Abstract
Police militarization is prevalent worldwide. Despite growing interest in the consequences of militarization, few studies focus on its causes. Through the case of Brazilian municipal police forces, we explore the connection between partisan ideology and police militarization. Some research shows that leftist administrations in democracies are more likely to emphasize human rights and less likely to militarize the police. However, leftist administrations might militarize police forces more than centrist and right-wing parties to compensate for perceptions of lower competence to deal with security issues, especially when they face high levels of violent crime. Using inverse probability of treatment weights, we show an association between leftist mayors and higher militarization of municipal police forces. Furthermore, in contexts of high homicide rates, mayors of Brazil’s leading leftist party, the Workers' Party, tend to militarize local police more than others.
From the incorporation of military materials and tactics in the police, 1 to the domestic deployment of military units and socialization processes, the militarization of internal security has become common across the world. Examples of this phenomenon are present in countries as diverse as the Philippines (Kenny and Holmes 2020), South Africa (Lamb 2018), or Mexico (Espinosa and Rubin 2015). Recent research suggests that militarizing internal security can lead to an increased likelihood of abuse by security forces (Espinosa and Rubin 2015; Flores-Macias and Zarkin 2024; Magaloni and Rodriguez 2020) without reductions in crime or violence (Muggah, Garzón, and Suárez, 2018; Mummolo 2018). In the United States, for example, there is increasing scrutiny of how the growing militarization of police departments is related to excessive use of force (Delehanty et al., 2017; Gunderson et al., 2021) and the violent repression of individuals based on their race (Ross 2015). In Latin America, many countries have militarized public security provision (Flores-Macías and Zarkin 2021).
Despite the prevalence of militarization and its pernicious consequences, we know very little about its causes. In this paper, we focus on police militarization—the process whereby police come to resemble the military in the use of materials, tactics, and norms (Go 2023; Magaloni and Rodriguez 2020; Mummolo 2018; Tiscornia 2024)—as a phenomenon that, although related, is distinct from the use of the military for public security provision. Specifically, we explore the political determinants of police militarization and the limitations of electoral processes as mechanisms for police reform.
Theoretical explanations for the causes of the militarization of public security in democratic contexts mostly treat it as a policy response to the rise in violent criminal organizations (Magaloni and Rodriguez 2020) and crime victimization more generally (Visconti 2020). Other research suggests that increases in militarization are the result of authoritarian policy preferences of right-leaning governments (Holland 2013). Following these arguments, one might expect right-wing governments to increase police militarization, and leftist ones to decrease it and emphasize the respect of human rights in policing. Not surprisingly, advocates of criminal justice and police reform often promote the election of progressive politicians as a strategy to curtail abusive police practices and reform police forces. Yet, research demonstrates that left-leaning governments are no less likely to militarize the police than their counterparts (Davis 2006; Durazo-Herrmann 2017; Muller 2017). Building on theories of issue ownership, we develop and test an explanation of militarization resulting from lack of ownership of security issues. We argue that left-leaning politicians, often perceived as less capable to deal with security threats (Green and Hobolt 2008), are forced to signal their competence in this area by militarizing police more than non-leftist politicians, especially in contexts of high levels of violent crime where security is a central issue.
We focus on the case of Brazilian municipal police forces and their growing importance in public security as our source of evidence. Previous studies have centered on Brazilian state police forces and their entrenched militarized logic. This logic relies on perceptions of citizens as threats (Magaloni and Rodriguez 2020), and it emphasizes the control, and often elimination, of social groups historically conceived as “enemies” of social order: slaves, political dissidents, the poor, and Afro-Brazilians (Machado 2014). These historical roots crystalized in bureaucratic structures, making state police forces difficult to reform (González 2020). However, municipal police bureaucracies are less entrenched, which gives local politicians more leeway to implement policing policies and practices that conform to their and their voters' preferences.
Brazilian leftist political forces have been associated with a rhetoric and policy preferences that emphasize respect for human rights and police reform. During and after democratization, center-left and centrist forces were committed to the defense of human rights as a pillar of Brazilian democracy. While their ideological commitment might create the expectation that leftist parties are associated with lower levels of police militarization at the municipal level, we argue that they will militarize local police more than right-wing or centrist parties to compensate for perceptions of lower competence to deal with security issues, especially when they face high levels of violent crime.
