The People’s Captain: Understanding Police Officers as an Electoral Brand

Abstract

In Brazil, police officers have increasingly competed in elections, mirroring expanded roles in government for coercive agents in cases such as the United States, the Philippines, and Mexico. In this paper, I develop a theory of how voters perceive these candidates. The symbol of the police officer represents the abstract idea of order and stability, more than just a crime fighter. Corruption scandals and insecurity in Brazil and Latin America have given rise to “order” in response to the chaos of politics as usual as a compelling electoral brand. I therefore expect that voters associate police officer candidates with strong, decisive leadership, social conservatism, a strong commitment to the rule of law, honesty in government, and a weak commitment to democracy. An image-based conjoint experiment fielded to a representative sample of Brazilian voters (N = 3,098) finds support for the theoretical framework, but respondents did not perceive the police officer candidate as less democratic than other professional profiles. The results help to inform our understanding of electoral politics in post-transition democratic regimes, demonstrating how armed agents of the state might remain popular and influential within the context of a consolidated democracy.

Keywords

Latin American politics elections and voting behavior police conjoint experiment

Introduction

In 2018, Brazil elected Jair Bolsonaro, a long-serving Federal Deputy from Rio de Janeiro but a then relatively unknown politician. He ran on a platform of anti-corruption, evangelism, and aggressive anti-crime rhetoric including a notorious phrase that came to define his campaign, bandito bom é bandito morto (the only good criminal is a dead criminal). One important feature of Bolsonaro’s candidacy and his subsequent government has been his connection with the security forces. Not only is Bolsonaro himself a former captain in the Brazilian army, but both of the vice presidential candidates he ran with, Hamilton Mourão in 2018 and Walter Braga Netto in 2022, were generals, and pro-Bolsonaro factions in congress are populated with former military officers. Campaigning for reelection, Bolsonaro’s official campaign jingle was even titled Capitão do Povo (The People’s Captain) to further emphasize this military connection.

However, Bolsonaro also projected a strong alliance with a different security sector corporation - the police. As Bolsonaro’s opponent Luiz Inácio da Silva (Lula) put it: Bolsonaro “doesn’t like the people, he likes the police” (Azevedo 2022). While the military and police are both agents of the state charged with wielding the state’s monopoly on the legitimate use of coercion (Bittner 1970), the police’s mission of maintaining public order within the state contrasts with the military’s outwardly-facing mission, meaning that these groups occupy similar, but not identical, spaces in the public’s imagination.

Arguably, Bolsonaro’s alliance with the police was just as pivotal during the campaign and transition than his relationship with the military, as high-profile incidents such as highway patrolmen in the Northeastern state of Pernambuco temporarily blocking Workers’ Party (PT) voters from the polls (Camargo and Falcão 2022), and the alleged complicity of some Brasília police in the January 8th riot (Teixeira and Alcântara 2023) dominated media coverage for a time. Police officers also ran for political office in the 2022 elections in record numbers (Zanini 2022). This growth in the number of police officers running in elections and the discrete incidents of political interference present important questions about the proper role of the security sector in democratic politics and how voters perceive this phenomenon. The growing participation of the police in elections at all levels of government in Brazil (Berlatto, Codato, and Bolognesi 2016; Pagnan and Rodrigues 2022; Zanini 2022) is striking, especially in a country like Brazil that has a long and unresolved history with military government (Atencio 2014), police violence (Magaloni et al. 2020; Arias and Barnes 2017; Fórum Brasileiro de Segurança Pública 2022), and collusion between law enforcement and armed criminal groups (Flom 2022).

This paper focuses specifically on police candidacies, developing a theory of the police electoral branding and testing the expectations empirically through an original image-based candidate choice experiment. While other scholars have treated the professional heuristic of police officer as interchangeable with “tough-on-crime” or mano dura policies, I argue that crime fighting is actually a single component of a broader idea: that of order. The police, the most visible organ of the state, occupy a complex position in society, and reducing their image down to simply a signal for punitive public security policies is missing important components of their electoral brand. I suggest that other features associated with this particular understanding of order such as strong leadership skills, credible claims to enforce a rule of law and fight corruption, and associations with traditional values and powerful social groups are important qualities that voters associate with the police.

This paper presents novel evidence from an image-based conjoint experiment fielded in July 2022 to support the theory. Respondents were presented with three pairs of randomly generated pairs of election handbills known in Brazil as santinhos and then asked to evaluate the candidate pairs along several theoretically motivated dimensions. Results lend support to the theory, although interestingly voters did not perceive police candidates as any more or less democratic than others.

In addition to providing the most complete understanding of police officer candidates to date, this paper presents two innovations. First, voters can and do differentiate between “tough-on-crime” policies and the professional signal of being a police officer. While the professional signal and policy signal are related, understanding the electoral appeal of security sector candidates, especially police officers, is conceptually distinct from studying the appeal of punitive public security policies.

Second, voters do not, on average, find police officer candidates to be undemocratic. This “null” finding indicates that voters are not rebelling against democracy in the abstract by supporting these candidates, and find their participation in the electoral process to be normal and possibly even desirable. The presence of politicians in uniform may not be as large of a threat to democratic consolidation as traditionally thought, at least in terms of the intentions of voters.

Security Sector Candidates

The blurring of lines between elected officials and state security agents is a familiar phenomenon in the Americas (Jaffe 2013; Trejo and Ley 2018). Frequently, the police in Brazil are responsible for much violent crime (Bueno 2014; Willis 2015) and are often perceived as inefficient. However, despite relatively low levels of trust in the police, police officer candidates have enjoyed electoral success in Brazil. The phenomenon is not unique to Brazil. Elsewhere in Latin America, candidates frequently play up or exaggerate their connections to law enforcement. In Argentina, Patricia Bullrich made liberal use of her public security background during electoral campaigns, including a viral moment where a group of police officers saluted her at a restaurant as if she were their commander (Clarín.com 2021).¹ Even left-leaning politicians such as Claudia López, mayor of Bogotá, have appropriated symbols of the security sector for campaign usage, dressing up like an officer and pretending to drill new recruits (Reynoso 2023).

Consolidated democracies without histories of military rule such as the United States can also find candidates from the security sector appealing. For instance, a majority of Congresspeople in the 1970s were veterans, as was the president (Teigen 2007). Police offers have also successfully run for office, notably Eric Adams in New York City. However, in a country like Brazil that has a difficult and largely unresolved history of military government and repression where the police played an important role, these candidacies have raised significant attention among the media and civil society (Pagnan and Rodrigues 2022; Passarinho 2022; Zanini 2022). These new candidacies lead directly to the question of how voters perceive a police officer on their ballots. What does the police officer “brand” mean to Brazilian voters? What advantages, disadvantages, strengths, and weaknesses does it imply?

The police officer candidate is a common occurrence in Brazil across all regions, and is becoming more frequent. This is true at all levels of government: federal, state, and municipal. See Figure 1 for plots of this growth for federal deputies and mayors and Figure 2 for examples of campaign material from these candidates.² The rise of police candidacies at the municipal level (Figure 1(b)) is particularly striking because the Brazilian constitution gives municipalities a very minor role in the provision of public security. The police forces of Brazil are commanded at the state and federal levels—by the governor for Military Police (PM) and Civil Police (PC) and the president for Federal Police (PF). Mayoral and city council candidates cannot credibly promise very much action in the realm of public security. For this reason, focusing on municipal level candidates is of particular theoretical interest, as respondents will have to focus at least partly on the non-crime fighting characteristics of the candidate, revealing more clearly how police officers are perceived along different dimensions.

