessoa P (2023) Political Competition When Gangs Rule. Effects of Removing Armed Groups’ Territorial Control in Brazil Polícia F (2024) Relatório Final do Inquérito Policial 2023.0059871-SR/PF/RJ – Inq n.º 4954/DF. Superintendência Regional no Estado do Rio de Janeiro, Rio de Janeiro Ribeiro E, Aparecido J, Borba, Rafael da Silva (2015) Comparecimento Eleitoral na América Latina: Uma Análise Multinível Comparada. Revista de Sociologia e Política 23(54):91–108 Romero B, More COPS, Higher, Turnout? (2025) APSA Preprints. doi: 10.33774/apsa-2024-bwwpd-v2 Rozema R (2007) Paramilitares y Violencia Urbana en Medellín. Colombia Foro Int : 535–550 Schumpeter JA (1942) Capitalism, Socialism and Democracy. Harper & Brothers, New York Snyder R, Angélica D-M (2009) Does Illegality Breed Violence? Drug Trafficking and State-Sponsored Protection Rackets. Crime Law Social Change 52:253–273 Trejo (2020) Guillermo, and Sandra Ley. Votes, drugs, and violence: The political logic of criminal wars in Mexico. Cambridge University Press Trelles A (2012) Bullets and votes: Violence and electoral participation in Mexico. J Politics Latin Am 4(2):89–123 Trudeau J (2022) How Criminal Governance Undermines Elections. Politics and International Relations Valdini ME, Lewis-Beck MS (2018) Economic Voting in Latin America: Rules and Responsibility. Am J Polit Sci 62(2):410–423 Zaluar A, Isabel Siqueira Conceição (2007) Favelas sob o Controle das Milícias no Rio de Janeiro. São Paulo em Perspectiva 21(2):89–101 Map 1 Map 1 is available in the Supplementary Files section. Additional Declarations The authors declare no competing interests. Supplementary Files Map1.png Map 1 - Territorial control and electoral participation Source: author's elaboration based on data from TSE, TRE and Fogo Cruzado. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9534248","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":629860132,"identity":"09a91c20-f9b5-403d-8e3c-7120515ee764","order_by":0,"name":"Igor Novaes Lins","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA90lEQVRIiWNgGAWjYBACNgSTueEAA4MNkMHYeIBILYwgLWkwBlGAsQFIHAYz8Wrh41987HNBxWF5fv6DjYcL/py3W9t+GGhLjU00TodJPEuePePMYcOZMxIbDs9su5287UwiUMuxtNwGnFrOGDPztqUxbrjB2HCYt+F2stkBoBYgm6AW+/3nDzYc5vlzLtns/EMCWvh7QFpsEjcwAB3Gw3bAzuwGQVvYkpl5ztgkzwCqPMzblpxgdgNoSwIev8j3Hz7MzFMhYdsPZHzm+WNnb3Y+/eGDDzU2OLUwSCSg8hPBKhMw1CEB/gOofHt8ikfBKBgFo2BkAgCO6WOxMptrnQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-0510-8355","institution":"Universidade de Brasília","correspondingAuthor":true,"prefix":"","firstName":"Igor","middleName":"Novaes","lastName":"Lins","suffix":""}],"badges":[],"createdAt":"2026-04-26 19:51:50","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9534248/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9534248/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107992468,"identity":"144d9bb1-ad15-4d3c-bec7-47326d614703","added_by":"auto","created_at":"2026-04-28 10:25:48","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":135459,"visible":true,"origin":"","legend":"\u003cp\u003eMechanism: how criminal governance shapes electoral participation\u003c/p\u003e\n\u003cp\u003eSource: author's elaboration.\u003c/p\u003e","description":"","filename":"fig1mechanismen2.png","url":"https://assets-eu.researchsquare.com/files/rs-9534248/v1/5f48176c522d4eb01b30c871.png"},{"id":107992467,"identity":"4200646c-0a9f-411e-be31-4e86f7f328e0","added_by":"auto","created_at":"2026-04-28 10:25:47","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":78124,"visible":true,"origin":"","legend":"\u003cp\u003eEngagement of peripheral governance under integrated brokerage\u003c/p\u003e\n\u003cp\u003eSource: author's elaboration.\u003c/p\u003e","description":"","filename":"fig2brokerageen1.png","url":"https://assets-eu.researchsquare.com/files/rs-9534248/v1/f9523db3bdb9a67f533ffe3d.png"},{"id":107992617,"identity":"a2124e7f-4ff8-40fe-90b5-97b64ca1d154","added_by":"auto","created_at":"2026-04-28 10:26:11","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":742346,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9534248/v1/57ef2b2c-b2eb-499c-8620-30fdde11f924.pdf"},{"id":107992591,"identity":"3bb19d33-0fa8-439e-a254-1a005a7d446b","added_by":"auto","created_at":"2026-04-28 10:26:04","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":361308,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMap 1 - Territorial control and electoral participation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSource: author's elaboration based on data from TSE, TRE and Fogo Cruzado.\u003c/p\u003e","description":"","filename":"Map1.png","url":"https://assets-eu.researchsquare.com/files/rs-9534248/v1/ed8087eaa32d8915722c294d.png"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eVoting under criminal governance: election mobilization by criminal organizations\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCriminal organizations govern elections. Voters stay away from the polls out of fear (Trelles \u0026amp; Carreras, 2012), responding to criminal activity beyond direct victimization (Ley, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), or facing political violence (Van Baalen, 2024; Lins, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This governance goes beyond demobilization. Criminals mediate candidate access to voter groups in urban peripheries (Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Trudeau, 2025), condition the preferences of voters living under armed control (Pessoa, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Hidalgo \u0026amp; Lessing, 2015), restrict political campaigns (Arias, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Bullock, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and systematically structure electoral dynamics (C\u0026oacute;rdova, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Accardo; De Feo; De Luca, 2023).\u003c/p\u003e \u003cp\u003eI ask why groups with similar territorial control produce different political effects on elections. These effects depend not only on armed control, but on the political governance criminal organizations accumulate over electoral processes over time.\u003c/p\u003e \u003cp\u003eRio de Janeiro offers a singular analytical opportunity. Paramilitaries rooted in police forces and drug-trafficking factions with distinct organizational trajectories coexist within the same municipal electoral system. This allows comparison across modes of criminal governance without the institutional confounders of cross-national designs, isolating political integration as the explanatory dimension.\u003c/p\u003e \u003cp\u003eIn the city, politicians, police officers, and local administrators maintain longstanding alliances with criminal organizations. These ties regulate who can campaign and channel votes through selective coercion and protection. Electoral strongholds in paramilitary-controlled neighborhoods such as Gard\u0026ecirc;nia Azul and Rio das Pedras show how political families tied to criminals expand their vote share where armed groups govern (Pantale\u0026atilde;o \u0026amp; Montini, 2025). They signal voter access and the capacity to mobilize through coercion or clientelist relations (Trudeau, 2025).\u003c/p\u003e \u003cp\u003eThis pattern extends beyond episodic scandals of violence. Across Latin American cities, armed actors and political elites form durable agreements that intertwine illicit markets, local bureaucracies, and electoral strategies. These configurations allow criminal groups to influence who participates, who runs, and who wins. Democratic representation comes to reflect the degree of political governance exercised by criminal organizations. This produces systematic advantages for embedded actors (Pessoa, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) and persistent barriers for those who oppose them (Albarrac\u0026iacute;n, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Albarrac\u0026iacute;n, Karolczak, \u0026amp; Wolff, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). After elections, these arrangements shape the distribution of public offices and guide policymaking (Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Pessoa, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eI argue that criminal organizations vary systematically in their degree of political involvement and in the mechanisms of electoral intervention. Some criminal organizations maintain durable ties with political elites, mediate access to voters, and mobilize them. They participate directly in campaigns by promoting or imposing a candidate. Other groups operate with more limited political structures, imposing selective restrictions, blocking rivals, and intervening only when threatened.\u003c/p\u003e \u003cp\u003eFrom this variation, I propose a typology with three levels of political governance. Politically integrated organizations participate systematically in electoral processes. Politically peripheral organizations interfere episodically and reactively. Hybrid configurations identify contingent opportunities for electoral action according to context.\u003c/p\u003e \u003cp\u003eI show that political integration, defined by durable ties with electoral elites, stable candidate alliances, and accumulated mobilization capacity, predicts the political effect of territorial control. Where governance is politically integrated, electoral competition is compressed and candidates outside criminal networks face access barriers that convert territorial domination into political exclusion.\u003c/p\u003e \u003cp\u003eWith approximately 7,000 polling station observations across five municipal elections, the empirical strategy combines regression with two-way fixed effects, interaction models, and a stacked differences-in-differences design. Militia presence is associated with higher participation and voter registration. This effect attenuates as lethal violence increases but is amplified in areas with active Pacifying Police Units (UPPs). Comando Vermelho and Amigos dos Amigos are associated with declines in participation. Terceiro Comando Puro shows unstable effects near zero, consistent with the prediction for the hybrid case. Militia exposure predicts territorial dependence of candidates' electoral bases, and UPP entry shocks disrupt these ties. Qualitative evidence shows how militias organize the vote logistically, from directed registration to election-day transportation.\u003c/p\u003e\n\u003ch3\u003eCriminal governance and elections\u003c/h3\u003e\n\u003cp\u003eCriminal governance shapes elections through territorial control, voter mobilization, and restricted neighborhood access. In Rio de Janeiro, these practices let criminal groups intervene politically without formal party ties, expanding their independence and influence. Criminal actors interact directly with voters, often outside institutional channels (Trudeau, 2025). Their capacity to impose rules and fill governance gaps reflects a broader pattern in which selective state absence, institutional weakness, or collusion allows illicit groups to exercise political power (Lessing, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Recent research shows how these groups regulate community behavior (Feltran, 2019), restrict elected authorities (C\u0026oacute;rdova, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and influence police operations (Misse, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Their interactions with formal institutions alter democratic governance (Blattman et al., 2021; Melnikov et al., 2022; Manso, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis governance arises through networks connecting criminals, community leaders, elected politicians, and law enforcement agents. These networks protect traffickers and paramilitaries from repression and integrate favelas and urban peripheries into the local political system (Arias, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). These groups coerce or buy votes and supply services typically associated with the state or the private sector, such as cooking gas and public transportation. According to Lessing (2020), despite continuous repression, these groups seek recognition. They achieve it through practical authority, acquired by regulating violence and social life (Lins \u0026amp; Machado, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Lessing \u0026amp; Willis, 2019). From this interaction comes 'violent democracy' (Arias, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2006\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Arias describes systems in which armed criminal actors routinely intervene in elections, negotiate directly with public authorities, and integrate into local political structures.