We test our hypotheses using municipal-level data from the Brazilian Institute for Geography and Statistics' (IBGE) Municipal Information Dataset (MUNIC). MUNIC compiles data on whether a municipal police force (Guarda Municipal) exists in a given year, how long it has existed, whether it uses weapons (lethal and non-lethal), and other characteristics of its training and operations. We exploit the availability of time-series cross-sectional data and employ inverse probability of treatment weights (IPTW) to achieve covariate balance between treated and untreated observations while avoiding post-treatment bias (Blackwell 2013). With this pre-processed sample, we test the effect of electing a leftist mayor on the training and functions of the municipal police. After balancing on relevant covariates, our results show that having a leftist mayor increases the degree of militarization of the municipal police. Leftist mayors, in particular mayors of the Workers' Party (PT), are more likely to promote actions that improve police capacity for the use of violence and the violent regulation of social conflict, especially when they face high homicide rates. This behavior is striking since other studies have shown that belonging to a leftist party is a significant predictor of the enactment of progressive social policies and democratic practices at the municipal level (Sugiyama 2008). Historically, voters have associated these policies with PT. However, when PT politicians are forced to engage with a salient issue their party does not own, they signal their commitment to security by militarizing local police forces more than mayors from other parties.
Our study sheds light on the political determinants of militarization processes and the limitations of police reform carried out by elected officials. Confronted by public demands for more security, leftist politicians might militarize police forces, prioritizing electoral success over their programmatic beliefs. Theoretically, we provide an argument for the choice of security policy that incorporates politicians' strategic electoral considerations in contexts of high criminal violence. We also highlight the role of local politics in the definition of security policy, an arena typically left for the national or state level. Empirically, our study of Brazilian municipalities and their policing policies can provide important evidence about the role of partisan ideology and issue ownership on security policies.
In the following section, we define police militarization and examine how it occurs. We then theorize the limits of partisan ideology in shaping leftist office holders' choice to reduce the militarization of police forces. We subsequently discuss Brazil’s legacy of police militarization and municipal governments' increasing involvement in public security through municipal police forces. Next, we present our research design, discuss in more detail our outcome of interest (the level of militarization of municipal police forces), our independent variable of interest (the presence/absence of a leftist mayor), and relevant covariates. Since our initial independent variable might not quite capture “true” leftist parties, in a second set of analyses, we explore the impact of having a PT mayor on the level of police militarization. We conclude with a discussion of the results and offer some remarks about the limits of electing leftist politicians to enact significant police reform.
Police Militarization as a Signal of Competency in Security Policy
Civilian control of security institutions is a prerequisite for the democratic rule of law (Kohn 1997). Concerns about the role of the security apparatus and the rule of law in Latin American democracies date back to the end of authoritarianism (Kohn 1997; Stepan 1988). Debates on the need to create democratic police forces gained traction in the early 1990s in response to generalized increases in crime and violence (Bailey and Dammert 2005), the end of the civil wars in Central America (Call 2007), and the limited development of the rule of law in the region’s democracies (Méndez, O’Donnell, and Pinheiro, 1999). Those discussions took place in the more general context of the end of the Cold War and the international community’s commitment to security sector reforms as mechanisms to increase state capacity and ensure peace and stability (Call and Cousens 2008; Paris 2004).
In Latin America, militarized police forces are a prevalent feature in the security policy landscape (Flores-Macías and Zarkin 2021), even after police reform. Since the 1990s, democratic governments have put in place a plethora of initiatives designed to increase police capacity and effectiveness to implement the prescriptions of democratic policing. Such prescriptions refer to the adherence to norms like the impartial application of the law, as well as standards of professionalization, accountability, and respect for human rights (Bayley 2005, 8). Yet, because capacity and professionalization were treated as technical aspects of policing, many of these reforms might have increased, rather than reduced, militarization (Tiscornia 2024).
Discussions of police militarization have focused on the involvement of the military in internal security (Blair and Weintraub 2023; Espinosa and Rubin 2015). This involvement can happen either because the police is subordinated to the ministry of defense, often under a military general’s command, or because the military oversee patrolling or other domestic security tasks. Other research has emphasized militarization as the incorporation of a military model, as an organizational logic, which can happen with or without participation of the military (Flores-Macías and Zarkin 2021; Magaloni and Rodriguez 2020).
We understand police militarization as an organizational logic that entails the preparation for the use of violence as the first recourse (Tiscornia 2024). Police militarization is often equated with the adoption of military equipment and tactics. This material dimension is one aspect of the process; however, police militarization also involves socialization and specific organizational features and practices (Flores-Macías and Zarkin 2021; Go 2023; Magaloni and Rodriguez 2020).
Socialization happens through the adoption of martial language and military uniforms, as well as by cultivating the belief that there are “enemies” who should be eradicated (Kraska 2001). This belief stands in contrast with the idea of police as agents who prevent and regulate conflict with and between citizens. Whereas the use of force—particularly lethal force—should be the last recourse for police, war and war-like contexts and rhetoric “normalize” the use of violence by public security agents (Wood 2008). In a democracy, increased crime, and politicians' references to the need to “fight a war” against crime, can contribute to normalize the militarization of the security apparatus. Organizational features refer to structural elements of the militarization process, such as whether there is a centralized structure or elite units modeled after military special operations units (Balko 2013). Practices are the specific functions or activities of the police: officers might be involved in intelligence gathering or in counter-insurgency type operations.
