The Electoral Support for Law and Order Candidates: Violence and Local State Capacities^*

Abstract

What motivates voters to support candidates with backgrounds in public security for Brazil’s National Congress, the so-called Law & Order (L&O) politicians? The 2018 elections witnessed a surge in parliamentarians from police and military backgrounds and the election of a president closely aligned with this agenda. We explore whether this trend correlates with municipal-level violence indicators and the presence of municipal capacity in the security sector, employing various regression models that account for socioeconomic, demographic, political, and geographic factors. Our findings from the additive model reveal a negative association between the level of violence and votes for L&O candidates. This suggests a potential disconnection between fear, victimization, and support for repressive policies, while also indicating that local capacity in the security sector tend to enhance L&O candidates’ performance. However, our model incorporating the interaction between both variables uncovers that violence exacerbates the negative impact on L&O voting when the city has capacity in the security sector. This may reflect a municipal-level preference for preventive measures, particularly in more violent areas, which contrasts with the generally repressive stance of L&O candidates.

Keywords
Elections; law and order candidates; violence; public security; state capacity

The 2018 elections marked a pivotal moment in Brazil’s political history. Jair Bolsonaro, then affiliated with the Social Liberal Party (PSL), a retired Army captain and former congressman who championed the causes of the Armed Forces and the police, was elected President on a radical right-wing platform. His main agendas included public security and fighting corruption. Additionally, this election saw a wave of candidates with strong ties to security forces and tough-on-crime rhetoric elected to Congress – referred to in the literature as Law and Order (L&O) or ‘mano dura’ candidates. This phenomenon reinforces a trend observed by scholars over the past decade (LIMA, 2020), highlighting the increasing political engagement of public security professionals, particularly from the civil police, federal and military police, federal highway police, and members of the Armed Forces.

Building on the trend from previous elections, the 2022 election saw a record number of candidates from the police force, representing a 27% increase compared to 2018. A total of 1,866 police officers ran for office, including 807 military police officers. This surge also underscores the strong preference of these candidates for right-wing parties: in 2018, 89.9% were affiliated with such parties, a number that rose to 94.7% in 2022 (COSTA and MARQUES, 2022).

Several studies have sought to map these parliamentarians, tracing their social and organizational connections as well as their agendas, which encompass advocating for corporatist issues, conservative values, and law and order (BERLATTO et al., 2016; MUNIZ et al., 2018). However, there has been limited headway in grasping the determinants or foundations of this electoral phenomenon. What explains the upsurge of elected politicians from the security sector in 2018?

One potential explanation involves the increased visibility garnered by police investigators due to major corruption investigations conducted by the Federal Police (FP) in the years leading up to 2018, particularly within the scope of the Car-Wash Operation (Operação Lava-Jato in Portuguese), coupled with a reaction to the pervasive level of violence in society (MACAULAY, 2019). Interestingly, both international literature and studies focused on the Brazilian context have found divergent results concerning the relationship between crime rates and voting for Law and Order (L&O) candidates. While recent research on Brazil or studies addressing the Brazilian context, such as Luz and Sant'anna (2019), suggest that higher homicide rates correlate with an increase in electoral support for security sector candidates. And Ventura et al. (2023) indicate a positive relationship between victimization and support for punitive measures, other works have actually found a negative correlation (VENTURA, 2022; NOVAES, 2023¹).

The literature offers an explanation for this latter observation, which involves distinguishing between victimization, fear of violence, and support for tough-on-crime policies. Candidates from the security sector may garner more support for their agenda from voters who experience less victimization, but demand punitive responses due to their perception of the causes and nature of violence. Conversely, while voters from more vulnerable demographics suffer more from violence (including police violence), they tend to reject solutions reliant on the state’s repressive apparatus (ZILLI and COUTO, 2017). A similar phenomenon occurs on a comparative scale. According to Ivkovic (2008), individuals with more frequent interactions with the police tend to have less trust in this institution, particularly in countries where it exhibits low levels of transparency.

In this article, we advance on at least two fronts. Firstly, we reassess the connection between homicide rates and support for ‘mano dura’ candidates, with a particular focus on the 2018 elections – a timeframe not yet explored in existing studies and characterized by the aforementioned context. Secondly, we integrate our analysis with an institutional explanation. Since the latter part of the 2000s, municipalities have assumed a more proactive stance in tackling crime, albeit with a predominant emphasis on prevention rather than repression, especially in more violent settings. This shift reflects several federally-driven initiatives, with the primary landmark being the PRONASCI program (MADEIRA and RODRIGUES, 2015). Thus, drawing on studies emphasizing the importance of mayoral strategies for the election of federal legislators, we explored whether the presence of municipal-level capacity in the security sector influences voting patterns for L&O candidates. Furthermore, we posit the hypothesis that this variable alters the relationship between violence and voting patterns, insofar as it introduces alternative approaches to combating crime compared to the typical proposals of security sector candidates. Testing this hypothesis demands the application of a model incorporating variable interaction. Thus, we may summarize the two hypotheses presented in this article: 01. the higher the level of violence in the municipality, the lower the proportion of votes for security sector candidates; 02. this negative effect is magnified in municipalities with their own state capacity in the security sector.

We developed several empirical regression models, both additive and interactive, applied at the municipal level to test these hypotheses on the electoral outcomes for the Lower House in 2018. We established criteria to identify candidates associated with the public security agenda based on their occupation and ballot name. Additionally, we identified variables representing municipal capacity, including the type of municipal security secretariat as well as the presence of five municipal public security structures, ranging from the existence of a Public Security Plan and Fund, Council and Community Council for Public Security, to the inclusion of municipal guards. All models include control variables related to socioeconomic, geographic, demographic, and political characteristics of the municipalities.

The results suggest that higher levels of violence did indeed result in fewer votes for security sector candidates in 2018. While the additive model indicates that local state capacities increase support for these candidates, we contribute with specialized knowledge by highlighting that the interactive effect between these variables is significant and has a negative direction: the decrease in votes resulting from increased violence is more pronounced in municipalities with their own state capacity in the security sector.

From a theoretical perspective, on the one hand our results reinforce existing studies that propose a potential dissociation between the preferences of populations exposed to extreme levels of violence and the strength of electoral platforms – and by extension, public policies (NOVAES, 2023) – related to public security. On the other hand, we contribute to the debate by underscoring that this dissociation may be more accentuated in contexts with a heightened emphasis on violence prevention (i.e., when municipalities implement local security structures), thus in stark contrast with the reactive public policies typically championed by L&O candidates.

The present article is structured as follows. We begin by contextualizing the political influence of the military and the police in Brazil in the next section. Following this, we explore the literature on violence and voting, laying the groundwork for our theoretical framework. We then describe the data and methods utilized, followed by the presentation and analysis of our findings. Lastly, in the concluding section, we summarize our discussion, highlight the advancements and limitations of our study, and suggest directions for future research.