\u003c/p\u003e \u003cp\u003eCriminal groups differ from other political mediators in their capacity to coerce voters beyond the limits of partisan institutions. They reach areas other intermediaries cannot, especially zones affected by violence. In these contexts, they press for more votes without facing significant institutional costs (Albarrac\u0026iacute;n, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This influence stems from their role in criminal governance. Leaders build social capital through kinship ties (Misse, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) and create monitoring networks based on extortion and surveillance (Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eElectoral behavior does not result only from voters' responses to violence or the presence of criminal groups. The active intervention of these groups as vote brokers also conditions it. Criminal governance goes beyond occupying a territory. It requires authority, the regulation of political life, and the delivery of services, including political ones, for residents under its influence (Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Lins \u0026amp; Machado, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Manso, 2022).\u003c/p\u003e \u003cp\u003eCriminal organizations use politics strategically. Understanding how illicit activities overlap with formal structures, and how organized crime sustains itself through political contracts, is essential (Pantale\u0026atilde;o \u0026amp; Montini, 2025). Politics becomes a commodity in the illicit economy of armed groups (Lins \u0026amp; Machado, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zaluar \u0026amp; Concei\u0026ccedil;\u0026atilde;o, 2007; Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). They become efficient political intermediaries because governing marginalized urban territories develops these capacities (Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Pessoa, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBeyond violence, research shows that criminal groups mobilize voters by buying votes or persuading them that certain candidates will help the community while others may cause harm (Manso, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Because these groups meet residents' basic needs, their promises and threats carry credibility. Once established, the alliance between criminal actors and politicians spreads rapidly across the territories under their control (Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003ePoliticians pursue office through criminally supported electoral structures while criminal actors protect or expand illicit profits, producing mutually beneficial relations (Albarrac\u0026iacute;n, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). In some cases, the lines between crime and politics blur into a stable arrangement that does not depend on electoral cycles alone, though elections remain a central asset.\u003c/p\u003e \u003cp\u003eCriminal groups manage voter access and mobilize participation in exchange for tangible benefits. Politicians, in turn, offer public goods or private advantages through these networks, positioning criminal actors as electoral intermediaries (Albarrac\u0026iacute;n, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Both sides calculate the value of intermediation by the results it produces. For candidates, higher participation reduces the cost of bringing voters to the polls. For intermediaries, controlling voter access and blocking rivals' information increases the chances of electing aligned candidates (Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Manso, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCriminal groups depend on these practices to reinforce authority and expand influence relative to the state (Lins \u0026amp; Machado, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). They use violence and illicit resources to co-opt community leaders and alter electoral outcomes (Arias, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Albarrac\u0026iacute;n (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) calls the situation where organized crime violence reduces candidates and reshapes voter demand 'criminalized electoral politics.' These groups frequently serve as political intermediaries who exchange goods and services for votes (Cano \u0026amp; Duarte, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Zaluar \u0026amp; Concei\u0026ccedil;\u0026atilde;o, 2007; Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Manso, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Paramilitary groups and factions, my focus here, engage with the political system through distinct forms and levels of governance.\u003c/p\u003e \u003cp\u003eParamilitary groups, typically composed of active and former agents of state repressive forces, maintain formal and stable ties with local politicians (Daly, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Manso, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). To legitimize their authority and intervene in electoral processes, they offer security services, public goods, and conflict mediation (Arias, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Drug-trafficking factions display more fluid and situational political connections. Their actions concentrate on territorial domination and protection of illicit markets (Misse, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Duarte, 2019). Paramilitary actors actively promote allies and suppress rivals. Factions prioritize local hegemony and do not always participate directly in electoral contests (Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Lins \u0026amp; Machado, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Hirata et al., 2022).\u003c/p\u003e \u003cp\u003eThese groups also differ in their relationships with the communities they control. Drug factions treat the local population as an operational base and source of legitimacy, though their primary markets are often outside the territory they govern (Misse, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2007\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Paramilitary networks, in turn, extract resources directly from residents by charging fees, selling services, and exercising social control. This structure reflects a clientelist logic rooted in police officers who operate in the same neighborhoods (Cano \u0026amp; Duarte, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The guiding hypothesis is that these differences produce distinct patterns of electoral participation and registration. Paramilitary control raises participation and voter registration. Factions show more volatile patterns, as previous studies suggest.\u003c/p\u003e \u003cp\u003eElectoral violence by armed actors discourages voters by raising the cost of voting, especially when targeting electoral authorities or polling stations (Ley, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Van Baalen, 2024). It undermines trust in elections, weakens beliefs in the efficacy of the vote, and reduces participation in violent environments (Trelles \u0026amp; Carreras, 2012; Ley, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). In Colombia, territories under paramilitary control, analogous in function to Brazil's militias, reduce competition. In those areas, the pressure to vote outweighs the risks. Guerrilla groups, like Brazilian factions, tend to suppress participation by disrupting the electoral process (Gallego, 2018). Coercion and material incentives, such as favors or community benefits, create distinct modes of mobilization (Accardo, De Feo, \u0026amp; De Luca, 2023). Participation helps reinforce local authority and legitimize candidates aligned with criminal structures. The literature on Italian mafias shows that organized crime affects candidate selection and electoral competition (Baraldi et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Daniele, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). This article shows that, with multiple types of armed actors, effects on participation and electoral organization vary with political integration, a dimension underplayed by literature focused on single actor types. Shocks to territorial control produce different effects by type. Lethal violence and armed competition between groups destabilize the mobilization infrastructure. They erode the territorial control that sustains electoral brokerage networks. Institutional police presence can coexist with this infrastructure or even complement it when formal forces and paramilitary groups share personnel. The first shock attenuates the militia effect on participation. The second coexists with it and may amplify it under institutional overlap. This leads to my hypotheses:\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 1\u003c/strong\u003e \u003cp\u003eTerritories under paramilitary control will exhibit higher electoral participation than those without criminal governance or under drug faction control.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 2\u003c/strong\u003e \u003cp\u003eTerritories under paramilitary control will exhibit higher voter registration than those without criminal governance or under drug faction control.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eThe type of criminal organization, its political engagement, and local context shape the outcome. Clear evidence on how these differences affect voting behavior remains scarce, and the literature lacks consensus. Building on this distinction, I expect results consistent with my hypotheses.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eCriminal governance and modes of electoral incorporation\u003c/h2\u003e \u003cp\u003eWhy do criminal actors support politicians? Political alliances help stabilize illicit markets against state regulation, secure protection (often through local police), and grant access to drug-trafficking routes. Why do politicians seek criminal groups? Among other reasons, to reach voters under their direct political control where electoral competition is candidate-centered rather than party-centered (Albarrac\u0026iacute;n, forthcoming). Through criminal control as a mechanism of political mediation, candidates benefit from concentrated voter mobilization capacities.\u003c/p\u003e \u003cp\u003eEven individual coercion targets enough vote shifts to alter local results. Criminal governance conditions patterns of electoral participation. This effect depends on how criminal governance interacts with pre-existing political arrangements, which structure opportunities for electoral engagement. Paramilitary networks, more dependent on the state, and drug-trafficking factions operate under distinct political arrangements, producing divergent effects on participation. The degree of electoral governance varies across contexts, as Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSource: author's elaboration.