Because they are highly visible, several aspects of militarization, such as military uniforms, or heavy weapons, provide politicians with symbolic elements to show they are “tough on crime,” regardless of the effectiveness of these policies to combat crime (Lamb 2018; Tiscornia and Perez Bentancur 2022). By increasing militarization, especially its most visible manifestations, politicians can simultaneously signal that they take voters’ preferences seriously, and that they are competent and can address security policy issues.
The Political Determinants of Police Militarization
It is often assumed that parties’ ideas, ideologies, or programmatic preferences translate into policymaking (Nice 1985; Schmidt 1996). Studies in the United States, for example, have found that the role of ideology in defining policies is more prevalent at the state level than at the local level, particularly regarding economic policy (Potrafke 2018). Many of these insights come from social or economic policy arenas, but we know less about security policy. We take these insights as a starting point to explore the role of ideology when defining security policy.
What explains the choice to militarize the police in democracy? Much like with other policies, the choice to militarize police forces may be intuitively associated with the ideological preferences of elected officials. One could expect, for example, that politicians on the right of the ideological spectrum privilege notions of “law and order” over human rights and therefore tend to create police forces that are tougher on “enemies of the social order.” In contrast, leftist politicians—who usually privilege notions of human rights and due process—tend to support police forces that are embedded in the communities they police and resort to less coercive tactics of conflict resolution. We should therefore expect lower levels of militarization in police forces when leftist politicians are in charge.
While not all leftist parties share the same commitment to human rights and democratic policing or all right-wing parties equally support “law and order” policing policies, case evidence supports this assumption. After the civil war in El Salvador, the administration of the conservative political party ARENA began politicizing the issue of crime and implementing law and order policies that militarized its police (Holland 2013). In Mexico, during the conservative Calderón administration, the government resorted to the military as a tool to quell drug cartel violence (Escalante Gonzalbo 2011). Conversely, in South Africa, the left-leaning ANC’s ascent to power raised the concerns of developing a national crime prevention strategy that would alleviate the structural factors leading to crime and creating a police service respectful of human rights (Rauch 2000). Similarly, in Brazil, many right-wing politicians implemented policing policies and practices that increase militarization, while their left-wing counterparts did not (Caldeira 2000). 2
Despite existing case evidence, the assumption that policing policies can be mapped onto an ideological divide should be further scrutinized. According to theories of issue ownership, voters come to associate certain parties and politicians with competency in specific issues. This association does not necessarily result from good performance by the party or politicians on the issue, but rather from the persistent prioritization of the issue (Egan 2013). 3 If a particular issue is salient in an election, the party (or politician) who has a reputation for competency in this issue will have the upper hand (Petrocik 1996; Petrocik, Benoit, and Hansen, 2003). Although parties and politicians tend to emphasize issues that they “own,” (Egan 2013; Fagan 2021), they are often compelled to engage with issues they do not, especially when they are very salient to voters (Damore 2004). In this scenario, we argue, politicians who do not own the issue will devote more resources to signal their ability, as compared to parties and politicians who already own the issue and choose to invest limited resources elsewhere. This is the case when there are high levels of crime and security becomes a central issue for voters.
Conservative parties and politicians prioritize security issues and tend to be associated with the capability to deal with crime and other threats against public security (Egan 2013; Green and Hobolt 2008). In contexts of high levels of crime, left and right leaning voters will push government officials to enact visible policies to deal with crime (Laterzo 2024). In Latin America, public opinion data shows that an overwhelming majority of citizens, regardless of income level or ideology, support the militarization of public security (Blair and Weintraub 2023; González 2020; Laterzo 2024; Tiscornia and Perez Bentancur 2022). Under these conditions, leftist politicians, who are at a disadvantage when security becomes a central concern for voters, must signal to voters that they, too, can deal with public order by over-committing to a “law and order” agenda, which usually translates into implementing harsher policies.
Militarizing police forces by providing more powerful weapons, employing a war-like rhetoric when referring to crime, or expanding the spectrum of situations in which the use of force is allowed, are visible mechanisms to showcase this commitment. For example, evidence from the United States shows that Democratic governors who win close elections pursue more punitive policies than their Republican counterparts: they spend more on the prison system and incarcerate more people (Gunderson 2022). Evidence from Mexico also suggests that voters perceive militarized police as more effective (Flores-Macías and Zarkin 2022). Leftist politicians will choose to invest in police militarization over ideological commitments that emphasize less coercive aspects of conflict settlement. In doing so, they signal competency over security policy to voters. Yet, pursuing militarizing policies does not guarantee electoral victory, as public opinion might see it as “too little, too late,” and political opposition might frame it as ineffective or insincere. As a result, each time they win, leftist mayors might further entrench militarization in hopes of retaining power.