Militarization of politics

The militarization of police forces, along with the increasing involvement of the armed forces in law enforcement, are defining features of the Brazilian state. These developments are the result of a conservative democratic transition, which laid the structural groundwork for the intensification of this dual phenomenon in present-day Brazil. According to Souza and Serra (2020), "From the perspective of practices and discourses, security became the linchpin for propelling militarization, alongside the strengthening of conservative agendas in public security” (SOUZA and SERRA, 2020, p. 207). Additionally, Macaulay (2019) argues that the initial push for police forces to engage in politics can be traced back to the strikes of the 1990s and the formation of their unions, which began organizing collectively and electing members to municipal and state legislatures before eventually gaining representation in the National Congress.

The election of Bolsonaro to the Presidency in 2018 represents the culmination of this movement that has unfolded in Brazil over the past few decades, embodying aspirations for law and order, policing, and public security (MACAULAY, 2019). The preceding elections, in 2014, had already indicated a shift in Brazilian politics with the rise of the right-wing forces, resulting in a legislative election that purportedly elected "the most conservative Congress in the post-1964 period” (QUADROS and MADEIRA, 2018, p, p. 492). According to Quadros and Madeira (2018), this marked the end of an old right, whose historical association with the civil-military dictatorship had previously deterred openly right-wing politicians.

Explanations for the resurgence of overt right-wing politicians center around a generational shift – which contributed to diminishing the sense of shame – and the mobilization of a moral agenda capitalizing on the political crisis and the investigations of the Car-Wash Operation (QUADROS and MADEIRA, 2018). However, it is primarily within the context of the enduring crisis in public security and the alarming levels of violence and crime in the country that the literature has addressed the increased participation of candidates and elected officials from the state’s repressive forces (BERLATTO et al., 2016).

If we trace the evolution in the participation of these candidates, we observe a gradual increase until 2006, coinciding with the initial displays of strength by the PCC (Primeiro Comando da Capital, a Brazilian organized crime syndicate) (ADORNO and SALLA, 2007). This was followed by a plateau until 2014, with a similar number of candidates compared to 1998 (BERLATTO et al., 2016). However, during this period, we already witness a significant rise in the number of elected candidates. According to Faganello (2015), the 2014 election saw a 25% increase in former police officers elected as state or federal parliamentarians. Yet, the 2018 election marked a milestone, with members from the security sector achieving their largest presence in the legislature since the return to democracy (MACAULAY, 2019). Subsequent elections in 2020 and 2022 reveal the ongoing expansion of this trend, both in the number of candidates and the number of elected officials from the police forces (COSTA and MARQUES, 2022; LIMA, 2020).

Embedded within the conservative wave and positioned on the right side of the ideological spectrum, these parliamentarians, despite their diverse stances, discourses, goals, and audiences, have rallied around security-related agendas. These include advocating for the reduction of the age of criminal responsibility, abolishing pre-trial detention hearings, outsourcing core activities, lobbying for the firearms industry against the Disarmament Act, as well as resisting projects that seek to criminalize homophobia and include gender discussions in school curricula. This collective is informally referred to as the “Bullet Caucus” (FAGANELLO, 2015, p. 145).

An examination of these candidates’ profiles over the past two decades reveals their alignment with the right-wing. Initially affiliated with major conservative parties, they later gravitated towards smaller right-wing parties, representing authoritarian trends within society and advocating for the expansion of police forces and the escalation of violent crackdowns on crime (BERLATTO et al., 2016).

Former military police officers and military personnel were precisely the primary support groups for Bolsonaro in his election. However, the ‘Bullet Caucus’ extends beyond them, encompassing individuals who advocate for a tougher stance on crime and punishment (MACAULAY, 2019). Operating as a lobbying group for conservative agendas, the ‘Bullet Caucus’ represents segments of the population that demonstrate “intransigence towards criminals and hold the Armed Forces and similar institutions in high regard” (QUADROS and MADEIRA, 2018, p. 505).

Coming from the state’s repressive forces, they capitalize on this expertise to orchestrate their campaigns, promising more punitive forms of social control. However, they seldom exhibit party loyalty; instead,

…the top vote-getters from the police are thus the result of individualism or personalism that provides the foundation and structure for this entire system. This personalism, which produces the figure of the 'sheriff' of public security, the individual who takes it upon himself or is entrusted to authoritatively resolve public insecurity, aligns with voters' predisposition to identify the causes of crime either in the wickedness of people or in the lack of policing in cities (BERLATTO et al., 2016, pp. 89-90).

We must also consider the emergence of other groups with direct ties to public security, such as the Federal Police Front, whose prominence in recent decades, owing to the enhancement of their role during the PT presidential administrations and their focus on combating corruption, garnered significant popular support through the Car-Wash Operation (FAGUNDES and MADEIRA, 2021). This was a major phenomenon in the 2018 elections, but subsequent developments ultimately led to the downfall of the operation. In terms of agenda, all these groups have been and are likely to continue acting as promoters of law and order and as veto players on human rights issues (MACAULAY, 2019).

Theoretical approach

While violence remains a relatively underexplored variable in the literature on political behavior, studies examining the relationship between voting and violence can be categorized into three main groups. The first group focuses on estimating the effects of insurgent violence, civil wars, and terrorist attacks on elections. The second group explores the phenomenon of electoral violence itself. The third group, which forms the basis of this article, aims to understand the effects of violence on electoral participation and voting preferences for ‘mano dura’ politicians, which includes assessing how crime victimization influences support for candidates with backgrounds in law enforcement agencies, signaling a trend toward the militarization of politics. This group also encompasses studies on partisan behavior concerning violence and public security, shedding light on policy choices and expenditure in this domain, as well as geographical patterns of homicides (INGRAM and COSTA, 2019).

The first group of studies, focused on the role that legacies of violence play in electoral behavior, examines whether incumbent candidates face punishment or if voters lean towards right-wing or more conciliatory candidates who prioritize peace processes in conflict-ridden contexts. A study conducted in Colombia suggests the latter trend, indicating that insurgent violence encourages support for conciliatory candidates, contrary to the hypothesis that voters punish incumbents for security failures (WEINTRAUB et al., 2015). Other studies on the behavioral legacies of violence associated with civil wars or terrorist attacks suggest that victims of such conflicts often exhibit increased political engagement and participation compared to non-victims (BELLOWS and MIGUEL, 2009; BLATTMAN, 2009; VOORS et al., 2012, apud LEY, 2018).