\u003c/p\u003e \u003cp\u003eI identify three types of political governance structures that produce distinct forms of electoral intervention. The least common, politically integrated criminal governance, is frequently linked to paramilitary groups that systematically control electoral processes over time, from voter registration through election-day mobilization. Their role goes beyond coercion. They allocate economic revenues to local leaders, who connect voters and candidates through neighborhood associations and clientelist ties.\u003c/p\u003e \u003cp\u003eThe second configuration, politically peripheral governance, describes organizations that remain outside integrated political arrangements. These groups intervene only in response to specific threats, using vetoes to block access to politicians aligned with rivals or perceived adversaries. Without the incentives or networks for sustained engagement, they raise the costs of political participation and tend to suppress turnout.\u003c/p\u003e \u003cp\u003eFrom tactical agreements between peripheral and integrated criminal organizations, a hybrid configuration emerges. Peripheral groups operate under integrated electoral arrangements, generally mediated by politically integrated actors. These groups lack stable candidate alliances and rarely develop autonomous mobilization structures. Their electoral involvement is contingent, emerging through occasional alignment with integrated actors who provide candidate access and co-governance, as Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSource: author's elaboration.\u003c/p\u003e \u003cp\u003eI classify the groups a priori, drawing on qualitative literature about their organizational structures and political ties. Militias represent politically integrated governance. CV and ADA represent peripheral cases. TCP represents the hybrid case, given the historical overlap of its networks with militias in the West Zone. For the hybrid type, I predict inconsistent effects across specifications, with low magnitude and unstable direction.\u003c/p\u003e \u003cp\u003eEffects therefore depend on which mode of political governance dominates in each area: integrated groups raise participation, peripheral groups depress it, and hybrid arrangements cannot sustain mobilization over time.\u003c/p\u003e \u003cp\u003eThe process unfolds in two stages. The first establishes local political control. The second mobilizes voters and sets the incentives they face. In some areas, criminal actors exercise uncontested authority. In others, they contend with ongoing disputes involving the state or rival organizations. When control is stable, criminal groups can engage systematically in elections. In areas marked by instability or conflict, political arrangements are often fragile or altogether absent.\u003c/p\u003e \u003cp\u003eGroups under politically integrated governance are directly present in the political arena. They control local economic flows, charge for services, regulate movement, and determine who can participate in community networks. During elections, they suppress competition by blocking rival campaigns and directing votes toward allied candidates (Trudeau, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Voters are incentivized through clientelist ties or coerced into voting, and coercion is sustained over time. These groups support voter registration, monitor who voted, and intimidate residents. They apply sanctions when necessary or suspend access to public goods they control. Control involves material incentives such as benefits for merchants, infrastructure promises, and service delivery, often mediated through residents' associations. In some cases, they receive payments to authorize campaigns or impose electoral discipline. Together, these practices form an electoral governance with consistent mobilization capacity.\u003c/p\u003e \u003cp\u003eGroups under politically peripheral governance, by contrast, intervene in politics only reactively. The absence of stable political structures and the focus on self-protection condition their logic of interference in the electoral process. They impose vetoes, restrict movements, obstruct campaigns, and limit the number of candidates. Criminal presence reduces political options, increases perceived risks, and tends to suppress electoral participation. When mobilization occurs, it responds to specific disputes and lacks continuity.\u003c/p\u003e \u003cp\u003eThe control strategy goes beyond managing illicit markets and monopolizing state presence. Community leaders treat threats to criminal authority as dangerous, reinforcing a clientelist system where residents trade loyalty for basic services. For politically integrated groups, selling services and cultivating political ties follow a market-preservation logic. Violence is explicit and instrumental, used to expand political influence and secure aligned candidates without undermining the conditions needed for voting. Non-electoral criminal organizations rely on more indirect control, prioritizing territorial defense and neutralizing threats. Some areas feature unstable political governance, where disputes between groups generate volatility and fuel recurrent violence. This instability raises voting costs and undermines campaign and voting conditions. In these contexts, participation remains persistently low, shaped by violence.\u003c/p\u003e \u003cp\u003eThe second stage concerns mobilization and incentives. Under politically integrated governance, coercive and clientelist networks maintain continuous relations with candidates, pressure voters to vote, and monitor election-day presence. Voters may be rewarded or punished. By limiting competition, the networks reduce cognitive costs and make support for the endorsed candidate more predictable.\u003c/p\u003e \u003cp\u003eIn contexts with hybrid political configurations, electoral interference is selective and contingent. The effects on participation vary with the degree of alignment with integrated actors, the nature of ties with candidates, and residents' perceived risks, which form through their understanding of governance over time. These groups limit available options and occasionally mobilize to block rivals. Interventions are typically episodic.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eOrganized crime and elections in Rio de Janeiro\u003c/h3\u003e\n\u003cp\u003eThe typology serves as a schema of expectations for interpreting the heterogeneity of results. I treat militias as the closest case to politically integrated governance, CV and ADA as politically peripheral, and the TCP, whose territorial networks historically overlap with militias in parts of the West Zone, as the hybrid case.\u003c/p\u003e \u003cp\u003eBetween 2002 and 2003, the Rio das Pedras Residents' Association organized voter registration campaigns in a neighborhood populated predominantly by migrants from northeastern Brazil. The initiative aimed to elect a 'favela candidate,' a label adopted by paramilitary leader Nadinho. Backed by local leadership, Nadinho had already run for the state legislature in 1998, framing his campaign around the need for political representation to address local demands. The strategy proved effective. With 3,624 votes, he became an alternate member of the PTdoB (Zaluar \u0026amp; Concei\u0026ccedil;\u0026atilde;o, 2007) and was later elected city councilman with strong community support.\u003c/p\u003e \u003cp\u003eMembers of the residents' association went door to door to convince neighbors of the importance of electing a representative from the periphery. Once convinced, residents were directed to designated meeting points, where vans waited to transport them to the city's electoral registry offices (Zaluar \u0026amp; Concei\u0026ccedil;\u0026atilde;o, 2007). The message centered on basic needs: lack of sanitation, infrastructure, and public services. Community members trusted the association and its intermediaries despite their known ties to armed groups. These campaigns successfully mobilized a significant number of new voter registrations.\u003c/p\u003e \u003cp\u003eTwo decades later, the data show the scale of criminal governance over Rio's electorate. Of the more than 4.9\u0026nbsp;million registered voters in 2020, approximately 1.4\u0026nbsp;million, or 29.38% of the electorate, lived in areas under paramilitary control. Another 15.25% resided in territories controlled by Comando Vermelho (CV), while 5.50% lived in areas dominated by Terceiro Comando Puro (TCP). Another 5.11% were in zones disputed by rival groups. Only 41.68% of voters lived in neighborhoods without reported criminal control. In total, nearly 60% of the city's voters were exposed to environments where coercive electoral strategies could be applied or some form of criminal governance operates.\u003c/p\u003e \u003cp\u003eThe main criminal organizations operating in Rio de Janeiro are Comando Vermelho (CV), Amigos dos Amigos (ADA), and Terceiro Comando Puro (TCP). Paramilitary groups, commonly called militias in Brazil, compete for territory and control local markets through extortion schemes targeting essential services such as water, electricity, internet, and informal urban transportation (Arias \u0026amp; Barnes, 2016). Unlike drug factions, which focus on retail drug and firearms trafficking, militias are composed largely of active or former state agents, including police officers and firefighters. They exploit these institutional ties to expand both political and economic influence (Zaluar \u0026amp; Concei\u0026ccedil;\u0026atilde;o, 2007; Manso, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThese groups follow distinct strategies of territorial control and political engagement. Factions rely on direct coercion and maintain relations with security forces through corrupt agreements involving extortion (Misse, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Hirata; Grillo, 2017). Militias benefit from institutional protection by state actors who enable their integration into formal politics and shield their illicit operations (Hirata, Cardoso, \u0026amp; Grillo, 2022). This control lets militias convert neighborhoods into electoral strongholds and secure seats for allied candidates (Motta, 2024).\u003c/p\u003e \u003cp\u003eFactions frequently negotiate with candidates to secure campaign access in the territories they dominate (Arias \u0026amp; Barnes, 2016). They also regulate illegal markets through informal coercion and negotiated order (Hirata \u0026amp; Grillo, 2017). Militias, for their part, maintain organic ties with politicians at different levels of government, pursuing their interests through legislative influence and electoral success (Hirata, Grillo, \u0026amp; Telles, 2022). Militiamen are also known for their connections within the judicial and legislative branches (Arias, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCoercion is central to the electoral strategies of both types of groups. Paramilitary actors impose threats, restrict campaign activity, and apply targeted pressure to ensure residents vote for their preferred candidates. Armed presence on the streets is a form of intimidation. Research has documented threats of eviction for residents who fail to comply with voting instructions, as well as efforts to block rival campaigns in contested areas (Hidalgo \u0026amp; Lessing, 2015).