Figure 1.

(a) Number of police officer candidates for federal deputy nationally, 2002–2022. (b): Number of police officer candidates for mayor nationally, 2000–2020.

Figure 2.

Two examples of campaign materials for police candidates, all three from the Military Police of São Paulo (PMSP) and using their rank and uniform.

Penal Populism

Traditional approaches to voter choice assume that voters select their candidates based on policy positions, selecting the candidate that is closest to them in the policy space. However, scholars have long recognized that there are simply too many candidates and too many issues for voters to reasonably process, especially in relatively low-information local elections. As a result, voters frequently rely on heuristics to choose from groups of relatively unknown candidates. Previous studies have uncovered several heuristics used to infer ideology or competence in low-information settings such as party, race (Krupnikov, Piston and Bauer 2016; Weaver 2012), gender (Teele, Kalla, and Rosenbluth 2018), religion (Berinsky and Mendelberg 2005; Calfano and Djupe 2009; Campbell 2008), and social class (Carnes and Lupu 2016; Carnes and Sadin 2015; McDonald, Karol, and Mason 2020).

One such heuristic is occupation. McDermott (2005) provides an early exploration of candidate occupation as a voting heuristic, demonstrating that voters infer candidate qualifications from their occupational cues. Voter abstention decreases when occupational cues are shared, and voters prefer candidates with such a professional cue. Atkeson and Hamel (2020) expand upon this finding, distinguishing incumbency effects from occupational effects and specifying that voters prefer candidates with what they perceive to be the most relevant job experience for the office. In Brazil, Boas (2014) demonstrated that using professional titles in ballot names is a meaningful and important heuristic, with candidates listed as “Pastor” performing better among evangelical Christians than otherwise identical candidates.

There has been some experimental research on military or police experience as a voting heuristic, mostly in the United States. Teigen (2013), for example, explores the impact of veteran status on vote choice and perceived issue competencies in US congressional elections. The findings of this study conform more or less to the framework advanced in Atkeson and Hamel (2020). That is, the strongest effect of veteran status was on voters’ perceptions of competency with national security or foreign policy issues. On aggregate, voters did not give veteran candidates an across the board boost, but rather gave favor in specific areas where the candidate’s professional experience was most likely to be a clear benefit.

Theory

Police as Symbols

Policing varies dramatically around the world, drawing on traditions as diverse as the largely unarmed London Metropolitan Police, the repressive Royal Irish Constabulary, and the militarized French Gendarmerie (Mawby 1999). However, the police consistently represent one of the most visible faces of the state apparatus (González 2017) and the symbols of policing such as the badge, the patrol car, and the uniform are powerful cultural symbols that transmit information beyond crime fighting. These symbols and associations are likely to influence how voters evaluate police officer candidates independent of their other features.

Research on the police has found that citizens associate the police with the status quo and dominant social groups, reading repressive police behaviors as signals from those groups Bradford, Murphy, and Jackson (2014). Policing therefore has a difficult relationship with liberal democratic ideals. The police often understand their role as Arthur’s knights in The Once and Future King, as “Might used for Right” (White 1958, 253). However, just as in that novel, the contradictions inherent in supporting a liberal order through violence and coercion generate weak points in the regime. Contact with the criminal justice system has been found to have negative impacts on democratic attitudes and behaviors from voting (Lerman and Weaver 2014; Morris 2021) to civic activism (Bateson 2012; Weaver, Prowse and Piston 2020) to general feelings of citizenship (Soss and Weaver 2017; Tyler, Fagan, and Geller 2014).

Given Brazil’s difficult history of military rule in which the police played an important role (Costa and Thompson 2011), a largely unreformed post-transition security sector, dramatic levels of police violence (Magaloni, Franco-Vivanco, and Melo 2020; Willis 2015) and police officer deaths in the line of duty, the public is likely to be skeptical of a police officer candidate’s commitment to democracy.

This negative relationship with democracy contrasts directly with an expected positive relationship with another concept central to modern democratic statehood, the rule of law. I suggest that voters are drawn to law enforcement professions because they draw a connection between a career spent enforcing the rules of the state and anti-corruption policies. That is to say a society that follows the law and a government that respects the formal rules of governing.

While modern scholars consider the rule of law to be an integral part of democratic governance, some even suggesting that the rule of law cannot exist outside of democracy (O’Donnell 1993), others have separated the concepts. Garrido (2021), for instance, outlines a similar divergence between citizen perceptions of democracy and the rule of law to explain the meteoric rise of Rodrigo Duterte in the Philippines. For these largely middle class voters, democracy had led to corruption and disorder in government and a strong, decisive leader like Duterte was the only one who could restore order. However, this tension may not exist if threats to the rule of law from corruption, rampant crime, or weak leadership is seen as the main threat to democracy. In such a case, profiles such as the police and military generally perceived as undemocratic may be seen as democracy’s best defenders.

This idea of “democracy as disorder” and the need of a strong, decisive leader to correct it therefore leads directly to the concept of leadership. Strong leadership is not necessarily tied to authoritarian attitudes, but in this case individuals skeptical of democracy are searching for leadership qualities. Influences from the media (Wolf 2012) and police public relations materials (Bonner 2013) emphasize the values that police training supposedly brings, among them the ability to effectively receive and follow orders, a necessity to operate within a strictly hierarchical system. Given the pervasiveness of true crime shows and cop dramas reinforcing these ideas in the media (Jewkes 2011), I expect for voters to transfer associations with values emphasized by police training, including leadership, to candidates with a professional background in the police forces. If affiliation with the police does indeed provide a comparative advantage on the issue of corruption, this could help to explain why these candidates were particularly well able to capitalize on the post-Lava Jato electoral environment, as they would not be perceived as “equally corrupt” (Pavão 2018).

Together, these associations based in the cultural significance of the police officer create a recognizable brand that defends social conservatism, and is comparatively strong on security, social order, the rule of law, and rooting out corruption, even if some measure of democratic norms are sacrificed. Figure 3 summarizes these expected qualities.

Figure 3.

Aggregate changes in perceptions for a candidate with security sector connections.

Empirical Expectations

Given the theoretical framework presented above, I derive several hypotheses regarding voter perceptions of security sector candidates.

First, I expect for voters to perceive these candidates as more competent on questions of public security. This expectation follows the literature on professional heuristics which demonstrates that issue areas closely related to a candidate’s professional experience such as education for teachers (Atkeson and Hamel 2020) or national security for veterans (Teigen 2013). Therefore:

H1a

Respondents will rate candidates with security sector connections as more competent on security-related issues.

H1b

Respondents will rate candidates with security sector connections as equally competent on non security-related issues (education, healthcare, and unemployment).

Relatedly, I expect for the professional experience of working within a strict hierarchy receiving and giving orders in high stress environments to lead voters to believe that police officer candidates are generally strong leaders.

Respondents will rate candidates with security sector connections as strong leaders.

Next, since police are frequently seen as defenders of the status quo, and traditionally powerful groups and values, I expect for police officers to be associated with the ideological right, independent of the policies they adopt. This could suggest that a center-seeking strategy based around security sector candidacies and “iron fist” crime fighting policies, as described in Holland (2013), could indeed be a winning strategy for left-wing parties. Conversely, it could suggest that voters are confused by these contradictory candidates.

Respondents will rate candidates with security sector connections on the ideological right.

Next, given Brazil’s troubled history with military government and state repression, intimately intertwined with the police forces (Ahnen 2007; Costa and Thompson 2011; Willis 2015), I expect for security sector connections to strongly impact perceptions of the candidate’s commitment to democratic norms and ideals. Specifically:

Respondents will rate candidates with security sector connections as less committed to democratic principles.