Not all leftist governments have pursued police reforms to reduce militarization, and those that have were not always successful. Brazil’s PT, whose commitment to the respect of human rights and police reform was a programmatic cornerstone (Fundacão Perseu Abramo 2017), has not consistently pressed for police reform when governing states and municipalities, particularly when political alliances with right-wing parties were electorally advantageous (Alves 2020). Research on the behavior of Brazilian political parties shows this pattern is consistent across policy arenas: when politicians from programmatic parties face electoral pressures, they will take nonprogrammatic actions once in office to secure reelection (Johannessen 2020).
While expressing a commitment to reduce militarization and curtail human rights violations, under conditions of high violence, strategic choices may overpower these ideological preferences. Because citizens have little information about the actions that politicians pursue at the local level, they tend to rely on those actions that are highly visible (Johannessen 2020), such as providing local police with larger weapons and military uniforms. When the electorate does not consistently support police reform (González 2020) and overwhelmingly supports “iron-fist” type policing regardless of ideology or income (Rosen and Cutrona 2021), there is a strong incentive for leftist politicians to choose police militarization over ideological attachments.
Police Militarization in Brazilian Municipalities
In Brazilian federalism, public security and policing policies are predominantly decided at the state level. Historically, Brazilian police forces have been charged with the repression of social groups that were deemed dangerous for the status quo and fulfilled this function with little constraints (Holloway 1993; Jakubs 1977). Since Brazil’s transition to democracy in 1985, despite several reform attempts (Durazo-Herrmann 2017; González 2020; Magaloni and Rodriguez 2020), only minor changes have taken place with little impact (Brinks 2008).
Modifying the structure of Brazilian policing requires changes in the Federal Constitution and state legislation. The Constitution predetermines a high level of militarization of state police in their role as auxiliary and reserve forces of the Armed Forces (Soares 2019). Moreover, military police forces have longstanding bureaucratic structures with corporativist interests with little to gain from police reforms. In addition, Brazilian public opinion has not pushed for reform in a consistent manner—in many cases, it has even supported punitive policing—resulting in Brazilian state police forces experiencing more continuity than change since 1985 (González 2020).
Since the 1980s, Brazil has undergone considerable decentralizing reforms, giving states and municipalities more fiscal resources and policy competencies (Falleti 2010). Reforms allowed municipal authorities to implement innovative policies, such as conditional cash transfers, which later inspired federal programs (Sugiyama 2008). This increase in policy competencies is also reflected in public security matters. In the late 1990s and early 2000s, municipal authorities started to make increasing use of their constitutional prerogative to create and empower guardas municipais, or municipal guards (Figure A1, Appendix A). These police forces are the most visible—and arguably, the most important—component of municipal activity in public security (Miranda, Freire, and Paes, 2008).
Decision-making regulating the attributions of guardas lies primarily with the mayor, and it is done through executive action. 4 Although local councils influence the decision to create guardas, executives are in charge of regulating their structure, organization, and practices. Thus, a change in local government can shift the orientation of a guarda to a more or less militarized perspective (Martins 2018). The fact that mayors have so much discretion creates a perfect context to study the impact of partisan ideology on policing policies at the local level.
Number of Municipalities With Guardas and Their Size.
Municipal guards have also shifted their focus. Originally, most guardas were tasked with protecting public property and community outreach. In recent years, they have become more involved in “traditional” police tasks, such as ostensive policing and responding to criminal incidents (De Oliveira Junior and Alencar 2016). Recent federal legislation (Law 13.022 of 2014) regulating the organization and tasks of municipal police has codified this expansive view of guardas' tasks by giving them preventive policing roles, enabling them to support state police forces and to apprehend suspects in flagrante, accepting their use of force, and authorizing their use of weapons.
Over time, guardas have incorporated more weaponry. Law 10.826 of 2003, also known as “the Disarmament Statute,” allowed municipal guards in municipalities over 50,000 residents to carry firearms. A 2018 ruling of Brazil’s Supreme Court extended municipal forces' right to bear firearms to all municipalities, regardless of population size. Table 1 shows the proportion of guardas that acquired weapons (lethal and non-lethal) over time. Whereas in earlier years about 85% of the municipal police forces did not have weapons, by 2014, the distribution is roughly 60–40. 6
Taken together, these changes suggest that many guardas are increasingly militarized. Although these changes could be reflective of a broadening of functions in the municipal police, considerable variation in the type of guardas suggests that changes go beyond the mere incorporation of duties. Municipal leaders tend to adhere to notions derived from the template of state military police forces about what municipal police should look like and how they should behave. Consequently, some guardas model their uniforms, internal structures, and leadership after the state police (De Oliveira Junior and Alencar 2016; Martins 2018). In fact, according to MUNIC data, more than 30% of the guardas in each of our mayoral periods of study were led by a military police commander. By increasing the number of officers and equipping them with firearms, the goal is to complement rather than substitute state police, as a response to demands from citizens, who expect guardas to act like the military police (Martins 2018).