\u003c/p\u003e \u003cp\u003eThe effects of this control are visible in voting patterns. Neighborhoods under militia domination show higher support for candidates affiliated with security forces (Hidalgo \u0026amp; Lessing, 2015), reinforcing the notion that these territories operate as controlled electoral districts. Weak law enforcement and clientelist urban networks, often supported by political allies, sustain these arrangements and strengthen paramilitary power (Arias \u0026amp; Barnes, 2016). Rio de Janeiro reflects a broader pattern of state capture by organized crime. Criminal groups convert territories into coercive electoral bases, secure political representation, and protect their interests through sustained influence over formal institutions.\u003c/p\u003e\n\u003ch3\u003eMap 1 - Territorial control and electoral participation\u003c/h3\u003e\n\u003cp\u003e \u003c/p\u003e \u003cp\u003eSource: author's elaboration based on data from TSE, TRE and Fogo Cruzado.\u003c/p\u003e \u003cp\u003eThe map shows electoral participation rates overlaid on areas under criminal control in Rio de Janeiro. The West Zone (shown on the western, left-hand portion of the map), which concentrates the main militias, especially in Campo Grande and Santa Cruz, shows a concentration of blue circles indicating high participation. In the North Zone, participation is mixed, as is the presence of competing factions.\u003c/p\u003e\n\u003ch3\u003eEmpirical strategy\u003c/h3\u003e\n\u003cp\u003eThe explanatory variable is exposure to criminal governance. The main measure is a binary variable indicating whether a polling station falls within 500 meters of a documented territorial control polygon. As a dynamic complementary measure, I use georeferenced Fogo Cruzado events classified by criminal group in the same radius, capturing temporal variation that static polygons miss. The two types of measures enter the analyses separately. Polygons feed the base specification for mobilization and registration. Fogo Cruzado events feed the robustness tests and the candidate panel.\u003c/p\u003e \u003cp\u003eThe base specification regresses each station's electoral outcome on group exposure, with standardized demographic controls (proportion of elderly voters, low-education voters, women, and biometrically registered voters), neighborhood and year fixed effects, and standard errors clustered by electoral zone. The sample includes approximately 7,033 polling-station observations from the 2008 to 2024 municipal elections.\u003c/p\u003e \u003cp\u003eBeyond the base model, I use four complementary identification strategies. I vary the exposure radius and use continuous log distance to test whether results are sensitive to the window. I estimate interaction models to test whether the militia effect is conditional on lethal violence and on Pacifying Police Unit presence. I use a stacked differences-in-differences design comparing stations moving from low to high militia exposure between consecutive elections with stations that did not change. Finally, in the electoral organization analysis, I shift the unit to examine whether candidates with a higher proportion of votes in militia areas develop a distinct pattern of territorial dependence.\u003c/p\u003e"},{"header":"Data","content":"\u003cp\u003eThe analysis combines official electoral records from the Superior Electoral Court (TSE) with georeferenced data from territorial control polygons in Rio de Janeiro. Electoral data cover the municipal elections of 2008, 2012, 2016, 2020, and 2024. For each year, I use voter profiles and ballot-box vote counts aggregated to the polling-station level, yielding the number of registered voters and the turnout rate. The panel includes approximately 7,033 polling station observations per year, after harmonizing identifiers across elections and keeping only stations with complete information.\u003c/p\u003e \u003cp\u003eI operationalize exposure to criminal governance using georeferenced territorial-control polygons compiled from the Instituto de Seguran\u0026ccedil;a P\u0026uacute;blica (ISP) and Pista News, complemented by armed-incident records from Fogo Cruzado. These sources allow me to classify each polling-station catchment as either under the control or under the influence of one of the criminal organizations under study.\u003c/p\u003e \u003cp\u003eFor the candidate-level analysis, I build a candidate \u0026times; year panel restricted to candidates with at least two contests. Territorial dependence is the proportion of votes obtained at stations under militia control. Spatial concentration is the HHI index over the distribution of votes across stations. UPP deployment provides the exogenous variation for examining how territorial control shocks affect candidates' electoral bases.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eCriminal groups differ systematically in their modes of governance and therefore in their electoral effects. I expect militias, as politically integrated actors, to be associated with higher participation and registration, conditional on territorial stability and attenuated where lethal violence is more intense. For CV and ADA, as peripheral actors, I expect negative or null effects. Given the TCP's ambiguous network character, I do not formulate a directional expectation. These contrasts inform the interpretation of the models below.\u003c/p\u003e \u003cp\u003eI present the results in three blocks. The first documents the effects on mobilization (participation and registration) in the main panel model. The second examines how lethal violence and UPP presence moderate militia mobilization capacity. The third addresses electoral organization, examining whether and when militia presence affects vote concentration and candidates' territorial bases. Evidence is more limited at the aggregate level but more consistent when focused on intra-candidate dynamics.\u003c/p\u003e"},{"header":"Findings","content":"\u003cp\u003eThe main panel model shows that proximity to a militia control polygon is associated with a significant increase in participation (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The result is robust to different exposure radii. The coefficient grows monotonically as the window expands, suggesting that the mobilizing effect extends beyond the immediate vicinity of the station. When all groups enter simultaneously (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), the result holds. CV and ADA are associated with declines in participation. TCP shows effects close to zero and unstable across specifications, consistent with the prediction for the hybrid case. Results remain stable when territories with pre-established militia control are excluded (β\u0026thinsp;=\u0026thinsp;1.42 vs 1.45 in the full model). Estimators robust to cohort heterogeneity confirm the direction and magnitude of the effects. Patterns by alternative exposure radius are detailed in the Appendix.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMilitia and electoral mobilization\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia (500 m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.4546**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0354***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.4297)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0055)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7,033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,033\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eNotes: Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60\u0026ndash;69), low education, women, and biometric registration. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. Results hold with Conley standard errors at radii of 1 to 10 km (Appendix A.2) and with spatial SAR and SEM models (Appendix A.4).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison across Criminal Groups (Joint Model)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCriminal group\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.3117**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0329***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.4155)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0055)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.8132*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0144**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3351)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0047)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTCP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.0421\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0024\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3510)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0069)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.8645*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0019\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3792)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0056)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7,033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,033\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eNotes: Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. All criminal groups enter simultaneously in the same model. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60\u0026ndash;69), low education, women, and biometric registration. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. Results hold with Conley standard errors at radii of 1 to 10 km (Appendix A.2) and with spatial SAR and SEM models (Appendix A.4).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eRobustness: Alternative Measures of Militia Proximity\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpecification\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e250 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.6198\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0240***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3969)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0056)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e500 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.3117**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0329***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.4155)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0055)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1,000 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.9827***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0450***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.5542)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0071)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDist. (log)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.2879***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0065***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0742)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0010)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7,033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,033\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eNotes: Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. Each row is an alternative specification of the militia treatment variable. Criminal controls for other groups included in all specifications.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eVoter registration follows the same pattern. Militia areas have more registered voters. Unlike the turnout result, this effect is already significant at the smallest exposure radius. CV shows a negative and significant effect on registration, reinforcing the interpretation that these territories have smaller electorates. TCP and ADA show no effect distinguishable from zero. Controls follow the expected pattern. Biometric registration is associated with a smaller electorate due to the institutional cost of re-registration.