Finally, while democracy and the rule of law are closely interrelated, I expect for voters to perceive police officer candidates to be more respectful of the rule of law. This is because even if an individual voter distrusts the police, a professional history of enforcing the law should theoretically place them ahead of other candidates who have no specific claim to being able to maintain the rules. A more specific implication of this expectation is that police officer candidates will be better able to combat corruption.

H5a

Respondents will rate candidates with security sector connections as more respectful of the rule of law.

H5b

Respondents will rate candidates with security sector connections as better able to combat corruption.

The combination of voters perceiving security sector candidates as more capable on questions of safety and crime but less respectful of democratic values and norms suggests that there exists a trade-off in the minds of voters between these two values. That is that symbols and indicators of safety and effective crime fighting simultaneously suggest weaker democratic values.

Hypotheses 1a, 1b, 3, and 4 were pre-registered on the Evidence in Governance and Politics (EGAP) platform, while hypotheses 2, 5a, and 5b were included post-registration as helpful extensions of the framework at the recommendation of workshop and conference presentation discussion.³

Experimental Methodology

While observational methods can shed light onto where security sector candidates run, how well they perform in these elections, and what contextual features correlate with emergence and performance, there remain several unanswered questions that cannot be satisfactorily answered through observational methods. These questions probe at the psychology of voting and the internal decision processes of voters. Experimental methods that randomly assign various stimuli to respondents are well suited to capture these individual-level mental associations between candidate brands and outcomes of interest.

I therefore field a candidate choice experiment. The experiment follows the design of the conjoint experiment, popularly introduced to the field of political science in Hainmueller, Hopkins, and Yamamoto (2014), though the conjoint’s history in marketing and sociology research extends quite a bit further back (Green, Krieger, and Wind 2001; Green and Rao 1971). The basic setup of the conjoint experiment is to present experimental subjects with two profiles of products, or candidates, and ask them to select between the two. The products have several features, usually seven or fewer, which randomly vary between levels. In this way, the researcher can precisely and accurately estimate the independent effect that each feature has on a product’s desirability, and also how features interact with one another.

This experiment presents respondents with two different candidates varying along several different features. See Table 1 for a description of the levels between which each feature will vary. To see the exact operationalization of each level, see the Appendix. The feature of most theoretical interest is Occupation, which signals police connections. As noted in the previous section, professional experience is very important for Brazilian candidates and candidates frequently signal these connections in campaign materials through their name, employing ranks such as Coronel, Tenente (Lieutenant), or Delegado (Sheriff), or using police symbols such as their uniform.⁴

Table 1.

Summary of Conjoint Levels.

Feature	Levels
Demographic features
Race	Black, White, Pardo
Gender	Male, Female
Professional features
Political experience	Served in state legislature, political novice
Occupation	Police officer, Military officer, Doctor, Professor, Business owner
Policy features
Economic policy	Lower taxes, Expand welfare
Security policy	Deterrence, Strain, Urbanist
Other
Slogan	Anti-crime, Unemployment, Corruption, Generic (see Appendix for slogan text)
Issue emphasis	Crime, Employment, Corruption, Inflation [Blocked with Slogan]
Party	(See Appendix for unregistered numbers)
“Control”
Name	(See Appendix for full list)
Santinho Design	(See Appendix for examples)
Actor	(see Appendix to see actors)

While party ID does not play an important role in the theoretical framework, I include party in the experimental design to increase realism. While parties are not identified by name, the stimuli have two-digit codes which correspond with parties in the Brazilian system. Mayoral candidates in Brazil are represented by this two-digit number and campaign heavily with that number so that voters know what to type into the electronic voting machine. I therefore include two-digit codes in the stimulus to make them even more believable as campaign propaganda. Since the experiment was conducted during a hotly contested presidential election, I omitted 13 and 22, the codes of the Workers’ Party (PT) and Liberal Party (PL), the parties of Lula and Bolsonaro, respectively. This choice ensures that respondents were reacting to the candidate profiles and not simply association with national-level figures.

In a highly fragmented party system such as Brazil’s with 16 parties represented in the lower house and upwards of 30 regularly competing in elections, voters would not find an unrecognized two-digit code to be unusual. However, they would recognize immediately that this candidate is not a PT or PL partisan. To make sure respondents don’t assume a right-wing or left-wing ideology for the unknown parties, I include left-wing and right-wing policy signals as an independent feature described in the next paragraph.

In order to break the correlation between security sector connections and certain policy dimensions, I include as separate levels both Security Policy and Economic Policy. Including these as separate dimensions breaks the correlation between left-wing economic policies and punitive security policies, allowing to estimate the independent effect of each. While observational data does suggest that police candidates are more right-wing on average, the left-wing cop is not a mirage. In fact, just under 10% of police officer mayoral candidates ran with the leftist PSOL, PC do B, or PDT parties in the 2020 municipal elections.

Since the Brazilian constitution carves out a relatively small role for municipal governments in the realm of public security, the security policy signals given by the fictional mayoral candidates have to be carefully chosen to maintain realism. It might arouse suspicion or confusion in some respondents if the candidates made proposals about some of the most pressing hot-button criminal justice issues that remain squarely in the jurisdictions of the state or federal governments such as lowering the age of criminal responsibility, reintroducing the death penalty, or reductions to prison furloughs, so-called saidinhas. Instead, I selected three policy signals drawn from real mayoral platforms that include a “punitive” approach, forming an armed municipal guard force,⁵ as well as two alternative policies: expanding social assistance and educative programs and installing a network of surveillance cameras in public areas.

The information in this experiment will be presented with both images and an accompanying small vignette reinforcing the information. While most candidate choice experiments have conveyed information to respondents in table or vignette form (e.g., Carnes and Lupu 2016; Hainmueller and Hiscox 2006; Horiuchi, Smith, and Yamamoto 2020; Kirkland and Coppock 2018; Mummolo, Peterson, and Westwood 2019; Teele, Kalla, and Rosenbluth 2018), this experiment presents the relevant information in a visual medium to approximate the ways in which voters actually absorb political information. Examples of conjoint experiments attempting to mimic political advertising include Poertner (2020), who presents respondents with handbills, Krupnikov, Piston and Bauer (2016), who creates fictional campaign websites, and Teigen (2013) who uses television ads from a real candidate modified to emphasize or omit their veteran status. Flores-Macías and Zarkin (2021) also conducts an image-based conjoint experiment to better understand perceptions of militarized policing in Latin America.

This experiment presents voters with information in the form of the santinho, handbills commonly distributed during elections in Brazil. This form of political advertising will be familiar and understandable to respondents, as they might be offered dozens over the course of a single day during the height of election season. Apart from intelligibility, an advantage of the santinho is that its visual nature allows for a very high density of information. Nearly all demographic information about the candidate (race, gender, occupation) will be absorbed nearly instantly and simultaneously by the respondent simply by looking at the candidate photo. See Figure 4 for two examples of possible santinhos. To further aid the transmission of information, short vignettes repeating information contained in the santinho will be displayed underneath the images. See below for further discussion of the image-based nature of the experiment. Figure 5 shows a full representation of how the santinhos are presented to the respondents during the experiment.

Figure 4.

An example of two possible santinhos identical in all features except professional signal. The left model is dressed as a police officer, and the right as a teacher.

Figure 5.

Example of image-based stimulus with accompanying vignette.