A second group of studies, primarily focused on Mexico, Colombia, and Brazil, gauges the impact of armed territorial control and violence by organized crime on the electoral behavior of residents in these areas (GARCIA-SANCHEZ, 2014), including crimes targeting politicians and candidates (BORBA et al., 2022). Ley’s study (2018) analyzes how organized crime strategically employs violence during electoral campaigns to discourage voter participation, highlighting how this criminal-electoral violence reduces voter turnout and influences individual electoral choices. Trudeau (2022) explores how criminal groups act as brokers in elections through two main mechanisms: coercing voters to the polls and restricting access to rival candidates. The first mechanism increases voter turnout and shapes voting behavior, while the second limits candidate options, reducing competitiveness and boosting the chances of victory for candidates supported by organized crime. Additionally, studies such as Borba et al. (2022) contribute to the theoretical debate on electoral violence as a form of political violence by examining the prevalence of such violence on local political leaders during Brazil’s 2020 municipal elections.

Lastly, the third group of studies focuses on the impact of criminal victimization on political participation and electoral choices. Ley (2018) observes that victims often experience a loss of political power in areas with high homicide rates, whereas Berens and Dallendörfer (2019) compare how feelings of suffering, anger, and apathy influence electoral engagement among victims of both violent and non-violent crimes.

A portion of the literature within this third group of studies aims to understand not only voter turnout but also the direction of the vote, particularly the preference for Law and Order candidates. In the Mexican context, Ventura et al. (2023) examine how direct experiences with violence and the profiles of candidates influence electoral choices. Their findings indicate that it is not possible to assess the effects of violence on support for security policies as a whole, emphasizing the need to differentiate between repressive and preventive policies. While there is a notable positive effect on support for candidates with experience in public security, this support extends beyond candidates from security forces to include human rights activists as well.

In the same study, the authors argue that it is essential not only to distinguish between repressive and preventive measures but also to determine which punitive measures will be implemented and by whom. They point out that certain repressive actions, such as militarization, can have varying electoral effects, ranging from neutral to adverse, depending on the electorate's level of exposure to crime and their fear of violence.

In the Brazilian context, public security is a major component of the electoral agenda, largely driven by the disturbingly high levels of crime and violence that thrust certain cities into the spotlight. This situation, intensified by media sensationalism, amplifies the widespread feeling of insecurity. Recent research has specifically examined how violence influences the electoral performance of Law and Order candidates (LUZ and SANT'ANNA, 2019; VENTURA, 2022) and, to a lesser extent, the inverse correlation (NOVAES, 2023).

However, in contexts characterized by profound inequality in the distribution of violence, such as Brazil, it becomes crucial to consider how this phenomenon impacts different segments of the population. Studies indicate that victimization from violence erodes trust in democratic institutions and the criminal justice system, leading to decreased electoral participation. Additionally, the fear of crime itself, irrespective of victimization, prompts approval for repressive measures by law enforcement agencies (VENTURA, 2022).

Sociological literature on violence has long distinguished between the fear of crime and actual victimization, emphasizing that these are not necessarily linked; fear is not solely shaped by direct or indirect victimization. In this context, individuals with higher socioeconomic status may perceive violence in a way that leads them to demand a stronger police presence and more visible security measures in the streets. In contrast, vulnerable populations, who frequently witness police violence in their neighborhoods, are less likely to support such measures (COSTA and DURANTE, 2019; OLIVEIRA JR., 2011; ZILLI and COUTO, 2017).

Luz and Sant'Anna (2019) emphasize the importance of this distinguishing between actual violence and the perception of violence, asking “whether the key factor for the success of L&O candidates is actual violence (or variations thereof) or the perception of violence, which may not actually correlate with real crime levels” (LUZ and SANT'ANNA, 2019, p. 03). While their analysis points to actual violence as the significant factor, other studies have disputed this conclusion.

Macaulay (2019) suggests that fear of violence – rather than actual homicide rates – had a greater impact in favor of Bolsonaro in the 2018 elections. Ventura (2022) adds that theoretical approaches to securitization, which claim that voters most affected by violence become more concerned with security and therefore more likely to support punitive policies and L&O candidates, must be tempered by the fact that the costs of such policies are primarily borne by the most socially and racially vulnerable groups (VENTURA, 2022). Ventura’s analysis (2022) reveals that the appeal and support for L&O candidates have a positive electoral effect in wealthy, conservative areas, thereby perpetuating existing inequalities in violence within broader society. Therefore, the first hypothesis proposed in this article is that elevated municipal violence indicators correlate negatively with the proportion of votes for L&O candidates.

In addition to municipal-level violence rates, we introduce another vital but often overlooked dimension in the analyses of the local bases of support for L&O candidates: municipal capacity in the security sector. For instance, Novaes (2023) observed an increase in homicides of non-white men in municipalities that elected Law and Order councilors, implying that this outcome may have resulted from changes in local security structures implemented by municipal legislators.

The integration of security structures and policies at the municipal level reflects a trend of induced municipalization by the federal government in the late 1990s and early 2000s, particularly following the establishment of the National Secretariat of Public Security (SENASP) and the subsequent introduction of the PRONASCI (National Program of Public Security and Citizenship) (MADEIRA and RODRIGUES, 2015). This initiative directed municipalities to engage in security efforts, believing that local authorities could better address specific factors to improve indicators in this arena, particularly through preventive policies (OLIVEIRA JR. and ALENCAR, 2016, pp. 26-27). The municipalization of public security aimed to address the highly uneven manifestation of violence in Brazil, often targeting vulnerable municipalities and territories with high rates of violent crime, through preventive policies (MESQUITA NETO, 2011).

The introduction of national security plans and the rationalization of resources from the National Public Security Fund encouraged municipalities to establish security secretariats, develop local plans, and deploy municipal guards, positioning municipalities as key players in violence prevention (OLIVEIRA JR., 2017, p. 94). The municipalization of public security involved creating these and other institutional structures, followed by the adoption of preventive actions driven by the need to develop policies tailored to local realities and responsive to community expectations (MADEIRA and RODRIGUES, 2015). It's worth noting that while the establishment of security sector capacity in municipalities was not directly tied to violence indicators, the adoption of preventive policies responded to federal encouragement for areas with high susceptibility to violence. Despite a decrease in federal funding to municipalities from 2015 onward, they continued to increase their expenditures on local security policies in response to growing demands from the population (PERES et al., 2016, p, pp. 76-77).

In this article, we consider the following structures as components of municipal state capacity in public security: Public Security Plan and Fund, Public Security Council and Community Council, and the existence of a municipal guard. A particularly important criterion in our assessment is the presence of a municipal security secretariat, whether exclusive or directly subordinated to the mayor, due to its role in formulating and coordinating public policies, akin to state security secretariats (COSTA, 2015). This framework enables us to more effectively measure the coercive, fiscal, administrative, and relational dimensions of municipal state capacity in the field of public security (MADEIRA et al., 2018).

Lastly, the literature has long emphasized the role of municipalities in the election of federal parliamentarians. While the entire state serves as the electoral district for Lower House elections, the expansive and diverse territories of these states, coupled with the significance of municipalities in Brazil’s federal structure, result in many candidates for the Federal Congress having backgrounds and strategies tied to specific municipalities (AMES, 2002). Hence, local characteristics are crucial for understanding the performance of candidates.