\u003c/p\u003e \u003cp\u003eThe Appendix details robustness by radius. The pattern is monotonic, with the coefficient growing as the exposure window expands. The effect radiates beyond the immediate vicinity of the polling station. Using log distance as a continuous variable, the sign reverses as expected, reinforcing the spatial interpretation.\u003c/p\u003e \u003cp\u003eTo check whether the effect depends on comparing consolidated militia territories with uncontrolled ones, I re-estimated the main model excluding stations already militia-controlled in the first panel period. The coefficient is stable (β\u0026thinsp;=\u0026thinsp;1.42 vs 1.45 in the full model). Identification does not rest on structural differences between old and new militia territories.\u003c/p\u003e\n\u003ch3\u003eEffect conditions\u003c/h3\u003e\n\u003cp\u003eThe mobilizing effect of militias is not constant. I examine two conditions that may modify it. The first is the intensity of lethal violence around the polling station. The second is the presence of Pacifying Police Units. The first test rests on the idea that mobilization capacity depends on territorial control, which violence undermines. The second examines whether organized state presence disrupts or coexists with militia mobilization networks. The UPP result is counterintuitive and deserves particular attention.\u003c/p\u003e \u003cp\u003eThe argument behind the interaction with lethal violence is straightforward. If militia mobilization depends on stable territorial control, then areas of high violence, where control is contested, should exhibit a smaller effect. The interaction with UPPs is less obvious. The conventional expectation is that organized police presence would reduce militia mobilization capacity. But militia clientelist networks may survive and adapt to police presence, especially when UPP officers themselves have paramilitary ties.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eConditional Effect of Militia. Interaction with Lethal Violence\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.5186***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0308***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3389)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0053)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLethal violence (z)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.8399***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0343***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.2851)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0049)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia \u0026times; Violence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;1.0801***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0147**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3030)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0050)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7,011\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,011\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eNotes: Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60\u0026ndash;69), low education, women, and biometric registration. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. Moderator (lethal violence\u0026thinsp;=\u0026thinsp;intentional homicides ISP/CISP) standardized to mean zero and standard deviation one within year. Other criminal groups (CV, TCP, ADA) included as controls.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ea. Militia effect by level of lethal violence\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLethal violence percentile\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ep10 (low violence)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.2798***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0412***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.4827)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0080)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ep25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.9932***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0372***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.4217)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0069)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ep50 (median)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.4126***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0293***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.3250)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0051)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ep75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.4963\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0169***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.3156)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0045)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ep90 (high violence)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.1820\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0076\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.4246)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0063)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eNotes: Marginal effect\u0026thinsp;=\u0026thinsp;β(militia) + β(militia \u0026times; violence) \u0026times; z_intentional_homicides, with standard errors by delta method. Moderator: intentional homicides (ISP/CISP), standardized within year. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e4\u003c/span\u003ea details the marginal effect of militia presence by level of lethal violence. In low-violence contexts, the positive effect on participation is substantial. As violence intensifies, the effect diminishes. At the 90th percentile of lethal violence, the effect loses significance and turns negative. Militia control ceases to operate as a mobilization structure where armed conflict is intense. This gradient shows that militia mobilization efficacy depends on territorial stability. Where other armed groups contest control, the incentives sustaining electoral brokerage weaken. The result on voter registration follows the same direction, though less pronounced.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eConditional Effect of Militia. Interaction with UPP Presence\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.1417***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0293***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3177)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0054)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eActive UPP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;2.6416***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0347***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.4821)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0058)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia \u0026times; UPP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.9272**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0090\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.7188)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0097)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7,012\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eNotes: Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. Exposure variable: presence of polygon at 500 m (dummy). Demographic controls: proportion of elderly (60\u0026ndash;69), low education, women, and biometric registration. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. Active UPP\u0026thinsp;=\u0026thinsp;dummy for presence of active Pacifying Police Unit in the electoral period. Other criminal groups (CV, TCP, ADA) included as controls.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ea. Marginal Effect of Militia by UPP Presence\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInstitutional context\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Turnout\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Log Electorate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo UPP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.1417***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0293***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.3177)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0054)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWith Active UPP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.0688***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0382***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e(0.7345)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e(0.0102)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eMarginal effect calculated at upp_active\u0026thinsp;=\u0026thinsp;0 and upp_active\u0026thinsp;=\u0026thinsp;1. Standard errors by delta method. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. Notes: Each interaction column corresponds to a separate model including the other criminal exposures, controls and fixed effects.\u003c/p\u003e \u003cp\u003eThe interaction with UPPs reveals a counterintuitive result. As Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows, stations near militia polygons with an active UPP exhibit even higher participation than those without. On voter registration, the interaction is not significant. This suggests that militia mobilization networks are stable enough to persist, and possibly intensify, when coexisting with formal police presence. This is unsurprising because, like UPP officers, militiamen are themselves drawn from police forces.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eElectoral organization and base territorialization\u003c/h2\u003e \u003cp\u003eThe evidence on electoral organization, vote concentration and candidate territorialization, is more heterogeneous. In the aggregate model with neighborhood and year fixed effects, militia presence has no significant effect on the most common concentration indicators. The Herfindahl-Hirschman Index and top candidate's vote share are close to zero and insignificant. The most consistent result is for the effective number of parties. Stations near militias have a slightly more restricted competitive field, though the effect is modest. Results appear in Table\u0026nbsp;\u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e6\u003c/span\u003e in the Appendix and serve as complementary evidence. The static specification with 2019 polygons limits what can be inferred about dynamics. I cannot separate militia presence effects from pre-existing neighborhood characteristics that the polygons map.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eElectoral Organization. Criminal Groups (Joint Model)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCriminal group\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) HHI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) ENP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3) share_top1\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;1.6576\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0062\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0017)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.1212)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0038)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.0031*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.6039\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.0068*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0013)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.9264)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0034)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTCP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0058\u0026dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;3.3851*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0149*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0033)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.3150)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0068)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;2.5110\u0026dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0059\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0029)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.4303)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0069)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5,660\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5,660\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5,660\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003eNotes: All criminal groups enter simultaneously in the same model. Treatment variable: territorial dominance (polygon Pista News / ISP, ~\u0026thinsp;2019). Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. Demographic controls: proportion of elderly (60\u0026ndash;69), low education, women, and biometric registration. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab9\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eElectoral Organization. Robustness by Specification\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpecification\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) HHI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) ENP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3) share_top1\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia dom. (polygon)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;1.6398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0062\u0026dagger;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0017)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.1073)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0037)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFC militia (dummy, 500 m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0007\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;1.9921*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0059\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0016)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.8146)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0036)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFC militia (intensity, 500 m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;1.8179\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0010\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0047)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.9608)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0094)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5,660\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5,660\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5,660\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eneighborhood\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003eNotes: Each row is an alternative specification of the militia treatment variable. Unit: polling station \u0026times; year. FE: neighborhood\u0026thinsp;+\u0026thinsp;year. Standard errors clustered by electoral zone (G\u0026thinsp;=\u0026thinsp;97), in parentheses. Demographic controls: proportion of elderly (60\u0026ndash;69), low education, women, and biometric registration. Significance: *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe design that captures temporal variation produces more consistent evidence of competitive reorganization. The stacked differences-in-differences compares stations whose militia exposure radius changes between consecutive elections to stable stations in the same period. Table\u0026nbsp;\u003cspan refid=\"Tab10\" class=\"InternalRef\"\u003e8\u003c/span\u003e shows that when a station moves from low to high militia exposure, the effective number of parties falls significantly. The HHI and the top candidate's share point in the expected direction. Together, these results suggest that expanding militia presence narrows the competitive field, even if the effect does not appear uniformly across the three measures.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab10\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eElectoral Organization. Stacked DiD (FC Exposure Switchers)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) HHI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) ENP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3) share_top1\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia \u0026times; Post (DiD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0045\u0026dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;5.7631**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0107\u0026dagger;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0026)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.8054)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0064)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4,122\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4,122\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4,122\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003estation\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003estation\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003estation\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eStacked Differences-in-Differences (DiD) design. Treatment\u0026thinsp;=\u0026thinsp;station moving from low to high militia exposure (Fogo Cruzado, 500 m) between two consecutive elections. Control\u0026thinsp;=\u0026thinsp;stations without exposure change in the same period. Each electoral pair constitutes an independent stack; FE station\u0026thinsp;+\u0026thinsp;year. HHI\u0026thinsp;=\u0026thinsp;electoral Herfindahl-Hirschman index; ENP\u0026thinsp;=\u0026thinsp;effective number of parties; share_top1\u0026thinsp;=\u0026thinsp;top candidate's vote share. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10.\u003c/p\u003e \u003cp\u003eThe strongest evidence of territorialization appears at the candidate level. Table\u0026nbsp;\u003cspan refid=\"Tab11\" class=\"InternalRef\"\u003e9\u003c/span\u003e shows that candidates closer to militia polygons get a larger share of their votes from those territories, regardless of total vote count. A UPP entry shock significantly reduces this territorial dependence, suggesting militias structure specific candidates' electoral bases and police presence disrupts this tie. Table\u0026nbsp;\u003cspan refid=\"Tab12\" class=\"InternalRef\"\u003e11\u003c/span\u003e completes the picture. Candidates with more spatially concentrated bases depend more on militia territories, reinforcing the link between electoral territorialization and militia control. This pattern is hard to explain by demographic characteristics alone. It points to a structural link between vote organization and territorial control.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab11\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 9\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCandidates' territorial base\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpecification\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Militia territorial dep.\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMilitia intensity (FC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.7221***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.1676)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e% militia stations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.2284***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0707)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUPP shock (% stations)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;0.9764***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.1474)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,945\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDem. controls\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ecandidate\u0026thinsp;+\u0026thinsp;year\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eUnit: candidate \u0026times; year (candidates with \u0026ge;\u0026thinsp;2 elections, N\u0026thinsp;=\u0026thinsp;827). Territorial dependence\u0026thinsp;=\u0026thinsp;proportion of votes obtained at stations under militia control. FE: candidate\u0026thinsp;+\u0026thinsp;year. Control: log(total votes). Militia intensity (FC) = mean intensity of militia Fogo Cruzado events at candidate's polling stations. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab12\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 11\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eTerritorialization of the electoral base\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1) Militia dep.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2) Spatial HHI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3) Militia dep.\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLog(votes)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0032\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;0.0066***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0077\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0030)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0005)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0033)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpatial HHI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.6798***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.1077)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6,203\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6,203\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6,203\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eyear\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eyear\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eyear\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCluster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eelectoral zone\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eUnit: candidate \u0026times; year (candidates with \u0026ge;\u0026thinsp;10 votes, N\u0026thinsp;=\u0026thinsp;6,203). Year FE. Territorial dependence\u0026thinsp;=\u0026thinsp;proportion of votes obtained at stations\u0026thinsp;\u0026le;\u0026thinsp;500 m from militia polygon. Spatial HHI\u0026thinsp;=\u0026thinsp;Herfindahl-Hirschman index calculated over the distribution of candidate votes across stations. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; \u0026dagger; p\u0026thinsp;\u0026lt;\u0026thinsp;0.10.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eQualitative evidence\u003c/h2\u003e \u003cp\u003eTo describe the mechanism tested by the statistical models, I added a qualitative component explaining the political processes underlying the estimated effects. The qualitative corpus consists of triangulated secondary sources. I consulted official documents produced by the Parliamentary Commission of Inquiry on Militias in the Rio de Janeiro State Legislative Assembly.\u003c/p\u003e \u003cp\u003eThrough a survey of 105 news articles in four local newspapers (Jornal do Brasil, O Dia, O Globo, Extra) covering 2006 to 2020, I built a qualitative database. It documents voter co-optation and electoral capture by organized crime. I selected articles using keywords related to militias, elections, voter registration, and residents' associations. I included only concrete records, excluding non-investigative articles. The analytical unit is the event. Reports function as documentary sources.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eMechanisms\u003c/h2\u003e \u003cp\u003eIn Rio de Janeiro's West Zone, and often in the North Zone, criminals go beyond direct coercion in their mobilization strategies on and before election day. Criminal organizations govern the electoral process. Residents of Campo Grande, Santa Cruz, and Jacarepagu\u0026aacute;, to name only a few neighborhoods, report that militiamen map individual families through residents' associations. These organizations conduct community population censuses and monitor local political behavior.