To match the theoretical framework, the experiment will measure several outcomes to analyze the perception process and how perceptions of candidates interact with preferences to predict vote choice. Therefore, while respondents will be asked to select between pairs of candidates to support in a hypothetical election, they will be asked separately to select between pairs of candidates along other theoretically interesting dimensions.

Sampling

Sampling was carried out in July and September 2022 by survey firm Netquest. This firm, specializing in online surveys, maintains a panel of over 500,000 respondents in Brazil. From this panel, a quota sample of 3098 complete responses were drawn. Sampling took place in two waves: a preliminary wave in July (n = 1,598) followed by a supplemental sample in September (n = 1,500).⁶ Experimental results from both samples were pooled for the analysis, although analyzing the results from each wave individually makes little impact on the substantive conclusions.

Online samples have become a very common feature of the survey experimental literature in recent years as the cost and difficulty of traditional face-to-face and telephone surveys has increased dramatically, especially during the COVID-19 pandemic. The advantages of online sampling are not only cost and convenience, but also the ease of including relatively complex treatments. The image-based conjoint design of this article, for example, was easily executed through the online Qualtrics platform, but would be logistically more difficult with an in-person survey.

The representativeness of a survey sample is not necessarily a major concern for experimental research. Due to the randomization process, demographic features do not correlate systematically with treatment conditions, allowing for unbiased estimation of the sample average treatment effect (Franco et al. 2017; Imbens 2010). Research investigating the utility of convenience samples has also found that concerns about generalizability may be overblown in certain situations, as Mullinix et al. (2015) demonstrated that estimates from many influential studies based on convenience samples could be consistently replicated with population-based samples.

The Appendix contains a complete description of the sample. While certain demographic characteristics such as region, age, and gender were quite close to population proportions, the sample was indeed wealthier more highly educated than the population on average. While this unrepresentativeness could present some challenges for extrapolating the estimated effects to the Brazilian public as a whole, I do not believe they threaten the analysis presented in this paper for the reasons presented above.

Estimation

The main quantity of interest estimated from the experimental data will be the average marginal component effect (AMCE) which represents the marginal effect of any given attribute averaged over the joint distribution of the remaining attributes on each outcome of theoretical interest (Hainmueller, Hopkins, and Yamamoto 2014). That is to say, the expected change in the outcome variable for a change from a pre-defined reference level of a certain feature to another level.

Since profiles are generated under full randomization, the AMCE can be estimated using a non-parametric difference in means estimator as outlined in Hainmueller, Hopkins, and Yamamoto (2014). That estimator, for attribute l from N respondents rating J profiles in K selection tasks is:

\hat{π_{l}} (t_{1}, t_{0}, p (t)) = \frac{\sum_{i = 1}^{N} \sum_{j = 1}^{J} \sum_{k = 1}^{K} Y_{i j k} 1 {T_{i j k} = t_{1}}}{n_{1}} - \frac{\sum_{i = 1}^{N} \sum_{j = 1}^{J} \sum_{k = 1}^{K} Y_{i j k} 1 {T_{i j k} = t_{0}}}{n_{0}}

where n_d is the number of profiles, T_ijkl = t_d, Y_ijk is the outcome for the ijk^th profile, and 1{} is the indicator function. These values can be most efficiently estimated with OLS regression, regressing the binary choice outcome onto dummy variables representing the treatment conditions. Standard errors are clustered at the respondent level to account for the multiple per-respondent observations. The AMCE estimates in this paper are produced with this estimation procedure.

While there has been some debate about the appropriate estimator and interpretations of these estimators, the AMCE retains its causal interpretation for mean preferences of the sample. However, when exploring subgroup preferences, I will instead employ the marginal mean (MM) as recommended by Leeper, Hobolt, and Tilley (2020). The marginal mean is defined for a single feature-level as the average outcome value of all profiles with that feature-level. In a forced-choice setup like this one, a value of 0.5 indicates that there was no systematic preference for that feature-level. The MM therefore does not have the same straightforward causal interpretation as the AMCE, but is a more robust estimator for comparing subgroup preferences because it has no reference category, allowing each estimate to be interpreted in isolation and to be directly comparable to others. For full transparency, marginal mean estimations will also be offered in the Appendix for all outcomes.

Results

Experimental results fall in line with theoretical expectations, with some caveats. Respondents perceived police candidates as more conservative, stronger leaders, more likely to support the rule of law and combat corruption, and better able to promote public security within the municipality. See Figure 6 for a summary of these results. The magnitudes of these effects are in line with canonical experimental results as well as “control” features included in this conjoint such as race, gender, and political experience. For example, the “police” effect for anti-corruption competency was 0.104, indicating an estimated 10% increase in the probability that the profile was chosen as more competent to deal with corruption, all else equal. The well-explored “female” anti-corruption effect (Barnes and Beaulieu 2019; Watson and Moreland 2014), by contrast, was estimated to be 0.056. In fact, the professional signals consistently generated the largest effects across outcomes. Plots of effects for all features and levels across all outcome measures can be found in the Appendix for reference.

Figure 6.

Summary of results for all hypotheses. 95% confidence intervals plotted.

The one outcome with unexpected findings was when respondents were asked about each candidate’s commitment to democracy. Respondents were on average neutral on the subject with regards to the security sector professional signals. While punitive crime fighting policies received a significant penalty on this dimension, professional affiliation with the police or military did not, suggesting that voters do not perceive connections to the security sector as inherently undemocratic, and they are able to parse out “tough on crime” policies from the candidates themselves.

This divergence between policy and professional affiliation suggests that the iron fist candidate and the police candidate are distinct profiles, and treating them as interchangeable captures a limited view of what it means to be a police candidate in the eyes of voters. In fact, the respondents in this experiment reacted negatively to the iron fist policy cue when asked about the public security dimension and positively to the police and military professional cue. If professional connections were simply heuristics for crime fighting policies, we would expect for the direction of these effects to be the same. The identity of the police candidate is therefore complex and multidimensional, and cannot be fully understood as simply a “tough-on-crime” brand.

While reducing crime and defending citizens from violence is part of this brand, it is only a piece. Other important pieces include defending the law more generally, reducing corruption in government, and delivering strong leadership. While the theoretical framework presented earlier, and in the pre-analysis plan, suggested that these characteristics were in tension with the abstract idea of democracy, I find little evidence that the average Brazilian voter necessarily believes this. In fact, some respondents suggested that they went together. For example, one respondent explained that the police candidate would be more respectful of democracy as so: “a delegado would know how to strictly follow the rules of the law without going outside the lines of the constitution.”⁷ Somewhat humorously, another suggested that “he already knows the law, if he errs, he will do so knowingly.”⁸ Finally, and perhaps most suggestively, one respondent explained choosing the police candidate on the democracy question because they are “the representative closest to the people, a professional who has the greatest contact with the reality and necessities of the people.”⁹

The results presented in this paper therefore suggest that we should not necessarily understand the rise of Bolsonarismo, police officer participation in the electoral process, or the popularity of that profile as necessarily driven by authoritarian impulses or democratic skepticism. Instead, this profile is an answer to specific demands and perceived needs and demands within democracy. In a context such as Brazil that faces acute security and corruption challenges, has recent history of military government, and a culture of respect for the armed forces among a large segment of the population, electing politicians in uniform could simply be a natural outcome of voter preferences without intention to subvert the democratic process. In fact, the characteristics of knowing the law, professional experience enforcing it, and being in regular contact with citizens and their day-to-day needs means that for at least a segment of the population these candidates are defenders of democracy.

In the following subsections, I explore the “police effect” on each outcome implied by my theoretical framework.