Assessing the Relationship Between Ideology and Municipal Police Militarization
We leverage the spatial and temporal variation in Brazil’s municipalities to examine if partisan ideology is a significant factor shaping municipal guards' levels of militarization. Concretely, we evaluate if having a leftist mayor is associated with a reduction in the level of militarization and if this effect holds even when the municipality faces high homicide rates. We use IBGE’s MUNIC dataset, available every 2 years since 2001 for municipalities across Brazil’s 26 states. More complete data, however, exist from 2006 onward. Given the large number of observations for each country-year at the subnational level, we can exploit variation across time, sections, or both.
The universe of analysis consists of the 1382 Brazilian municipalities that had a municipal guard at some point in or before 2014, over four time periods between 2001 and 2016. The IBGE collected information about guardas from municipal governments in the years 2001, 2002, 2004, 2006, 2009, 2012, and 2014. We identified four mayoral periods between 2000 and 2016. Mayoral elections in Brazil take place in October every 4 years, and the elected mayor’s tenure begins on January 1st of the year immediately following the election and lasts until December 31st of the subsequent election year. For our analysis, we select one survey per mayoral period: 2004 (for the 2001–2004 Mayor), 2006 (for the 2005–2008 Mayor), 2012 (for the 2009–2012 Mayor), and 2014 (for the 2013–2016 Mayor). We take each survey as representative of the state of the municipal police during the 4 years. When multiple surveys were available for a mayoral period, we selected the survey year that included the most information and was closest to the end of a mayoral period to account for the time it takes a mayor to implement policies. 7 For our analysis, we average the covariates over each four-year mayoral period.
Research Design
We assess the relationship between the presence of a left-leaning mayor and the extent of militarization in the municipal police, using time-series cross-sectional data between 2000 and 2016, averaged over 4 mayoral periods. We are interested in explaining variation in the extent of militarization of local police, given that there is one. Although it may be possible that the factors leading to the creation of a guarda also drive militarization, our research design allows us to ensure there is balance in terms of those likely drivers between treated and control groups. 8
The dynamic nature of our data—repeated treatments over time that do not occur at the same time for all municipalities, nor at the same rate—makes the likelihood of post-treatment bias high. To take advantage of the panel structure of our data while avoiding post-treatment bias, we propose employing a weighting approach which can be used to estimate effects in dynamic settings. 9
Following Ladam, Harden, and Windet (2018), we incorporate inverse probability of treatment weights (IPTW) to balance treated and untreated observations (Blackwell 2013). This method, routinely employed in biostatistics, is increasingly used in political science (Blackwell 2013; Ladam, Harden, and Windet, 2018). We use IPTW to address confounding by reweighting the data. In our case, confounding could occur, for example, if higher GDP per capita is positively correlated with electing a leftist mayor. Confounding can be eliminated if electing a leftist mayor is equally likely in richer and poorer municipalities, which can be achieved by reweighting the data. Furthermore, leftist mayors can be modeled without controlling for GDP per capita directly, thus avoiding post-treatment bias, which occurs if part of the treatment operates through the control variable (Blackwell 2013; Ladam, Harden, and Windet, 2018). IPTW addresses issues resulting from known confounders. If there is an unknown confounder that is not considered in the analysis, omitted variable bias can still be present.
Using IPTW requires a two-step process. First, we estimate weights and then use them to model our outcome of interest. Because the correct weights are not known, we need to estimate them from the data. The weight for a given observation can be understood as the inverse of the probability of treatment. To avoid high variability in the computed weights, we follow Blackwell (2013) and employ stabilized weights. To create the weights, we model treatment status—having a left-wing mayor in a given period—as a function of a set of covariates using a logistic regression. We include regional fixed effects to account for different baseline probabilities of getting a left-wing mayor. Then we include treatment history. We can think of treatment history as the cumulative effect of having had left-wing mayors over the time periods in our analysis—there can be multiple combinations of this: a municipality may have had a left-wing mayor once, twice, up to four times, or never; treatment can occur consecutively, or alternately. We then include a set of the time-varying covariates that can be associated with the extent of militarization of the police. As Figures B1, B2, and B3 in Appendix B show, by computing weights we greatly improve balance in our sample.
In a second step, we include these weights to model our outcome of interest—the extent of militarization of the police—excluding the time-varying control variables, as they were included in the weighting process (Blackwell 2013; Ladam, Harden, and Windet, 2018). This way, we can achieve balance in the re-weighted data while avoiding post-treatment bias. Following the recommendations in Blackwell (2013) and Ladam, Harden, and Windet (2018), we bootstrap the whole process by resampling municipalities to compute uncertainty measures. Due to space constraints, we provide a more parsimonious explanation of the IPTW method in Appendix B.