\u003c/p\u003e \u003cp\u003eFirst, militiamen summon residents to meetings in social centers they control, usually the association headquarters. There, criminals demand that residents bring their identity documents and voter registration cards for entry into the militia's database, which identifies each voter's polling station. In Santa Cruz, for example, militiamen from the Babu clan collected voter card numbers under the promise of improvements. They conditioned basic urban services on proof of electoral regularity and support for the 'house' candidate, former state deputy Jorge Babu. Voters know that non-compliance will be discovered and punished, and they believe it. In more extreme cases, as reported in Muzema, the militia physically collected residents' voter cards weeks before the election. The documents were returned on the eve of voting, accompanied by the local candidate's flyer and instructions for attendance on election day.\u003c/p\u003e \u003cp\u003eWhen a voter is not registered\u0026mdash;common given the migrant population\u0026mdash;militiamen take them to electoral registry offices to regularize their status. Militias also register voters from other cities or neighborhoods through targeted strategies. Rio's electoral court itself reported cases of anomalous voter registration and the use of public vehicles, such as kombis, to transport them to registry offices. Militiamen also handle the bureaucratic side, sometimes providing false proof of residency for submission to the local electoral authority.\u003c/p\u003e \u003cp\u003eOn election day, the militia sets up real-time mobilization logistics. Alternative transportation, the militiamen's initial income source since the late 1990s, is mobilized in a kind of 'vote caravan.' The clandestine company 'Via\u0026ccedil;\u0026atilde;o Coringa,' run by militiamen in Santa Cruz and Paci\u0026ecirc;ncia, displayed a clown smile as a group identifier. It transported residents to their polling stations free of charge.\u003c/p\u003e \u003cp\u003eMilitiamen position themselves at the doors of schools and polling stations as parallel election monitors. According to residents' accounts, police officers assigned to election security are seen conversing with militiamen or deliberately ignoring their activity. This symbiosis discourages resistance and undermines ballot secrecy, even though the vote inside the booth cannot be identified. To validate who voted, militiamen demand the delivery of the voting receipt issued by the electoral justice. In other cases, there is even photo confirmation of votes sent from the ballot machines.\u003c/p\u003e \u003cp\u003eAfter the election, failure brings consequences. Reports range from suspended access to public goods like internet service to eviction of residents who refuse to cooperate. Victims report late-night approaches with 24 to 48-hour deadlines to leave under threat of death for all family members.\u003c/p\u003e \u003cp\u003eOutcomes can also be positive for the community when militia brokerage succeeds. Militiamen offer, in exchange for votes, the suspension of extortionate security fees and internet access for the months following the election, November and December. Militiamen create an atmosphere of pressure so that no one stays home.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eConcluding remarks\u003c/h2\u003e \u003cp\u003eI show that the political effect of territorial control on elections varies with the degree of political integration of criminal organizations, measured by the depth of ties to electoral elites, the stability of candidate alliances, and systematic mobilization capacity. Paramilitary groups, under politically integrated governance, mobilize voters and produce higher participation through coercion and clientelist incentives. Drug-trafficking factions, under politically peripheral governance, intervene more selectively, without systematic mobilization. Their influence on participation varies with context, ranging from suppression to no distinguishable effect. The TCP, whose territorial networks partly overlap with militias in the West Zone, shows inconsistent effects across specifications, with low magnitude and unstable direction. This pattern is consistent with the prediction for the hybrid case.\u003c/p\u003e \u003cp\u003eAcross multiple identification strategies, politically integrated groups exercise structured political control, act as intermediaries between voters and candidates, and produce higher participation in their areas of influence. This mobilization works through direct coercion, suppression of competition, and economic benefits, consolidating their role as local political actors. The militia coefficient is stable when territories with pre-established control are excluded and persists under estimators robust to cohort heterogeneity. Politically peripheral actors do not systematically mobilize voters. Their actions depend on territorial stability and the political interests at stake.\u003c/p\u003e \u003cp\u003eThe central contribution is the distinction between forms of political engagement by criminal organizations. Electoral behavior is predicted both by armed presence and by how crime articulates with institutional politics. This approach broadens analytical perspectives on Brazilian politics and enables comparisons with other contexts of territorial capture.\u003c/p\u003e \u003cp\u003eMy results challenge the assumption that higher electoral participation corresponds to greater democratic quality. Criminal groups reorganize voting incentives, distort competition, and depoliticize electoral choice. The inference on coerced voting rests on descriptive evidence such as individual family mapping, physical collection of voter registration cards, organized transportation on election day, attendance monitoring, and post-electoral sanctions documented in the qualitative corpus. Under these conditions, voting ceases to express the voter's will. It becomes a mechanism that legitimizes domination through power structures that permeate the state.\u003c/p\u003e \u003cp\u003eUnderstanding voting behavior under criminal governance requires abandoning linear conceptions of democratic participation. In many territories, voting is not a free choice but an imposition. Politics becomes an instrument of domination managed by a hybrid state where formal and informal authorities intertwine to shape democratic conditions. Measuring electoral participation without accounting for this context distorts the true state of democracy.\u003c/p\u003e \u003cp\u003eI contribute to the comparative politics of criminal governance, drawing on the case of Rio de Janeiro. Militias, composed largely of active and former agents of repressive forces, and drug-trafficking factions coexist in the same political system with opposite ties to the state and electoral machines. The distinction between politically integrated and politically peripheral governance, combined with the nature of state relations, helps explain the varied effects of criminal governance on electoral incorporation. The framework applies to contexts like Colombia, Mexico, or Italy. This lens applies to other Latin American countries where armed groups interact with public institutions and shape electoral processes.\u003c/p\u003e \u003cp\u003eElectoral behavior under criminal governance emerges from structured relations between the state, criminal actors, and voters. Higher participation in territories dominated by politically integrated actors reflects durable arrangements of coercion, intermediation, and institutional ties rooted in historical struggles over territory and authority.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research was supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, Brazil) through a doctoral sandwich fellowship (Programa de Doutorado Sanduíche no Exterior, PDSE, 2025–2026) at the University of Illinois Chicago, awarded to Igor Novaes Lins.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Competing Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author declares no competing financial or personal interests that could have influenced the work reported in this article.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlbarrac\u0026iacute;n J, Karolczak M, Rodrigo, Wolff J (2025) Violence against civil society actors in democracies: Territorialization of criminal economies and the assassination of social activists in Brazil. J Peace Res 62(5):1411\u0026ndash;1427. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/00223433251347784\u003c/span\u003e\u003cspan address=\"10.1177/00223433251347784\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlbarrac\u0026iacute;n J (2018) Criminalized Electoral Politics in Brazilian Urban Peripheries. Crime Law Social Change 69(4):553\u0026ndash;575\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlbarrac\u0026iacute;n J Forthcoming. Undermining Democracy from the Peripheries: Criminalized Electoral Politics in Brazil. Book Manuscript\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlesina A, Piccolo S, Pinotti P (2019) Organized crime, violence, and politics. Rev Econ Stud 86(2):457\u0026ndash;499\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnduiza E, and Guillem Rico (2022) Sexism and the Far-Right Vote: The Individual Dynamics of Gender Backlash. American Journal of Political Science\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArias ED (2006) The Dynamics of Criminal Governance: Networks and Social Order in Rio de Janeiro. J Latin Am Stud 38(2):293\u0026ndash;325\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArias ED (2013) The Impacts of Differential Armed Dominance of Politics in Rio de Janeiro, Brazil. Stud Comp Int Dev 48:263\u0026ndash;284\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaraldi A, Laura E, Papagni, and Marco Stimolo (2024) Neutralizing the Tentacles of Organized Crime. Assessment of the Impact of an Anti-Crime Measure on\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarnes N (2017) Criminal Politics: An Integrated Approach to the Study of Organized Crime, Politics, and Violence. Perspect Politics 15(4):967\u0026ndash;987\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlattman C et al (2024) Gang Rule: Understanding and Countering Criminal Governance. Rev Econ Stud : rdae079\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoas T, Smith AE (2015) Religion and the Latin American Voter. In: Ryan E, Carlin MM, Singer, Zechmeister EJ (eds) The Latin American Voter: Pursuing Representation and Accountability in Challenging Contexts. University of Michigan Press, Ann Arbor, pp 112\u0026ndash;134\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBohn SR (2004) Evang\u0026eacute;licos no Brasil: Perfil Socioecon\u0026ocirc;mico, Afinidades Ideol\u0026oacute;gicas e Determinantes do Comportamento Eleitoral. Opini\u0026atilde;o P\u0026uacute;blica 10:288\u0026ndash;338\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBullock J (2019) Criminal Dominance and Campaign Concentration\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCano I, Duarte (2012) Thais. No sapatinho: a evolu\u0026ccedil;\u0026atilde;o das mil\u0026iacute;cias no Rio de Janeiro (2008\u0026ndash;2011). LAV, Laborat\u0026oacute;rio de An\u0026aacute;lise da Viol\u0026ecirc;ncia. LAV-UERJ)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarreir\u0026atilde;o S\u0026eacute;rgio et al (2022) Polariza\u0026ccedil;\u0026atilde;o, Fragmenta\u0026ccedil;\u0026atilde;o