Policy Competencies

The first set of outcomes that I explore are issue competencies. The theoretical framework predicted that the police and military candidates should be perceived as more competent on public security, but not significantly different on other dimensions. Results suggest that respondents have a strong preference for candidates with security sector experience when asked about public security but are actually skeptical about competencies in other issue areas, giving penalties to police and military candidates in the fields of education, healthcare, and employment.

Figure 7 plots the estimated average marginal component effect (AMCE) for the “police” level across four different policy areas. The AMCE is an estimate of the causal effect of moving from the reference category, in this case “businessperson,” to the feature-level of interest, in this case “police.” As discussed in previous sections, the AMCE is the preferred estimator for average effects because it has a clear causal interpretation.

Figure 7.

Estimated AMCE for Police candidate in different policy areas. 95% confidence intervals plotted.

Results between the estimators tell a similar story, that respondents thought that police officers would be effective at providing security, but were skeptical of their abilities in other policy areas. This finding reconfirms the conventional wisdom that respondents perceive candidates with security sector backgrounds as competent in policy areas where they have direct professional experience.

However, this finding should not be interpreted as demonstrating that the respondents in the sample had, on average, punitive preferences. When looking at the AMCEs for criminal justice policy on public security effectiveness, we actually observe that candidates were penalized for endorsing “tough on crime” policies. Interestingly, both the preventative, social services-based and the punitive approaches to crime were not preferred by respondents relative to the video monitoring approach, suggesting that respondents were somewhat skeptical of the two approaches to crime reduction commonly explored in the literature and are receptive to new, technical approaches that privilege urban design and careful maintenance of public spaces. This outcome suggests a fruitful avenue for future research unpacking the policy preferences of voters independent of candidate choice.

If respondents perceived professional connections to the police as simply a stand-in for punitivism, respondents should have responded similarly to the police professional cue and the punitive security proposal cue. Instead, respondents gave police candidates a boost on this dimension and penalized punitive security proposals, suggesting that respondents processed the police cue as separate from the public security policy cue. See Figure 8 for a plot of these estimates.

Figure 8.

Estiamted AMCE for police candidates and security policies on security effectiveness. 95% confidence intervals plotted.

Ideology

While police candidates in Brazil do not always run with right-wing parties, there has been a shift toward running with such parties in recent elections (Berlatto, Codato, and Bolognesi 2016). Correspondingly, respondents indeed perceived police candidates as more right-wing, all else equal. However, the magnitude of this effect measured in this experiment was unexpected. According to the model results, respondents perceived police and military candidates as uniquely right-wing, with a right-wing effect significantly larger than any other profession, including the pastor and the businessman, both professions that are commonly associated with the right due to evangelical churches’ embrace of Bolsonarismo and business owners’ economic interests. See Figure 9 for plots of these results.

Figure 9.

Estiamted AMCE for all professions on ideology, 1-7 left-right scale. 95% confidence intervals plotted.

While we can conclude without concern that respondents indeed saw police and military officers as strongly right-wing, the fact that the magnitude of the right-wing effect dwarfed all others might be an artifact of the timing of the experiment. Since the experiment was fielded during the “pre-campaign,” respondents would have been well aware of the discourse of the 2022 presidential election, and the most prominent allies lined up on each side. Given the strong dose of attention the news media paid to police candidates, especially on the right, it is possible that the right-wing signal was exaggerated among respondents during treatment assignment.

However, it is also possible that police and military officers truly are a uniquely powerful right-wing symbol. As developed above, the police as an institution are fundamentally conservative in nature, defending the status quo with the most extreme measure, violence. While a businessman might be viewed as leaning right, the very nature of a police officer’s job might reinforce a rightist reputation.

Democracy, the Rule of Law, and Corruption

One of the most critical pieces of the theoretical framework presented in this paper is the conceptualization of democracy versus what voters understand to be the rule of law. I suggest that the rise of police officer candidates is partly due to the salience of precisely this trade-off in Brazilian political culture. I suggest that police candidates fit comfortably in the Bolsonarista movement and the post-Lava Jato era not because of policy affinity between the Bolsonaro administration and the police¹⁰ but rather because police symbols are powerful symbols for order, respect for the laws, and anti-corruption.

To probe to what extent respondents associated this brand of legality and questionable commitment to democracy, I asked respondents to evaluate the candidate pairs along four separate dimensions. To measure associations with democratic ideals, I asked which candidate would govern more in line with their idea of democracy. I also asked which candidate would be a stronger leader. While strong leaders can of course be committed democrats, the concept of a strong leader is frequently contrasted with democratic governance in survey questions (Latin American Public Opinion Project 2019; World Values Survey 1998).

As shown in Figure 10, results suggest that voters do not penalize police candidates on the democratic dimension, although police candidates do receive a small boost on the leadership dimension. This result suggests that respondents did associate police training and experience on the force as at least somewhat effective leadership training, but this kind of leadership was not necessarily associated with authoritarian, anti-democratic tendencies. Once again, it is instructive to compare how respondents reacted to the police cue and the punitive security policy cue. While police candidates faced no penalty on respect for democracy, Figure 11 shows that the punitive crime reduction signal corresponded with a 0.06 point drop in the probability of being chosen as more likely to respect democracy. This result suggests that, contrary to the theoretical expectation, voters do not necessarily associate police and military personnel with anti-democratic ideas, despite a history of military dictatorship. However, “tough on crime” policies do negatively impact democratic evaluations. Just as with the public security outcome, we can see that respondents can and do differentiate the police signal from the punitive signal, demonstrating that there is significantly more to the police identity than simply an anti-crime heuristic.

Figure 10.

Estiamted AMCE for police candidates on rule of law, corruption, leadership, and respect for democracy. 95% confidence intervals plotted.

Figure 11.

Estiamted AMCE for police and security policy on respect for democracy. 95% confidence intervalse plotted.

To investigate connections to order and the rule of law, I asked for two separate evaluations. First, I asked directly which candidate would be more respectful of the rule of law and second which candidate would be more effective at combating corruption in the municipal government. My theoretical framework suggests that voters perceive law enforcement officers as qualified on all questions of law and order, even though the typical police officer has little to no experience in anti-corruption operations.

Respondents did give police officers a clear advantage when asked about the rule of law and corruption, with the police officer professional cue corresponding to a 0.07 and 0.09 point increase relative to businessperson respectively. These two results suggest that the police officer brand goes beyond public security, extending to all issues that can be considered about “law and order.” This finding might seem counterintuitive to observers of Brazilian and Latin American politics, where the police are frequently considered to be a corrupt, unaccountable force (Willis 2015). However, the values popularly associated with a career in the police such as respect for authority and their professional history enforcing the law mean that the electoral brand can be, on average, associated with combating corruption and defending the system of legality. See Figure 11 for plots of these results.

However, on these features there is important heterogeneity among respondents, suggesting that what is considered “rule of law,” and its relationship with the police, might be understood differently by distinct ideological groups. First, I broke the 7-point ideological scale into three sections to represent left, right, and center. I then estimated marginal means for each of these groups (Leeper, Hobolt, and Tilley 2020). See Figure 12 for plots of the marginal means across the self-identified left, right, and center groups. These estimates suggest that the right-wing respondents responded most favorably to the police signal, and most strongly thought that the rule of law and anti-corruption would be served by such a candidate.

Figure 12.

Estimated Marginal Means for police on corruption and rule of law, conditional on respondent ideology. 95% confidence intervalse plotted.