In this context, several studies highlight a specific factor: the support of mayors, who can act as decisive brokers in federal legislative elections (AVELINO, BIDERMAN, and BARONE, 2012; NOVAES, 2018). One mechanism for this effect is the credit-claiming strategy, where mayors emphasize the legislative achievements of parliamentarians in their localities (BAIÃO and COUTO, 2017; VENTURA, 2021). While such studies consider a broad spectrum of candidates for federal congress, this article focuses solely on those associated with public security. Thus, we aim to deepen the understanding of the specific mechanisms connecting municipal attributes and the role of mayors in federal legislative elections. Consequently, the second hypothesis of this article is that the negative impact of violence on voting preferences is amplified by the presence of local security structures.

Data and methodology

One of the main methodological challenges consists in identifying which politicians belong to a particular thematic group. For studies focusing on elected officials, these criteria may include membership in formalized parliamentary caucuses, sponsorship of bills, drafting of legislation, participation in committees, speeches, among others. Some studies (e.g. QUADROS and MADEIRA, 2018), rely on characterizations provided by parliamentary monitoring bodies such as the Inter-Union Parliamentary Assistance Department (DIAP - Departamento Intersindical de Assessoria Parlamentar), while others undertake case studies of specific cities, conducting in-depth analyses of individual politicians in the security field Faganello (2015).

The greatest challenge lies in analyzing all candidates in a national election. Brazil’s Superior Electoral Court (TSE) does have a database of all candidates, which includes self-declared professions (CEPESP, 2022; TSE, 2019)². However, this information often fails to accurately reflect reality, as candidates might identify themselves as politicians due to having held prior political offices at any level, even if they have strong ties to another profession. Inspired by the approaches proposed by Ventura (2022), Novaes (2023), and Luz and Sant'Anna (2019), as well as those adopted by Boas (2014) and Lacerda (2018) for other groups, we included both the candidates’ ballot names and professions in our identification strategy. Thus, we identified federal congressional candidates in Brazilian states by grouping them based on their professional occupation and electoral pseudonym (ballot name). Table 01 summarizes the mechanisms used for identifying candidacies and collecting data on the number of votes for public security candidates.

Table 01
Mechanism for identifying Security Force candidates

Identification Mechanism	Reference Sources
Occupation listed on the candidate's registration form.	Occupations registered with the TSE: firefighter, civil police officer, military police officer, member of the Armed Forces, retired military personnel.
Occupation listed on the candidate's registration form, name used on the ballot, or membership in the Car-Wash Front.	(a) Ballot Name: Agent, Corporal, Captain, Capt., Commander, Colonel, Civil Guard, Federal Police Chief, Federal Police Officer, Guard, Investigator, Major, Police Chief, Police Officer, Civil Police Officer, Military Police Officer, Soldier, Sergeant, Sgt., Sublieutenant, Security Guard, Lieutenant; (b) Member of the Federal Police Front: a list of seventeen candidates, one per state (South - PR and RS; Southeast - ES, MG, RJ, SP; Midwest - DF, MS, and MT; Northeast - BA, CE, MA, PE; North - AC, AM, AP, and RR).

We utilized the sum of votes for candidates identified as Law & Order as the dependent variable, divided by the total nominal votes for federal legislators in the municipality³. As depicted in Figure 01 below, the distribution of votes for L&O candidates exhibits significant spatial variation⁴. The Southeast region stands out positively, contrasting with the South and Northeast regions.

Figure 01
Proportion of votes associated with security forces in municipalities (%)

In our study, we investigated two independent variables of interest: the violence rate and the presence of municipal capacity in the security sector. To calculate the violence rate, we utilized data from DataSUS (BRASIL, 2022), aggregating three datasets of deaths by municipality: two categories of ‘events (facts) of undetermined intent’ (codes A00 and D00, Major Group CID10 Y10-Y34); two categories of ‘assaults’ (codes B00 and C00, Major Group CID10 X85-Y09); and ‘legal interventions and operations of war’ (code E00, Major Group CID10 Y35-Y36). We collected data from 2018, the election year, to capture the nearly simultaneous impact of violence levels on candidates for federal legislative positions. Each municipality was assigned a unit value, which was subsequently weighted by 100,000 inhabitants and logarithmized.

We obtained data on the structure and capacities of municipal public security from Munic/IBGE 2019⁵, which is the dataset closest to the 2018 election year. Our focus was on identifying the presence of a municipal public security secretariat, its status, and other capacities in the security sector within the municipal administration. As mentioned earlier, some studies suggest that mayors play a significant role in the election of legislators in general. However, our analysis aims to deepen this understanding by specifically examining the case of public security.

We devised a dichotomous measure that assumes a value of 01 if there is an exclusive municipal public security secretariat or if the sector reports directly to the head of the municipal executive, along with the presence of at least one other municipal public security policy structure, as mentioned earlier⁶. Otherwise, the measure takes a value of ‘0’. This approach allows us to assess the significance attributed by local policymakers to the public security sector, with the goal of determining whether this structure influences electoral performance.

We incorporated a variety of controls encompassing socioeconomic, demographic, and political characteristics of municipalities to gauge the impact of violence and local municipal capacity. Per capita GDP (in logarithmic terms), the GINI coefficient, and the percentage of the population living in poverty serve as indicators of wealth, income distribution, and the level of deprivation within the population, all of which could potentially influence electoral outcomes, levels of violence, and the infrastructure of local government. Other factors such as population size (logarithmically), the proportion of evangelical Christians, non-white individuals (Black people, Brown people, and other racial groups), urbanization rates⁷, and regional disparities are included to capture dimensions that may affect federal representative elections, as well as levels of violence and state capacities. Furthermore, we introduced strictly political and partisan variables: the magnitude (measured by the number of parliamentary seats available) of each state, along with the Effective Number of Parties in the 2014 election for the Federal Congress (calculated for each municipality), serve to assess the potential receptiveness of the party-electoral system to diverse agendas. Additionally, the percentage of votes for parties grouped by ideological blocs in the 2014 election, as well as the ideology of the winning mayor’s party in 2016⁸, indicate the municipal inclination toward ideological blocs. Finally, we included the percentage of security sector candidates out of the total candidates in each state to adjust for performance relative to supply.

We applied additional specifications, including robustness tests of the results. Specifically, in models 02 and 04, we introduced votes for Jair Bolsonaro in the presidential election as a control variable, due to his association with several L&O politicians, such as his own son Eduardo Bolsonaro⁹, who secured the highest number of votes for office in both the state of São Paulo and nationwide. However, the inclusion of this variable likely introduces endogenous control issues, given the simultaneous relationship between this variable and the dependent variable. In the appendix, we present results from a model that incorporates clustered standard errors by state¹⁰, which aims to accommodate the geographical nature of the data and the rule designating this level as the electoral district for parliamentarians.