e Participa\u0026ccedil;\u0026atilde;o Eleitoral no Brasil. Revista Brasileira de Ci\u0026ecirc;ncia Pol\u0026iacute;tica 29(1):55\u0026ndash;80\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarreras M, N\u0026eacute;stor Casta\u0026ntilde;eda-Angarita (2014) Who Votes in Latin America? A Test of Three Theoretical Perspectives. Comp Polit Stud 47(8):1079\u0026ndash;1104\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCepaluni G, Daniel Hidalgo F (2016) Compulsory Voting Can Increase Political Inequality: Evidence from Brazil. Political Anal 24(2):273\u0026ndash;280\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCigtar V (2018) Party System Fragmentation and Electoral Volatility in Eastern Europe. East Eur Politics Soc 32(3):621\u0026ndash;648\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eC\u0026oacute;rdova A (2019) Living in Gang-Controlled Neighborhoods: Impacts on Electoral and Nonelectoral Participation in El Salvador. Latin Am Res Rev 54(1):201\u0026ndash;221\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorrales J (2020) The Expansion of LGBT Rights in Latin America and the Backlash. In The Oxford Handbook of Global LGBT and Sexual Diversity Politics, 185\u0026ndash;200\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCPI das Mil\u0026iacute;cias (2008) Relat\u0026oacute;rio Final da Comiss\u0026atilde;o Parlamentar de Inqu\u0026eacute;rito Destinada a Investigar a A\u0026ccedil;\u0026atilde;o de Mil\u0026iacute;cias no \u0026Acirc;mbito do Estado do Rio de Janeiro. Assembleia Legislativa do Estado do Rio de Janeiro (Alerj), Rio de Janeiro\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDahl RA (1998) On Democracy. Yale University Press, New Haven\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaly SZ (2022) How do violent politicians govern? The case of paramilitary-tied mayors in Colombia. Br J Polit Sci 52(4):1852\u0026ndash;1875\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaniele G (2019) Strike One to Educate One Hundred: Organized Crime, Political Selection and Politicians\u0026rsquo; Ability. J Econ Behav Organ 159:650\u0026ndash;662\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDettrey BJ, Leslie A, Schwindt-Bayer (2009) Voter Turnout in Presidential Democracies. Comp Polit Stud 42(10):1317\u0026ndash;1338\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeldmann AE, Juan PL (2022) Gobernanza Criminal y la Crisis de los Estados Latinoamericanos Contempor\u0026aacute;neos. Ann Rev Sociol 48(1):S\u0026ndash;1\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGallego A (2015) Unequal Political Participation Worldwide. Cambridge University Press, Cambridge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeys B (2006) Explaining Voter Turnout: A Review of Aggregate-Level Research. Electoral Stud 25(4):637\u0026ndash;663\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrupo de Estudos dos Novos Ilegalismos (GENI) and Fogo Cruzado (2024) Grande Rio sob Disputa: Mapeamento dos Confrontos por Territ\u0026oacute;rios. Niter\u0026oacute;i: GENI/UFF and Fogo Cruzado-RJ. Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://geni.uff.br/2024/06/05/grande-rio-sob-disputa-mapeamento-dos-confrontos-por-territorios/\u003c/span\u003e\u003cspan address=\"https://geni.uff.br/2024/06/05/grande-rio-sob-disputa-mapeamento-dos-confrontos-por-territorios/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed June 14, 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHealy A, and Neil Malhotra (2013) Retrospective Voting Reconsidered Annual Rev Political Sci 16:285\u0026ndash;306\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLessing B (2021) Conceptualizing Criminal Governance. Perspect Politics 19(3):854\u0026ndash;873\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLey S (2018) To vote or not to vote: how criminal violence shapes electoral participation. J Conflict Resolut 62(9):1963\u0026ndash;1990\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLijphart A (1997) Unequal Participation: Democracy\u0026rsquo;s Unresolved Dilemma. Am Polit Sci Rev 91:1\u0026ndash;14\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLins IN, Machado CAM (2023) O Crime \u0026Eacute; Pol\u0026iacute;tico: Elementos Te\u0026oacute;ricos para uma An\u0026aacute;lise Neoinstitucionalista das Mil\u0026iacute;cias no Rio de Janeiro. Revista Brasileira de Ci\u0026ecirc;ncia Pol\u0026iacute;tica. 42:e271780\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLins I, Novaes, and Carlos Machado (2024) A Geografia do Voto das Mil\u0026iacute;cias na Cidade do Rio de Janeiro. Revista de Ci\u0026ecirc;ncia Pol\u0026iacute;tica, Teoria \u0026amp; Pesquisa, p e024008\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLins IN (2023) Da Baixada \u0026agrave; Zona Sul: Caminhos da Viol\u0026ecirc;ncia Pol\u0026iacute;tica de Ra\u0026ccedil;a no Rio de Janeiro. Revista Brasileira de Seguran\u0026ccedil;a P\u0026uacute;blica 17(1):188\u0026ndash;207\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMadrid Ra\u0026uacute;lL (2008) The Rise of Ethnopopulism in Latin America. World Polit 60(3):475\u0026ndash;508\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMafia Violence in Italy J Econ Behav Organ 223: 57\u0026ndash;85\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMainwaring S et al (2015) The Left and the Mobilization of Class Voting in Latin America. In The Latin American Voter: Pursuing Representation and Accountability in Challenging Contexts, eds. Ryan E. Carlin, Matthew M. Singer, and Elizabeth J. Zechmeister. Ann Arbor: University of Michigan Press, 99\u0026ndash;105\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManso BP (2020) A Rep\u0026uacute;blica das Mil\u0026iacute;cias: Dos Esquadr\u0026otilde;es da Morte \u0026agrave; Era Bolsonaro. Todavia, S\u0026atilde;o Paulo\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMisse M (2011) Crime organizado e crime comum no Rio de Janeiro: diferen\u0026ccedil;as e afinidades. Revista de sociologia e pol\u0026iacute;tica 19:13\u0026ndash;25\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMisse M (2007) Mercados Ilegais, Redes de Prote\u0026ccedil;\u0026atilde;o e Organiza\u0026ccedil;\u0026atilde;o Local do Crime no Rio de Janeiro. Estudos Avan\u0026ccedil;ados 21(61):139\u0026ndash;157\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNorris P (2011) Democratic Deficit: Critical Citizens Revisited. Cambridge University Press, Cambridge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePessoa P (2023) Political Competition When Gangs Rule. Effects of Removing Armed Groups\u0026rsquo; Territorial Control in Brazil\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePol\u0026iacute;cia F (2024) Relat\u0026oacute;rio Final do Inqu\u0026eacute;rito Policial 2023.0059871-SR/PF/RJ \u0026ndash; Inq n.\u0026ordm; 4954/DF. Superintend\u0026ecirc;ncia Regional no Estado do Rio de Janeiro, Rio de Janeiro\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRibeiro E, Aparecido J, Borba, Rafael da Silva (2015) Comparecimento Eleitoral na Am\u0026eacute;rica Latina: Uma An\u0026aacute;lise Multin\u0026iacute;vel Comparada. Revista de Sociologia e Pol\u0026iacute;tica 23(54):91\u0026ndash;108\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRomero B, More COPS, Higher, Turnout? (2025) APSA Preprints. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.33774/apsa-2024-bwwpd-v2\u003c/span\u003e\u003cspan address=\"10.33774/apsa-2024-bwwpd-v2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRozema R (2007) Paramilitares y Violencia Urbana en Medell\u0026iacute;n. Colombia Foro Int : 535\u0026ndash;550\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchumpeter JA (1942) Capitalism, Socialism and Democracy. Harper \u0026amp; Brothers, New York\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSnyder R, Ang\u0026eacute;lica D-M (2009) Does Illegality Breed Violence? Drug Trafficking and State-Sponsored Protection Rackets. Crime Law Social Change 52:253\u0026ndash;273\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrejo (2020) Guillermo, and Sandra Ley. Votes, drugs, and violence: The political logic of criminal wars in Mexico. Cambridge University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrelles A (2012) Bullets and votes: Violence and electoral participation in Mexico. J Politics Latin Am 4(2):89\u0026ndash;123\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrudeau J (2022) How Criminal Governance Undermines Elections. Politics and International Relations\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eValdini ME, Lewis-Beck MS (2018) Economic Voting in Latin America: Rules and Responsibility. Am J Polit Sci 62(2):410\u0026ndash;423\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZaluar A, Isabel Siqueira Concei\u0026ccedil;\u0026atilde;o (2007) Favelas sob o Controle das Mil\u0026iacute;cias no Rio de Janeiro. S\u0026atilde;o Paulo em Perspectiva 21(2):89\u0026ndash;101\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Map 1","content":"\u003cp\u003eMap 1 is available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Coordenação de Aperfeicoamento de Pessoal de Nível Superior","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"criminal governance, electoral participation, militias, drug-trafficking factions, criminalized electoral politics","lastPublishedDoi":"10.21203/rs.3.rs-9534248/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9534248/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWhy do some armed criminal groups mobilize voters while others depress participation? Existing accounts treat criminal governance as electorally uniform, rarely distinguishing forms that differ in political integration into local electoral networks. I argue that the electoral effects of criminal governance depend on how deeply armed groups are embedded in those networks: durable ties with electoral elites, stable candidate alliances, and accumulated voter mobilization capacity. Politically integrated criminal governance increases participation by organizing territorial brokerage, candidate access, and election-day coordination. Peripheral criminal governance produces weaker, more unstable effects, including suppression. Using geocoded armed-control polygons and polling-place electoral returns across five municipal elections in Rio de Janeiro, I estimate the effects of exposure to different criminal groups on turnout and voter registration within a panel design with neighborhood and year fixed effects. Militia-controlled areas exhibit significantly higher turnout and voter registration, with effects that attenuate under lethal violence but amplify where police forces overlap with paramilitary networks. Drug-trafficking factions generate weaker, heterogeneous, and sometimes negative associations. These findings recast criminal governance as a heterogeneous political order whose electoral effects depend on territorial brokerage and political embeddedness, not merely coercive capacity.\u003c/p\u003e","manuscriptTitle":"Voting under criminal governance: election mobilization by criminal organizations","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-28 10:25:04","doi":"10.21203/rs.3.rs-9534248/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"03904c44-9060-412a-86c3-b7823c9870bb","owner":[],"postedDate":"April 28th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":67036153,"name":"Comparative Political Science"}],"tags":[],"updatedAt":"2026-04-28T10:25:04+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-28 10:25:04","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9534248","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9534248","identity":"rs-9534248","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}