On the question of corruption, centrist respondents aligned more closely with right-wing respondents, while on the more abstract rule of law question they were closer to the leftist respondents, giving police candidates no significant advantage on anti-corruption competency. Divergence across self-identified ideological groups could be the result of the timing of the experiment, being in the midst of a hard-fought presidential campaign where one side, that of right-wing Jair Bolsonaro, heavily emphasized corruption as an election issue and continuously hammered leftist candidate Lula for his previous corruption conviction. Future research could investigate the extent to which left-wing and right-wing voters truly disagree on corruption competencies outside of these circumstances (Barabas, 2010; Bradford, 2014; Pearce, 2010).

Discussion

With recent increases in the number of police officers running for office in a democratic context, it is important to understand not only what type of voter supports these candidates, but more fundamentally how do voters understand the choice they are presented with when a police officer is on the ballot. This paper developed a theoretical framework to understand this more fundamental question of branding, developing empirical expectations about these candidacies based on the complex role a police officer plays in society, and presented novel experimental evidence in support. Results of the experiment suggest that voters perceive police officer candidates as right-wing, strong leaders, and effective at promoting public security, defending the rule of law, and combating corruption. While the theoretical framework expected for police officer candidates to face a penalty on respect for democracy, no evidence was found to suggest that the police officer signal alone enough to constitute an anti-democratic brand.

This combination of reactions from survey respondents suggests that there is significantly more wrapped up in the symbols of the police than simply a punitive approach to fighting crime. Voters perceive support for traditional values, leadership skills, and, most importantly, a candidate with unique abilities to defend the rule of law and “clean up” society, whether that means reducing violent crime or fighting corruption in government. Voters who place value on these dimensions will be more likely to find police candidates attractive, regardless of their policy preferences on public security.

On several dimensions, there was clear divergence between the way that respondents reacted to the police signal and the punitive crime fighting policy signal. Most strikingly, respondents reacted positively to police candidates on the ability to effectively provide public security but negatively to the punitive policy signal. This divergence indicates that the professional experience that police gain from enforcing the law gives them credibility far beyond the punitive public security policies they are commonly associated with. In fact, a more consistent explanation for this trust on the public security dimension ties directly to the rule of law and anti-corruption finding. The police officer brand is strongly associated with enforcing the law, and this skill applies in the mind of the average Brazilian voter just as well to violent crime as it does to white collar crime such as government corruption.

The implications of these findings are twofold. First, the phenomenon of police officer candidacies cannot be simply reduced to a reaction to crime. While police officer candidates are trusted more strongly on questions of public security, they also hold advantages relative to other common candidate profiles in other areas. Support for police officers in government is therefore also about ideology more generally and preferences for a certain style of government that emphasizes decisive leadership and strict interpretation and enforcement of laws. This implies that the rise of police officer candidates in Brazil signifies a more fundamental shift in the structure of electoral competition and not simply a rise in the salience of public security as an election issue.

Second, the absence of a negative relationship between police officer status and respect for democracy implies that voters do not perceive a trade-off between this candidate profile and democracy. This is a reassuring sign, as voters do not seem to be deliberately sacrificing democracy for order. In fact, many respondents saw these two concepts as closely intertwined. As quotes from open-response questions in the Results section indicate, some voters thought that a professional background that gave police candidates knowledge of the laws, experience enforcing them, and proximity to “the people” would be a net positive for democracy. Other advantages in this area included abstract values: “I believe that a police officer has ethical and moral values,” shared one respondent.¹¹ Another expressed respect for the system of concursos (public service exams) that police officers have to pass: “since she is a police officer... it shows that she is intelligent (for having passed a concurso).”¹² Finally, one simply drew an equivalence between democracy and security, saying that the candidate was more democratic because of their “proposals focused on democracy, of high security and tranquility for the population.”¹³

Conclusion

The involvement of the wielders of coercion, the police and military, in democratic politics has long been a reality and a concern, especially in a region such as Latin America that has a dark history of police violence, military dictatorship, and brutal political repressions. Recent electoral success for police officer candidates in Brazil has therefore caused alarm, especially as some police engaged in anti-democratic acts during the 2022 federal elections and January 8 riot.

The results of this experiment expand our understanding of these candidacies in two major ways. First, voters are more than capable of distinguishing between “tough-on-crime” deterrence-based public security policies and the professional signal of being a cop. The fact that respondents treated these two signals differently suggests that while police officer candidates are frequently associated with mano dura policy proposals, their true appeal comes from the more general idea of order. Order sometimes goes along with being tough on crime, but not always. While there is much research on “tough-on-crime” candidates in Latin America, this study suggests that there is an important conceptual distinction with specifically the police officer candidates.

Second, the lack of evidence for a negative association between the police officer signal and expected respect for democracy indicates that voters do not on average perceive police officer candidates to be necessarily anti-democratic. Instead, the police officer profile is seen as a normal, possibly even desirable, profile within the framework of normal electoral politics. While this result presents no evidence about whether police officer politicians actually are anti-democratic, it does demonstrate that Brazilian voters are not thinking of democracy vs. security as a trade-off, at least in the context of candidate choice. Attempts from the PT to paint themselves as defenders of democracy or to proscribe police officer candidacies¹⁴ might therefore only appeal to their own partisans and not attract the kind of broad coalition that they hope for.

While the experiment was designed for and conducted within the context of the Brazilian case, these implications could very well travel to other cases where the security sector plays a role in electoral politics. As shown in cases as diverse as Mexico (Ventura, Ley, and Cantú 2023), the Philippines (Garrido 2021), and the United States (Ax 2021), police officers and public security professionals are an increasingly viable electoral profile. Ultimately, this study suggests that we need to update our understanding of candidates in uniform. While not what scholars and observers of democratic transitions might have envisioned, voters continue to demand this type of politician profile without turning against what they understand as the core principles of democracy.

Supplemental Material

Supplemental Material - The People’s Captain: Understanding Police Officers as an Electoral Brand

Supplemental Material for The People’s Captain: Understanding Police Officers as an Electoral Brand by Jacob R. Turner in Political Research Quarterly

Supplemental Material

Supplemental Material - The People’s Captain: Understanding Police Officers as an Electoral Brand

Supplemental Material for The People’s Captain: Understanding Police Officers as an Electoral Brand by Jacob R. Turner in Political Research Quarterly

Footnotes

Acknowledgments

I would like to thank the editors and the two anonymous reviewers. I would also like to thank Michael Coppedge, Luis Schiumerini, Aníbal Pérez-Liñán, Ian Rebouças Batista, Lucas Câmara, Luiz Vilaça, Abby Cordova and the Survey Research Methods Course, Juan Albarracín, the SIVIS Institute, and the panel participants at MPSA and APSA for their helpful comments on the research design, survey flow, and manuscript. I thank the Kellogg Institute, ISLA, and APSA for funding this project. Finally, I thank Lara Luz for the photography and graphic design that made this project possible.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication was made possible by a grant from the American Political Science Association, the Kellogg Institute for International Studies, and the Institute for Scholarship in the Liberal Arts. The statements made and views expressed are solely the responsibility of the author.

ORCID iD

Jacob R. Turner

Data Availability Statement

Replication files can be found on the following GitHub repository:

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Ahnen

Ronald E.

2007. “The Politics of Police Violence in Democratic Brazil.” Latin American Politics and Society 49 (1): 141–164.

Arias

Enrique Desmond

Barnes

Nicholas

. 2017. “Crime and Plural Orders in Rio de Janeiro, Brazil.” Current Sociology 65 (3):448–465.

Atencio

Rebecca J.

2014. Memory’s Turn: Reckoning with Dictatorship in Brazil. Madison: University of Wisconsin Press.

Atkeson

Lonna Rae

Hamel

Brian T.