Finally, before presenting the results, we must acknowledge an analytical limitation of our data. Since we aggregated data at the municipal level and relied on objective indicators, we were unable to observe differences within municipalities or directly measure individual and subjective perceptions of violence. However, this limitation is offset by employing the appropriate level to examine the impact of municipal public policies on the electoral performance of candidates from the security sector.

Findings and discussion

Table 02 below displays the results of four models: the first is purely additive; the second incorporates votes for candidate Jair Bolsonaro as a control variable; models three and four include the interaction between the level of violence and municipal capacity in the security sector, with the latter model also incorporating votes for Jair Bolsonaro.

Table 02
OLS Models: Dependent Variable: % of municipal votes for candidates from the security sector, with and without interaction, and controlled for votes for candidate Jair Bolsonaro (1st Round) - 2018 Elections

	Additive and Interactive Models
	(1) Additive	(2) Additive	(3) Interactive	(4) Interactive
Log Violence	-0.569***	-0.548***	-0.490***	-0.465***
	(0.106)	(0.105)	(0.111)	(0.111)
State Capacity in Security Sector	0.509**	0.520**	2.666***	2.768***
	(0.248)	(0.247)	(1.001)	(0.996)
Interactions
Log Violence x State Capacity in Security Sector			-0.663**	-0.691**
			(0.298)	(0.297)
Electoral-political dimension
Proportion of Law and Order (L&O) Candidates	0.449***	0.509***	0.452***	0.513***
	(0.049)	(0.049)	(0.049)	(0.049)
Effective Number of Parties - 2014	0.456***	0.442***	0.457***	0.443***
	(0.039)	(0.039)	(0.039)	(0.039)
Magnitude	0.084***	0.096***	0.084***	0.095***
	(0.008)	(0.008)	(0.008)	(0.008)
Center-Party Mayor - 2016	0.480***	0.468***	0.484***	0.472***
	(0.154)	(0.154)	(0.154)	(0.154)
Left-Party Mayor - 2016	0.600***	0.624***	0.602***	0.626***
	(0.190)	(0.190)	(0.190)	(0.190)
% of Votes for Center Parties - 2014	0.005	0.004	0.004	0.004
	(0.004)	(0.004)	(0.004)	(0.004)
% of Votes for Left Parties - 2014	0.001	0.004	0.001	0.004
	(0.005)	(0.005)	(0.005)	(0.005)
Municipal Dimension
Log Municipal GPD	0.634***	0.496***	0.622***	0.483***
	(0.160)	(0.160)	(0.160)	(0.160)
Log Population	0.098	0.022	0.109	0.033
	(0.079)	(0.079)	(0.079)	(0.079)
% Urbanization	0.033***	0.032***	0.033***	0.032***
	(0.005)	(0.005)	(0.005)	(0.005)
% Evangelicals	0.044***	0.025*	0.045***	0.025*
	(0.013)	(0.013)	(0.013)	(0.013)
% Non-White Population	-0.070***	-0.056***	-0.069***	-0.056***
	(0.006)	(0.006)	(0.006)	(0.006)
% Poor Population	0.010	0.035***	0.009	0.034***
	(0.011)	(0.012)	(0.011)	(0.012)
GINI Index	2.555*	2.604*	2.594*	2.644*
	(1.470)	(1.464)	(1.470)	(1.464)
Regions
Northeast	-0.784**	-0.061	-0.751**	-0.023
	(0.375)	(0.388)	(0.375)	(0.388)
North	-0.523	-0.573	-0.525	-0.575
	(0.384)	(0.383)	(0.384)	(0.382)
Southeast	-1.054**	-1.150***	-1.022**	-1.117***
	(0.427)	(0.426)	(0.427)	(0.426)
South	-5.514***	-5.194***	-5.468***	-5.144***
	(0.356)	(0.357)	(0.356)	(0.358)
Controls – Votes for Jair Bolsonaro
Valid Votes for Bolsonaro in the 1st Round		0.059***		0.059***
		(0.009)		(0.009)
Constant	-7.105***	-9.385***	-7.385***	-9.687***
	(1.849)	(1.870)	(1.852)	(1.874)
Observations (N)	5,554	5,554	5,554	5,554
Adjusted R2	0.328	0.334	0.329	0.335

• Source: Elaborated by the authors from the article’s database.
• Notes: (01): Standard errors in parentheses; (02): p<0.1*; p<0.05**; p<0.01***.

To begin with, we may analyze the outcomes of the additive model, the first model. Our estimates corroborate one side of the literature: there is a negative correlation between the level of violence and voting for L&O candidates in 2018. Violence exhibits highly unequal patterns across municipalities, primarily victimizing individuals from lower socioeconomic backgrounds, who bear the brunt of this phenomenon. Additionally, the presence of security forces also varies unevenly among municipalities. Moreover, violence and feelings of insecurity or fear are typically disconnected, resulting in an electorally favorable impact in conservative and affluent areas, thus perpetuating the existing inequality in society’s experience of violence.

The impact of local state capacities, still on the additive model, partially contradicts our expectations, showing a positive trend. Holding other variables constant, the existence of municipal security structures and policies is associated with a slight increase in the estimated vote for ‘mano dura’ candidates by just over 0.50%.

However, this finding only partially captures the relationship between these variables. Upon analyzing models 03 and 04, we find that not only is the coefficient of the interaction between the level of violence and local capacities in the security sector statistically significant, but it also intriguingly displays a negative sign. Put differently, the presence of a security secretariat, along with at least one additional structure from the sector, intensifies the adverse impact of violence on the electoral performance of L&O candidates. For a clearer understanding of the predicted effects, we may refer to Table 03 below, which shows the marginal effect of violence in scenarios with and without local state capacity, using model 03:

Table 03
Interaction - Marginal Effect of ‘Log Violence’ in Interaction with State Capacity

	Marginal Effect	Standard Error	t	P>|t	[95% CI]
State Capacity in the Security Sector = 0	-0.49	0.111	-4.40	0.000	-0.709	-0.272
State Capacity in the Security Sector = 1	-1.153	0.283	-4.08	0.000	-1.707	-0.598

The marginal impact for each 1-point increase in the (log) violence rate is -1.15 in municipalities with state security structures and -0.49 in those without them. The disparity between these two estimates, amounting to -0.66, is statistically significant at the 95% confidence level.

We propose a plausible explanation for this outcome: considering the inclination of municipal security policies to adopt a preventive approach in more severe contexts, it follows that candidates advocating for more punitive measures perform less favorably as violence increases compared to municipalities without a local security agenda. This also helps explain the positive outcome of the variable ‘state capacities’ in the additive model. Given that citizen security policies are typically formulated and executed in municipalities grappling with high rates of violence and crime, the estimated impact of local capacities, irrespective of the violence level, demonstrates a positive correlation.