By using IPTW, we can take advantage of the temporal nature of our data when estimating our effects of interest. We estimate the cumulative average treatment history effect (ATHE) for theoretical reasons: having mayors with the same political leaning repeatedly (not necessarily consecutively) may affect the ease of making changes to the police. In addition, it is possible that voters see leftist mayors' efforts to militarize as “too little, too late,” which can encourage mayors to continue to militarize the police when they are in office, producing a cumulative effect of having a leftist mayor. Including the ATHE is a mechanism recommended to account for a municipality’s full treatment history on an outcome (Blackwell and Glynn 2018), which allows us to incorporate the possibility that the effect of having a left-wing mayor could occur contemporaneously, or not. We also estimate the contemporaneous treatment effect (CET); however, as Ladam, Harden, and Windet (2018:7) point out, this is a more restrictive estimand: because it ignores treatment history, it assumes that the entire effect of treatment is contemporaneous. We present results with bootstrapped standard errors, with bootstrapping at the municipality level.
Variables and Data
Outcome of Interest: Militarization of Local Police
We compute our dependent variable, the militarization of local police, using the MUNIC dataset. IBGE collects census type data on numerous indicators of municipal governments' capacity and performance based on responses to a standardized questionnaire. For the existence of a guard, we use a dummy variable that is equal to 1 for the municipalities where there is a municipal guard in a given mayoral-year. The map in Figure A2 in Appendix A presents the contrast between the municipalities with and without municipal police for the four mayoral periods in our analysis (2000–2012) providing information on the time-series cross-sectional variation in our data.
We are interested in a set of variables that account for how militarized the municipal police are. The data contain dummy variables for several aspects of municipal policing. Unfortunately, many of these variables are only available for a few time periods. We selected the ones for which data availability extended furthest back in time. To analyze the extent of militarization, we constructed an additive index composed of four indicators capturing material, normative, and organizational dimensions of the concept, to account for changes in police capacity for violence: (1) whether the municipal police carry weapons, (2) assist the military police in their policing functions, (3) patrol, (4) engage in other policing activities beyond safeguarding buildings or public spaces. In Appendix A, we provide a more extensive discussion of the different indicators. We tested the internal consistency of our scale using the alpha coefficient, which measures the reliability of a given scale. The results suggest that our scale items are strongly correlated. 10 The temporal distribution of the index is presented in Figure A12 in Appendix A. The histograms in Figure A12 represent the variation in militarization across municipalities. Figure A13 in Appendix A displays maps with the distribution of militarization across Brazil.
Independent Variable of Interest: Leftist Mayor
Our main independent variable of interest is the mayor’s ideology, proxied by the ideology of the party that wins the mayoral election. Based on Power and Zucco (2009), we classify Brazilian political parties along the traditional left-right spectrum. For those parties that were not identified by Power and Zucco, we use the classification developed by Baker and Greene (2011). We create a dummy variable that identifies leftist parties; Figure A11 in Appendix A presents information on its spatial distribution.
Covariates
The IPTW approach assumes that assignment to treatment is independent of the potential outcomes conditional on a set of covariates (Blackwell and Glynn 2018), and that treated and untreated observations are comparable with respect to those covariates (Ladam, Harden, and Windet, 2018). To seek balance, our analysis includes a group of time-varying covariates that could affect assignment to treatment and our outcome of interest. The selection of covariates was informed by the literature on comparative policing and police reform, and reflects explanations commonly associated with the militarization of police forces: state capacity, arguments about development, and increases in violent crime. All covariates are averaged over the corresponding mayoral period.
First, we include a measure of state capacity from data from the Brazilian treasury’s FINBRA system. It represents the proportion of municipal revenue that comes directly from local taxes and measures the extractive capacity of the local state, that is, the amount of resources it can derive from the local population. We also include a covariate measuring the royalties received from oil production with data provided by the Brazilian National Agency of Petroleum and Gas (ANP), as well as the size of the bureaucracy per 1000 people using data from IBGE’s MUNIC. We add an indicator of intermediation (IBGE 2017) to measure the extent to which a municipality is connected to other municipalities via land or water to account for the possibility that very isolated municipalities may be driving either the outcome or assignment to treatment.
Other arguments link the nature of police forces to the extent of order in a municipality. Municipalities where disorder is high as a result of phenomena such as high criminal violence, low human development, or population increases, should be more likely to have a police force of a more militarized nature. We account for this possibility by including the municipal-level homicide rate (from IPEA and health system SUS), human development index (from IPEA and UNDP), GDP per capita, proportion of the municipal GDP resulting from agricultural production, industry, and the public sector, as well as population size (from IBGE).