. 2020. “Fit for the Job: Candidate Qualifications and Vote Choice in Low Information Elections.” Political Behavior 42 (1): 59–82.

Joseph

. 2021. “Democrat Eric Adams wins New York City Mayoral Election.” Reuters.

Azevedo

Victoria

. Bolsonaro não gosta de gente, gosta de policial, diz Lula. Folha de S.Paulo. 2022.

Barabas

Jason

Jerit

Jennifer

. 2010. “Are Survey Experiments Externally Valid?” American Political Science Review 104 (2): 226–242.

Barnes

Tiffany D.

Beaulieu

Emily

. 2019. “Women Politicians, Institutions, and Perceptions of Corruption.” Comparative Political Studies 52 (1): 134–167.

Bateson

Regina

. 2012. “Crime Victimization and Political Participation.” The American Political Science Review 106 (3): 570–587.

10.

Berinsky

Adam J.

Mendelberg

Tali

. 2005. “The Indirect Effects of Discredited Stereotypes in Judgments of Jewish Leaders.” American Journal of Political Science 49 (4): 845–864. doi:10.1111/j.1540-5907.2005.00159.x.

11.

Berlatto

Fabia

Codato

Adriano

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 21: 77–120.

12.

Bittner

Egon

. 1970. Functions of the Police in Modern Society. Chevy Chase, MD: National Institute of Mental Health.

13.

Boas

Taylor C.

2014. “Pastor Paulo vs. Doctor Carlos: Professional Titles as Voting Heuristics in Brazil.” Journal of Politics in Latin America 6 (2): 39–72.

14.

Bonner

Michelle D.

2013. “The Politics of Police Image in Chile.” Journal of Latin American Studies 45 (4): 669–694.

15.

Bradford

Ben

. 2014. “Policing and Social Identity: Procedural Justice, Inclusion and Cooperation between Police and Public.” Policing and Society 24 (1): 22–43. doi:10.1080/10439463.2012.724068.

16.

Bradford

Ben

Murphy

Kristina

Jackson

Jonathan

. 2014. “Officers As Mirrors Policing, Procedural Justice and the (Re)Production of Social Identity.” British Journal of Criminology 54 (4): 527–550.

17.

Bueno

Samira.

2014. “Bandido bom é bandido morto: a opção ideológico-institucional da política de segurança pública na manutenção de padrões de atuação violentos da polícia militar paulista.” Accepted 2014-03-19.

18.

Caldeira

Teresa P. R.

James

Holston

. 1999. “Democracy and Violence in Brazil.” Comparative Studies in Society and History 41 (4): 691–729.

19.

Caldeira

Teresa Pires do Rio

. City of Walls: Crime, Segregation, and Citizenship in São Paulo. Calif, Berkeley: University of California Press, 2000.

20.

Calfano

Brian Robert

Djupe

Paul A.

. 2009. “God Talk: Religious Cues and Electoral Support.” Political Research Quarterly 62 (2): 329–339.

21.

Camargo Isabela Falcao

Márcio

. 2022. “PRF descumpre ordem do TSE e para pelo menos 610 ônibus de eleitores em blitze; Moraes intima diretor-geral.” G1.

22.

Campbell

Liz

. 2008. “Criminal Justice and Penal Populism in Ireland.” Legal Studies 28(4):559–573.

23.

Carnes

Nicholas

Sadin

Meredith L.

. 2015. “The “Mill Worker’s Son” Heuristic: How Voters Perceive Politicians from Working-Class Families—and How They Really Behave in Office.” The Journal of Politics 77 (1): 285–298.

24.

Carnes

Nicholas

Lupu

Noam

. 2016. “Do Voters Dislike Working-Class Candidates? Voter Biases and the Descriptive Underrepresentation of the Working Class.” American Political Science Review 110 (4): 832–844.

25.

Clarín.com . 2021. Video: El sorpresivo saludo de los policías bonaerenses a Patricia Bullrich. Section: Clarín politica.

26.

Câmara dos

Deputados.

2021. Quarentena para militares, policiais e juízes é excluída do novo Código Eleitoral - Notícias. Portal da Câmara dos Deputados.

27.

Costa

Arthur Trindade Maranhão

Thompson

Timothy

. 2011. Police Brutality in Brazil: Authoritarian Legacy or Institutional Weakness? Latin American Perspectives 38 (5): 19–32.

28.

Davis

Darren W.

Silver

Brian D.

. 2004. “Civil Liberties vs. Security: Public Opinion in the Context of the Terrorist Attacks on America.” American Journal of Political Science 48 (1): 28–46.

29.

Flom

Hernán

. The Informal Regulation of Criminal Markets in Latin America. Cambridge: Cambridge University Press, 2022.

30.

Flores-Macías

Gustavo

Zarkin

Jessica

. Militarization and Perceptions of Law Enforcement in the Developing World: Evidence from a Conjoint Experiment in Mexico. British Journal of Political Science 2021; 52 (3): 1–21.

31.

Fórum Brasileiro de Segurança Pública . Anuário Brasileiro de Segurança Pública. Anuário Brasileiro de Segurança Pública. 2022.

32.

Franco

Annie

Malhotra

Neil

Simonovits

Gabor

Zigerell

L. J.

. 2017. “Developing Standards for Post-Hoc Weighting in Population-Based Survey Experiments.” Journal of Experimental Political Science 4 (2): 161–172.

33.

Garrido

Marco

. 2020. “A Conjunctural Account of Upper- and Middle-Class Support for Rodrigo Duterte.” International Sociology 35 (6): 651–673.

34.

Garrido

Marco

. 2021. “Democracy as Disorder: Institutionalized Sources of Democratic Ambivalence Among the Upper and Middle Class in Manila.” Social Forces 99 (3): 1036–1059.

35.

González

Yanilda María

. 2017. “What Citizens Can See of the State”: Police and the Construction of Democratic Citizenship in Latin America. Theoretical Criminology 21(4): 494–511.

36.

González

Yanilda María

. Authoritarian Police in Democracy: Contested Security in Latin America. Cambridge Studies in Comparative Politics Cambridge. Cambridge: Cambridge University Press.

37.

Green

Paul E.

Krieger

Abba M.

Wind

Yoram (Jerry)

. 2001. “Thirty Years of Conjoint Analysis: Reflections and Prospects.” Interfaces 31 (3): S56–S73.

38.

Green

Paul E.

Rao

Vithala R.

. 1971. “Conjoint Measurement for Quantifying Judgmental Data.” Journal of Marketing Research 8 (3): 355–363.

39.

Hainmueller

Jens

Hopkins

Daniel J.

Yamamoto

Teppei

. 2014. “Causal Inference in Conjoint Analysis: Understanding Multidimensional Choices via Stated Preference Experiments.” Political Analysis 22 (1): 1–30.

40.

Hainmueller

Jens

Hiscox

Michael J.

. 2006. “Learning to Love Globalization: Education and Individual Attitudes toward International Trade.” International Organization 60 (2): 469–498.

41.

Holland

Alisha C.

2013. ““Right on Crime? Conservative Party Politics and “Mano Dura” Policies in El Salvador.” Latin American Research Review 48 (1): 44–67.

42.

Horiuchi

Yusaku

Smith

Daniel M.

Yamamoto

Teppei

. 2020. “Identifying Voter Preferences for Politicians’ Personal Attributes: a Conjoint Experiment in Japan.” Political Science Research and Methods 8 (1): 75–91.

43.

Imbens

Guido W.

2010. “Better LATE Than Nothing: Some Comments on Deaton (2009) and Heckman and Urzua (2009).” Journal of Economic Literature 48 (2): 399–423.