These findings contribute to the existing literature, reinforcing the effects of the unequal distribution of violence (VENTURA, 2022), as well as the influence of authorities' responses to violence on electoral preferences. While Novaes (2023) noted that increased violence correlates with fewer votes for L&O candidates, and that electing councilors with this profile exacerbates violence in vulnerable sectors, our data suggest that the existence of municipal-level capacity in the security sector, diminishes electoral backing for these candidates in specific contexts. Presumably due to the implementation of policies diverging from those championed by this group.

The results of the other variables generally aligned with our initial expectations, albeit with some noteworthy exceptions. Concerning the electoral-political dimension, our findings indicate that party fragmentation, district magnitude, and candidate supply positively correlate with votes for candidates from the security sector. However, when examining the ideological aspect of municipalities, we encountered different outcomes depending on the electoral scope under consideration. The distribution of votes across ideological blocs in 2014 showed no correlation with votes for L&O candidates in 2018. Surprisingly, when considering the ideology of the parties of mayors elected in 2016, municipalities governed by left and center parties witnessed a higher vote share for security sector candidates two years later compared to those governed by right-wing parties. We currently lack an explanation for this particular finding.

Our findings from the various models suggest that voting for L&O candidates is more prevalent in wealthier and more urbanized municipalities. The multivariate models alter the conclusions derived from the simple visualization of the map concerning the vote for L&O candidates. Specifically, states in the South, Southeast, and Northeast regions exhibit a negative and significant coefficient compared to the Midwest. However, when controlling for the vote for Jair Bolsonaro in models 02 and 04, the significance of the Northeast region diminishes.

Population profile variables, such as the proportion of evangelicals, non-white population, and income inequality, reveal intriguing electoral behavior trends. The proportion of evangelicals and the GINI index both positively correlate with voting for L&O candidates, while the presence of racial minorities negatively correlates with the performance of these candidates. These findings provide evidence supporting the notion that the performance of L&O candidates is positively associated with conservative agendas and privileged groups.

We do find indications of a racial relationship with support for punitive measures and investment in crime prevention. According to Barkan and Cohn (2005), white people (and potentially other privileged groups) typically envision criminals as being Black, leading them to support harsher measures against criminals and greater investment in fighting crime. Regarding the connection to racial prejudices, the authors argue that “given the popular belief that street criminals tend to be African Americans, conflict theory would suggest that it is in whites’ self-interest to be punitive toward criminals” (BARKAN and COHN, 2005, p. 303). Not perceiving their own group as capable of being categorized as criminals leads these groups to not fear, but rather support, increased repressive measures and policing.

Lastly, in the appendix, we present the results from the model with standard errors clustered by state, as explained above. As shown, some important variables lose significance, but our conclusion regarding the interaction remains unchanged.

Concluding remarks

In this article, we analyzed the rise of Law and Order candidates elected in the 2018 Federal Congress elections in Brazil, aiming to identify the factors influencing voters’ choices at the municipal level.

There is no consensus in the literature regarding the impact of violence on the electoral performance of Law and Order candidates. While some studies suggest that higher homicide rates lead to increased support for these candidates, others indicate the opposite trend. Our findings align with the latter group, showing a negative correlation between violence rates and voter preference for candidates from the public security sector. This finding also reinforces interpretations regarding the uneven local distribution of violence and its varying effects across different socio-economic strata, underscoring the distinction between actual violence and feelings of insecurity or fear, suggesting that the latter may drive the preference for these candidates, while the former may contribute to their rejection.

We underline, nevertheless, that our work innovates in this research field by offering alternative explanatory possibilities for this phenomenon. By examining institutional factors, we find that the existence of a dedicated municipal secretariat for the security sector – or directly subordinate to the mayor – alongside another local security policy structure, increases support for L&O candidates when other variables are held constant and unconditionally. Interestingly, however, when we combine violence with this particular state capacity in the field of security, we observe a negative interactive effect, revealing that the decrease in votes caused by increased violence is more pronounced in municipalities with a municipal security secretariat.

We offer the following explanation for these findings: In unconditional terms, the presence of municipal capacities could bolster the electoral support for L&O candidates due to their direct association with mayors. However, in highly violent municipalities, particularly those witnessing elevated levels of state violence against specific groups – often vulnerable populations – local capacity and policies in the security sector tend to prioritize violence prevention. Consequently, the electorate is more inclined to reject the reactive policies typically championed by L&O candidates.

Lastly, we acknowledge that our analysis has certain limitations. Apart from the inevitable synchronization issues with some data pertaining to the 2018 elections, such as data from the 2010 Census, our violence indicator remains static. This limitation hampers our ability to ascertain whether the effects we identified vary across the temporal dynamics of this phenomenon in the pre-election period. Additionally, we must approach causal arguments based on observational data with caution and acknowledge the ecological nature of our information. Given that our observations are at the municipal level, we cannot draw inferences about intra-municipal regions or individual behaviors.

Future studies could further explore the relationship between voter preference for candidates linked to the public security sector and the local context of Brazilian municipalities across various electoral cycles. These studies could investigate whether the observed negative correlation between violence and support for L&O candidates in municipalities, a correlation that intensifies with the prominence of the municipal state's capacity in public security, occurred during the 2014 and 2022 congressional elections. This comparative analysis could shed light on whether these findings occurred in the period before the emergence of Bolsonarism, albeit already influenced by far-right movements, and during the time when Jair Bolsonaro fails in his bid for reelection, despite the continued strength of his current policies in Congress. Another promising analytical avenue involves examining the electoral performance of candidates associated with the security sector who advocate for civil and human rights.

Acknowledgments:

We thank Ludmila Mendonça Ribeiro and Cleber da Silva Lopes.