Finally, we include a series of covariates to account for institutional factors that might limit a mayor’s ability to enact their agenda. The greater the fragmentation of the party system, the more likely it is that mayors will have to form coalitions in the city council. As a result, they may have reduced leverage when attempting to reform municipal guards while maintaining their policy positions. To measure party system fragmentation, we calculate the effective number of parties in municipal council elections using data from the Federal Electoral Court (TSE). The local party system might also be poorly institutionalized, affecting the predictability of partisan competition and the ability of local governments to enact their policy priorities. We measure party system institutionalization using the Pedersen Index of electoral volatility in municipal council elections (Pedersen 1983). Turnover from one party to another can also affect the ability of a mayor to enact his agenda, since a new party in government will require more time to staff municipal administration and build a new coalition to support its agenda. Thus, we include a dummy variable indicating if there was a change in the party governing the municipality. Moreover, the greater the importance of the left in city council elections, the more likely it is that a leftist mayor will be elected who will implement a policy agenda that is aligned with its preferences. We therefore include the total percentage of votes for leftist parties in local legislative elections. Finally, to account for unobservables at the state level, we include state fixed effects. 11 We provide summary statistics in Table B1 in the online Appendix B.
Results
Figure 1 graphs the estimated IPTW effects of having a left-leaning mayor on police militarization. We consider the ATHE to capture municipalities' histories of previous treatment, that is, the possible combinations of left-leaning mayors over time. The ATHE estimates are positive and statistically significant at the 0.05 level. This means that having mayors who belong to left-leaning parties is associated with more militarized municipal police. Although the size of the coefficient is relatively small, it is important to note that our dependent variable—the militarization index—ranges from 0 to 4. Thus, a quarter-of-a-point change in ideology can represent a sizable change in a municipality’s militarization index. Although the CET’s direction is consistent with the ATHE, it is smaller and not statistically significant, underscoring the relevance of the treatment histories and the likelihood that not all the effect of treatment is contemporaneous. Regression analysis: Any leftist mayor.
To further explore these findings, we investigate what mayors do when they face a crisis of high violence. Following theories of issue ownership, it is possible that leftist mayors are perceived as less competent in dealing with public order. Facing a security crisis that stems from high crime and the saliency of this issue, leftist mayors could increase the militarization of police forces to show that they can govern effectively. To test this argument, we split the sample based on how high the murder rate is. According to the World Health Organization, a country faces a homicide “epidemic” when there are over 10 intentional homicides per 100,000 people in a given year. Since we are using municipality data, we preserve the ratio and group municipalities with more than 1 homicide per 10,000 people as cases with high murder rates.
12
Figure 2 graphs the results. While in municipalities with low murder rates, the cumulative effect of having a leftist mayor leads to an increase in the militarization of the police, the effect is not statistically significant in high violence municipalities. Although this result is surprising in light of our theory, one possible explanation for this finding is that in high violence contexts it is difficult to find an effect because our independent variable does not adequately capture “true leftist” parties that have developed a partisan identity around human rights issues and is negatively associated with competency and commitment to handle security issues. Regression analyses: Split sample any leftist mayor.
A Programmatic Leftist Party’s Electoral Constraints in High Insecurity Environments
Considering our previous findings, there is a strong possibility that our dummy variable for “leftist” is not quite isolating “true” leftist parties. At the local level in Brazil, non-leftist politicians might join a leftist political party for strategic reasons (for example, access to the ballot or to access resources from state or federal governments) without identifying with issue positions on human rights and policing. To examine if this is the case, we test if electing a PT mayor exerts an effect on the level of police militarization.
Founded in 1979, the PT developed out of a confluence of grassroots social movements, unions, and leftist intellectuals. The PT had a clear leftist agenda, strong ties to progressive social movements, a strong commitment to the defense of human rights, and the social, economic, and political inclusion of marginalized populations. Moreover, unlike most Brazilian political parties, it developed a partisan organization that fostered members' participation in its activities, expecting and enforcing party discipline. Although the PT has exhibited more “office-seeking” behavior in the past decades, the high costs of membership prevented ambitious but “disloyal” politicians from joining the party (Hunter 2010). Furthermore, the PT historically emphasized demilitarization and human rights as centerpieces of its programmatic agenda (Fundação Perseu Abramo 2017). If the PT’s ideology is relevant for its implementation of policing policies, we should expect PT mayors to have less militarized local police forces.
However, while in office, PT has often provided visible public goods as way to retain support from voters (Johannessen 2020). A case in point is the election of Rui Costa, a PT candidate who won the 2014 gubernatorial election in the state of Bahia advocating strong punitive measures (Laterzo 2024). Another example is the current administration of President Lula. Perceived to be less competent to manage public security than his right-wing predecessor (Veja 2024), he has supported further militarizing public security policies (Paes de Souza and Feltran 2023).
Given the PT’s historic lack of ownership of security issues and the prevalence of high levels of insecurity in many Brazilian cities, we expect to observe more militarization from PT mayors. If PT mayors have taken office in municipalities where the party’s programmatic orientation is not in line with citizens' preferences on security and policing, they will pursue policies that further militarize guards to signal to voters their commitment to the issue and remain attractive to them, particularly when levels of violence have made security an important topic for Brazilian public opinion. 13
Using a new independent variable—a dummy variable indicating the presence (1) or absence (0) of a PT mayor—we re-weighted our sample and re-estimated our models. Figures 3 and 4 present the results. The results in Figure 3 show patterns consistent with our first set of models: having a mayor from the Workers' Party is associated with higher levels of militarization. The coefficients are statistically significant at the 0.05 level and are of larger magnitude than those in the previous analyses. When considering the PT only, both the ATHE and the CET are statistically significant, suggesting that there is a cumulative effect on militarization, but also a contemporaneous one. Regression analyses: PT mayor.