44.

Jaffe

Rivke

. 2013. “The Hybrid State: Crime and Citizenship in Urban Jamaica.” American Ethnologist 40 (4): 734–748.

45.

Jewkes

Yvonne

. 2011. “Media & Crime: a Critical Introduction.” In Key Approaches to Criminology. 2nd ed. Los Angeles: Sage.

46.

Kirkland

Patricia A.

Coppock

Alexander

. 2018. “Candidate Choice Without Party Labels.” Political Behavior 40 (3): 571–591.

47.

Krupnikov

Yanna

Spencer

Piston

Bauer

Nichole M.

. 2016. “Saving Face: Identifying Voter Responses to Black Candidates and Female Candidates.” Political Psychology 37 (2): 253–273. doi:10.1111/pops.12261.

48.

Latin American Public Opinion Project . 2019. LAPOP Latin American Public Opinion Project. Nashville, TN.

49.

Leeper

Thomas J.

Hobolt

Sara B.

Tilley

James

. 2020. “Measuring Subgroup Preferences in Conjoint Experiments.” Political Analysis 28 (2): 207–221.

50.

Lerman

Amy E.

Weaver

Vesla M.

. 2014. Arresting Citizenship: The Democratic Consequences of American Crime Control. Chicago, USA: University of Chicago Press.

51.

Magaloni

Beatriz

Franco-Vivanco

Edgar

Melo

Vanessa

. 2020. “Killing in the Slums: Social Order, Criminal Governance, and Police Violence in Rio de Janeiro.” American Political Science Review 114 (2): 552–572.

52.

Mawby

R. I.

, ed. 1999. Policing across the World: Issues for the Twenty-First Century. London: Routledge.

53.

McDermott

M. L.

2005. “Candidate Occupations and Voter Information Shortcuts.” Journal of Politics 67 (1): 201–219.

54.

McDonald

Jared

Karol

David

Mason

Lilliana

. 2020. ““An Inherited Money Dude from Queens County”: How Unseen Candidate Characteristics Affect Voter Perceptions.” Political Behavior 42 (3): 915–938.

55.

Morris

Kevin

. 2021. “Welcome Home—Now Vote! Voting Rights Restoration and Postsupervision Participation.” Social Science Quarterly 102 (1): 140–153. doi:10.1111/ssqu.12901.

56.

Mullinix

Kevin J.

Leeper

Thomas J.

Druckman

James N.

Freese

Jeremy

. 2015. “The Generalizability of Survey Experiments.” Journal of Experimental Political Science 2 (2): 109–138.

57.

Mummolo

Jonathan

Peterson

Erik

Westwood

Sean

. 2019. “The Limits of Partisan Loyalty.” Political Behavior 43(3), 949–972.

58.

Nueva Rioja . Desafectaron a los policías que le hicieron la venia a Patricia Bullrich |Nueva Rioja. Nueva Rioja. 2021.

59.

O’Donnell

Guillermo

. 1993. “On the State, Democratization and Some Conceptual Problems: A Latin American View with Glances at Some Postcommunist Countries.” World Development 21 (8): 1355–1369.

60.

Pagnan

Rogério

Rodrigues

Artur

. 2022. PMs candidatos batem recorde em SP e miram voto bolsonarista. Folha de S.Paulo.

61.

Passarinho

Nathalia

. Como é ser policial militar de esquerda: ’Se é a favor de Bolsonaro, fala à vontade, se é contra, vai pagar por isso. BBC News Brasil. 2022.

62.

Pavão

Nara

. Corruption as the Only Option: The Limits to Electoral Accountability. The Journal of Politics. 2018; 80(3): 996–1010.

63.

Pearce

Jenny

. 2010. “Perverse State Formation and Securitized Democracy in Latin America.” Democratization 17 (2): 286–306. doi:10.1080/13510341003588716.

64.

Poertner

Mathias

. 2020. “The Organizational Voter: Support for New Parties in Young Democracies.” American Journal of Political Science. eprint. doi:10.1111/ajps.12546.

65.

Reynoso

Lucas.

2023. Claudia López, alcaldesa de Bogotá, se viste de militar: “¡Honren al Ejército!” | EL PAÍS América Colombia. El País.

66.

Shamir

Michal

Jacob

Shamir

. 1995. “Competing Values in Public Opinion: A Conjoint Analysis.” Political Behavior 17 (1): 107–133.

67.

Soss

Joe

Weaver

Vesla

. 2017. “Police Are Our Government: Politics, Political Science, and the Policing of Race–Class Subjugated Communities.” Annual Review of Political Science 20 (1): 565–591.

68.

Teele

Dawn Langan

Kalla

Joshua

Rosenbluth

Frances

. 2018. “The Ties that Double Bind: Social Roles and Women’s Underrepresentation in Politics.” American Political Science Review 112 (3): 525–541.

69.

Teigen

Jeremy M.

2007. “Veterans’ Party Identification, Candidate Affect, and Vote Choice in the 2004 U.S. Presidential Election.” Armed Forces & Society 33 (3): 414–437.

70.

Teigen

Jeremy M.

2013. “Military Experience in Elections and Perceptions of Issue Competence: An Experimental Study with Television Ads.” Armed Forces & Society 39(3):415–433.

71.

Teixeira

Isadora

Alcântara

Manoela

. 2023. Moraes diz que atos de 8/1 não ocorreriam “se PMDF tivesse atuado” | Metrópoles. Metrópoles. Section: Grande Angular.

72.

Trejo

Guillermo

Ley

Sandra

. 2018. “Why Did Drug Cartels Go to War in Mexico? Subnational Party Alternation, the Breakdown of Criminal Protection, and the Onset of Large-Scale Violence.” Comparative Political Studies 51 (7): 900–937.

73.

Tyler

Tom R.

Fagan

Jeffrey

Geller

Amanda

. 2014. “Street Stops and Police Legitimacy: Teachable Moments in Young Urban Men’s Legal Socialization.” Journal of Empirical Legal Studies 11 (4): 751–785. doi:10.1111/jels.12055.

74.

Ventura

Tiago

Ley

Sandra

Cantú

Francisco

. 2023. Voting for Law and Order: Evidence From a Survey Experiment in Mexico. Comparative Political Studies.

75.

Watson

David

Moreland

Amy

. 2014. “Perceptions of Corruption and the Dynamics of Women’s Representation.” Politics & Gender 10 (3): 392–412.

76.

Weaver

Vesla

Prowse

Gwen

Spencer

Piston

. 2020. “Withdrawing and Drawing in: Political Discourse in Policed Communities.” Journal of Race, Ethnicity, and Politics 5 (3): 604–647.

77.

Weaver

Vesla M.

2012. “The Electoral Consequences of Skin Color: The “Hidden” Side of Race in Politics.” Political Behavior 34 (1): 159–192.

78.

White

T. H.

1958. The once and Future King. New York: Putnam.

79.

Willis

Graham Denyer

. 2015. The Killing Consensus: Police, Organized Crime, and the Regulation of Life and Death in Urban Brazil. Oakland, California: University of California Press.

80.

Wolf

Sonja

. 2012. “Creating Folk Devils: Street Gang Representations in El Salvador’s Print Media.” Journal of Human Security 8 (2): 36–63.

81.

World Values Survey . 1998. “World Values Survey WVS.” Stockholm, Sweden.

82.

Zanini

Fábio

. 2022. Painel: Número ‘absurdo e exagerado’ de candidaturas de policiais preocupa parlamentares já eleitos em SP. Folha de S.Paulo.

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.

19.90 MB

0.00 MB

19.31 MB