References

• ADORNO, Sérgio and SALLA, Fernando (2007), Organized criminality in prisons and the attacks of the PCC. Estudos Avançados. Vol. 21, pp. 07-29.
• AMES, Barry (2002) The deadlock of democracy in Brazil. Ann Arbor: University of Michigan Press. 352 pp..
• AVELINO, George; BIDERMAN, Ciro, and BARONE, Leonardo S. (2012), Articulações intrapartidárias e desempenho eleitoral no Brasil. Dados Vol. 55, Nº 04, pp. 987-1013.
• BAIÃO, Alexandre Lima and COUTO, Cláudio Gonçalves (2017), Eficacia del pork barrel: la importancia de enmiendas presupuestarias y alcaldes aliados en la elección de diputados. Opinião Pública Vol. 23, Nº 03, pp. 714-753.
• BARKAN, Steven E. and COHN, Steven F. (2005), Why whites favor spending more money to fight crime: the role of racial prejudice. Social problems Vol. 52, Nº 02, pp. 300-314.
• BELLOWS, John and MIGUEL, Edward (2009), War and local collective action in Sierra Leone. Journal of Public Economics Vol. 93, Nº 11/12, pp. 1144-1157.
• BERENS, Sarah and DALLENDÖRFER, Mirko (2019), Apathy or anger? How crime experience affects individual vote intention in Latin America and the Caribbean. Political Studies Vol. 67, Nº 04, pp. 1010-1033.
• BERLATTO, Fábia; CODATO, Adriano, and BOLOGNESI, Bruno (2016), Da polícia à política: explicando o perfil dos candidatos das Forças Repressivas de Estado à Câmara dos Deputados. Revista Brasileira de Ciência Política Nº 21, pp. 77-120.
• BLATTMAN, Christopher (2009), From violence to voting: war and political participation in Uganda. American Political Science Review Vol 103, Nº 02, pp. 231-247.
• BOAS, Taylor C. (2014), Pastor Paulo vs. Doctor Carlos: professional titles as voting heuristics in Brazil. Journal of Politics in Latin America Vol. 06, Nº 02, pp. 39-72.
• BOLOGNESI, Bruno; RIBEIRO, Ednaldo, and CODATO, Adriano (2023), Uma nova classificação ideológica dos partidos políticos brasileiros. Dados. Vol. 66, Nº 02, pp. 01-29.
• BORBA, Felipe; ISRAEL, Vinícius; CARNEVALE, Miguel P., and BAHIA, Pedro (2022), Political and electoral violence in the 2020 municipal elections. Revista Brasileira de Ciências Sociais. Vol. 37, Nº 108, pp. 01-21.
• BRASIL, Ministério da Saúde (2022), Banco de dados do Sistema Único de Saúde-DATASUS. Available at ? http://www.datasus.gov.br?. Accessed on November 30, 2022.
  » http://www.datasus.gov.br
• CEPESP (2022), CepespR: R API for extrating data storaged on CEPESPData. Available at ? https://github.com/Cepesp-Fgv/cepesp-r ?, ? http://cepesp.io/ ?, ? http:// cepesp.fgv.br/ ?. Accessed on November 30, 2022.
  » https://github.com/Cepesp-Fgv/cepesp-r» http://cepesp.io/» http:// cepesp.fgv.br/
• CODATO, Adriano; BERLATTO, Fábia, and BOLOGNESI, Bruno (2018), Tipologia dos políticos de direita no Brasil. Análise Social Vol. 53, Nº 229, pp. 870-897.
• COSTA, Arthur Trindade (2015), Estado, governança e segurança pública no Brasil: uma análise das secretarias estaduais de Segurança Pública. Dilemas-Revista de Estudos de Conflito e Controle Social Vol. 08, Nº 04, pp. 607-632.
• COSTA, Arthur Trindade Maranhão and DURANTE, Marcelo Ottoni (2019). A polícia e o medo do crime no Distrito Federal. Dados Vol. 62, Nº 01, pp. 01-31.
• COSTA, Arthur Trindade Maranhão and MARQUES, David (2022), Número de candidaturas de policiais nas eleições gerais de 2022 cresce 27% em relação a 2018. Boletim Fonte Segura. Vol. 150. Available at ? https://fontesegura.forumseguranca.org.br/numero-de-candidaturas-de-policiais-nas-eleicoes-gerais-de-2022-cresce-27-em-relacao-a-2018/? Accessed on August 18, 2022.
  » https://fontesegura.forumseguranca.org.br/numero-de-candidaturas-de-policiais-nas-eleicoes-gerais-de-2022-cresce-27-em-relacao-a-2018/?
• FAGANELLO, Marco Antônio (2015), Bancada da bala: uma onda na maré conservadora. In: Direita, volver: o retorno da direita e o ciclo político brasileiro. Edited by CRUZ, Sebastião Velasco e; KAYSEL, André, and CODAS, Gustavo. São Paulo: Ed. Fundação Perseu Abramo. pp. 145-161.
• FAGUNDES, Andréa Lucas and MADEIRA, Lígia (2021), A Polícia Federal e o Governo Bolsonaro: duas décadas de desenvolvimento e dois anos de ataques e resistências. Governo Bolsonaro: retrocesso democrático e degradação política. Edited by AVRITZER, Leonardo; KERCHE, Fábio, and MARONA, Marjorie. Belo Horizonte: Autêntica Editora. pp. 175-185.
• FENAPEF (2018a), Mais Policiais Federais na Política em 2018. Revista FENAPEF em Ação. Vol. 02, pp. 10-21.
• FENAPEF (2018b), 04 Milhões de votos de confiança. Revista FENAPEF em Ação. Vol. 03, pp. 14-19.
• GARCIA-SANCHEZ, Miguel (2014), Violent contexts and vote choice in Colombia: a multilevel approach. Available at ? https://ssrn.com/abstract=2500237? Accessed on July 14, 2024.
  » https://ssrn.com/abstract=2500237?
• IVKOVIC, Sanja Kutnjak (2008), A comparative study of public support for the police. International Criminal Justice Review Vol. 18, Nº 04, pp. 406-434.
• INGRAM, Matthew C. and COSTA, Marcelo Marchesini da (2019), Political geography of violence: municipal politics and homicide in Brazil. World Development. Vol. 124, Issue C, pp. 01-15..
• LACERDA, Fábio (2018), Assessing the strength of Pentecostal Churches' Electoral Support: evidence from Brazil. Journal of Politics in Latin America Vol. 10, Nº 02, pp. 03-40.
• LEY, Sandra (2018), To vote or not to vote: how criminal violence shapes electoral participation. Journal of Conflict Resolution Vol. 62, Nº 09, pp. 1963-1990.
• LIMA, Renato Sérgio de (2020), Eleições de policiais no Brasil e a força do 'partido policial'. Anuário Brasileiro de Segurança Pública. Vol. 14, pp. 158-201.
• LUZ, Guilherme Gomes and SANT'ANNA, Pedro Carregã (2019), Crime e voto: efeito da violência pública na eleição de candidatos Law & Order no Brasil. Paper. Anais do 47° Encontro Nacional de Economia/Anpec.
• MACAULAY, Fiona (2019), Bancada da bala: the growing influence of the security sector in Brazilian politics. In spite of you: Bolsonaro and the New Brazilian resistance. Edited by FOLEY, Conor. New York: OR Books. pp. 56-70.
• MADEIRA, Lígia Mori; GELISKI, Leonardo; SALVARREY, Gabriela; BOESSIO, Ana Paula, and TRINDADE, Kalita (2018), Capacidade estatal na segurança pública municipal das capitais brasileiras. In: A política, as políticas e os controles: como são governadas as cidades brasileiras. Edited by MARENCO, André. and NOLL, Maria Izabel. Porto Alegre: Tomo Editorial. pp. 141-170.
• MADEIRA, Lígia Mori and RODRIGUES, Alexandre Ben (2015), Novas bases para as políticas públicas de segurança no Brasil a partir das práticas do governo federal no período 2003-2011. Revista de Administração Pública Vol. 49, Nº 01, pp. 03-21.
• MESQUITA NETO, Paulo de (2011), Ensaios sobre segurança cidadã. São Paulo: Quartier Latin/FAPESP. 472 pp..
• MUNIZ, Jacqueline de Oliveira; CARUSO, Haydée, and FREITAS, Felipe da Silva (2018), Os estudos policiais nas ciências sociais: um balanço sobre a produção brasileira a partir dos anos 2000. Boletim Informativo Bibliográfico Nº 84, pp. 148-187.
• NOVAES, Lucas Martins (2023), The violence of Law-and-Order politics: the case of Law enforcement candidates in Brazil. American Political Science Review Vol. 118, Nº 01, pp. 01-20.
• NOVAES, Lucas Martins (2018) Disloyal brokers, and weak parties. American Journal of Political Science. Vol. 62, Nº 01, pp. 84-98.
• OLIVEIRA JÚNIOR, Almir de (2017), A SENASP e o papel dos municípios na Segurança Pública: o caso das guardas civis municipais. In: Atlas da violência 2017: política nacional de Segurança Pública orientada para a efetividade e o papel da Secretaria Nacional de Segurança Pública. Edited by CERQUEIRA, Daniel Ricardo de Castro; FERREIRA, Helder; BUENO, Samira; SILVA, Fábio de Sá e; LIMA, Renato Sérgio de; LEMGRUBER, Julita; MUSUMECI, Leonarda; KOPITTKE, Alberto; FIGUEIREDO, Isabel; SANTOS, José Vicente Tavares dos; COSTA, Arthur Trindade M., and OLIVEIRA JR., Almir de. Brasília: IPEA. pp. 94-101.
• OLIVEIRA JÚNIOR, Almir de (2011), Dá pra confiar nas polícias? Confiança e percepção social da polícia no Brasil. Revista Brasileira de Segurança Pública Vol. 05, Nº 09, pp. 06-22.
• OLIVEIRA JÚNIOR, Almir de and ALENCAR, Joana Luiza Oliveira (2016), Novas polícias? Guardas municipais, isomorfismo institucional e participação no campo da segurança pública. Revista Brasileira de Segurança Pública Vol. 10, Nº 02, pp. 24-34.
• PERES, Ursula Dias; BUENO, Samira, and TONELLI, Gabriel Marques (2016), Os municípios e a Segurança Pública no Brasil: uma análise da relevância dos entes locais para o financiamento da Segurança Pública desde a década de 1990. Revista Brasileira de Segurança Pública Vol. 10, Nº 02, pp. 36-56.
• QUADROS, Marcos Paulo dos Reis and MADEIRA, Rafael Machado (2018), Fim da direita envergonhada? Atuação da bancada evangélica e da bancada da bala e os caminhos da representação do conservadorismo no Brasil. Opinião Pública Vol. 24, Nº 03, pp. 486-522.
• SOUZA, Luís Antônio Francisco de and SERRA, Carlos Henrique Aguiar (2020), Quando o Estado de exceção se torna permanente: reflexões sobre a militarização da segurança pública no Brasil. Tempo Social Vol. 32, Nº 02, pp. 205-227.
• TRUDEAU, Jessie (2022), How criminal governance undermines elections. Paper presented at American Political Science Association.
• TSE (2019), Base do TSE - Estatísticas Eleitorais - Candidaturas. Available at <http://www.tse.jus.br/eleicoes/estatisticas/estatisticas-eleitorais>. Accessed on November 30, 2022.
  » http://www.tse.jus.br/eleicoes/estatisticas/estatisticas-eleitorais
• VENTURA, Tiago Augusto da Silva (2022), Violent democracies: essays on crime, inequality and preferences for protection in Latin America. Doctoral thesis. Faculty of the Graduate School. University of Maryland.
• VENTURA, Tiago Augusto da Silva (2021), Do mayors matter? Reverse coattails on congressional elections in Brazil. Electoral Studies. Vol. 69, pp. 01-11.
• VENTURA, Tiago Augusto da Silva; LEY, Sandra, and CANTÚ, Francisco (2023), Voting for Law and Order: evidence from a survey experiment in Mexico. Comparative Political Studies Vol. 57, Nº 04, pp. 551-583.
• WEINTRAUB, Michael; DUQUE, Juan Fernando Vargas, and FLORES, Thomas (2015), Vote choice and legacies of violence: evidence from the 2014 Colombian presidential elections. Research and Politics. Vol. 02, Nº 02, pp 01-08.
• ZILLI, Luís Felipe and COUTO, Vinícius Assis (2017), Servir e proteger: determinantes da avaliação pública sobre a qualidade do trabalho das Polícias Militares no Brasil. Sociedade e Estado Vol. 32, Nº 03, pp. 681-700.