Figure 4 presents results for our split sample of municipalities with high and low homicides. The results here lend support to the argument that when facing a security crisis, programmatic left-wing parties will militarize the police. As Figure 4 shows, in municipalities facing epidemic levels of homicidal violence (i.e., more than 1 homicide per 10,000 residents), mayors from the PT will militarize their police more than other mayors (both the CET and the ATHE are significant). Importantly, while the estimated effect is larger in municipalities with high homicide rates, PT mayors also tend to militarize the police in municipalities with low levels of homicidal violence.
14
Regression analyses: Split sample PT mayor.
Because the IPTW approach assumes no confounding after weighting, in Appendix C, we conduct sensitivity analysis to assess the extent to which our estimates would change if there were a confounder we had not accounted for. Based on this analysis, it is unlikely that our results are driven by hidden bias.
Conclusion
The militarization of police forces has frequently been associated with higher levels of police violence (Uildriks 2010), especially against marginalized populations (Ross 2015). Despite growing scholarly interest in police militarization, few quantitative studies seek to understand the causes of police militarization in the developed or the developing world. In this article, we explore the connection between partisan ideology, issue ownership, and police militarization. It is often assumed that in democracies, leftist/progressive administrations are more likely to emphasize human rights in their public security strategies. They should be less likely to militarize police forces. However, in situations of epidemic criminal violence, leftist administrations might want to counter this perception and militarize police forces more than their centrist and conservative counterparts would, to signal their competence in public order affairs. We use the case of Brazilian municipal police forces (Guardas Municipais) to explore the association between leftist mayors and the militarization of local police.
Our results indicate that leftist mayors tend to militarize local police forces more than mayors from other political parties. These results are robust even when we use a more demanding measure for a leftist municipal administration and contrast municipalities governed by Worker’s Party mayors to municipalities governed by other political forces. While PT mayors tended to enact progressive social and participatory policies at the municipal level in the past (Sugiyama 2008) and have also campaigned on programmatic issues close to the preferences of their base, they have then governed by prioritizing issues that can run counter to their party platforms (Johannessen 2020). In the case of policing, PT mayors on average seem to be more responsive to the punitive policing preferences of voters than to the ideological commitments of their party.
The findings in this article raise important questions for the study of the causes of police reform. Existing research suggests that one of the obstacles for police reform is the existence of powerful police bureaucracies with entrenched interests and hard to change socialization patterns and practices. We analyzed cases where reformist politicians could have reduced police militarization, given the absence of entrenched police bureaucracies. By focusing on more malleable bureaucracies, we show that even when there is ample room to maneuver, leftist mayors choose to militarize the police. This dynamic suggests that mayors' electoral considerations are a powerful driver of police militarization, beyond policing practices.
Our findings also have practical implications. If politicians respond to citizens' demand for “tough on crime” policies by militarizing local police, even when they have the tools to do otherwise, the prospects for police reform are bleak. Leftist governments are under strong pressure to show their competency in security policy and have few electoral incentives to promote policing practices that are not predicated on the use of force.
Further research is required to disentangle whether partisan ideology impacts specific dimensions of militarization. It is possible that leftist administrations militarize in one dimension—for example, the acquisition of more lethal weapons—while also investing in human rights training more than centrist and conservative parties. Leftist parties might be attempting to “square the circle:” increasing the capacity to inflict physical harm to enhance their public security credentials while investing more in human rights training to appease their traditional constituencies.
The article also contributes to our understanding of the role of local level politics in decisions about specific policies. Particularly when it comes to security policy, local governments have great autonomy in implementing decisions made at other levels of government. Focusing on the Brazilian case, the results suggest that electing leftist or politically progressive local governments as a strategy to transform militarized police forces into more community and human rights-oriented organizations might not bring about the desired results. As observed in other contexts, leftist politicians appear to be—at the very least—complicit in the growing punitiveness of policing policies.
Supplemental Material
Supplemental Material - Mall Cop or Robocop? The Political Determinants of Police Militarization in Brazil
Supplemental Material for Mall Cop or Robocop? The Political Determinants of Police Militarization in Brazil by Juan Albarracín, and Lucía Tiscornia in Political Research Quarterly.
Footnotes
Acknowledgements
We are grateful to Abby Cordova, Jeff Harden, and Ana Petrova for their invaluable feedback on earlier iterations of the manuscript. A version of this article was presented at the 2018 Annual Meetings of the Latin American Studies Association and the Western Political Science Association. We also thank the editors and reviewers at PRQ for their insightful comments.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