• Data replication:
  https://doi.org/10.7910/DVN/PMNBY7
• *
  This article was presented at the 13th Meeting of the Brazilian Political Science Association (ABCP), held online from September 19 to September 23, 2022. It was honored as the best paper presented in the subfield of Public Security and Democracy.
• 1
  Faganello (2015) found a negative correlation between social vulnerability indicators and support for Law and Order (L&O) candidates in São Paulo.
• 2
  The candidate profiles and voting data by municipalities were retrieved from the CEPESP Data package using the R software. This package contains data sourced from the Brazilian Superior Electoral Court (TSE), compiled by the Research Center for Politics and Economics in the Public Sector (CEPESP) at the Getúlio Vargas Foundation (CEPESP, 2022).
• 3
  The descriptive statistics for the model variables can be found in Table A1 in the Appendix.
• 4
  Please refer to Figure A1 in the Appendix of this article for the distribution of the proportion of Law & Order candidates by state.
• 5
  The 2019 Municipal Basic Information Survey (MUNIC) included information on the presence of various public security mechanisms, such as the public security management agency, municipal structures, councils, funds, and plans related to the topic, as well as the presence of several other related structures.
• 6
  Reminder: Public Security Plan and Fund, Public Security Council and Community Council Public Security Council, and the existence of a municipal guard.
• 7
  The data source is the IBGE (Brazilian Institute of Geography and Statistics). GDP and population size data refer to 2018, while GINI, % of poor individuals, evangelicals, non-whites and urbanization come from the 2010 Census. This lag between the two datasets may introduce a non-negligible measurement error, which we have to acknowledge and accept in our analysis.
• 8
  We adhered to the typology developed by Codato, Berlatto, and Bolognesi (2018) and Bolognesi, Ribeiro and Codato (2023). Electoral data comes from CEPESPData (2022).
• 9
  In 2018, Eduardo Bolsonaro was one of the candidates of the Federal Police Agents Front (FENAPEF, 2018a, 2018b).
• 10
  See Table A2, ‘Model with cluster by state’ in the Appendix.

• Funding:
  National Council for Scientific and Technological Development (CNPq).

Data availability

https://doi.org/10.7910/DVN/PMNBY7

Publication Dates

• Publication in this collection
  04 Nov 2024
• Date of issue
  2025

History

• Received
  01 Dec 2022
• Accepted
  20 Jan 2024