Abstract
What are the effects of state repression on public perceptions of police? And to what extent are these effects uniform or conditional on individuals’ loyalty to political authorities? I argue that repression by the police negatively affects how people evaluate the police, especially among those who do not support the ruling party. People who oppose the regime are more likely to fear the police following a repressive event relative to regime supporters. To test this argument, I leverage a unique research design opportunity that emerges from the social media tax protest led by Robert Kyagulanyi Ssentamu (also known as Bobi Wine) and subsequent selective repression by the Uganda Police Force while a nationally representative survey on police and security was being administered in Uganda. I demonstrate selective repression of protesters decreased support for the police. These effects are largely driven by political loyalty; repression has a stronger effect on how members of the opposition evaluate the police relative to incumbent supporters.
Introduction
What are the effects of state repression on public perceptions of the institution that perpetrates it? 1 Does state repression by the police at the national-level affect individual-level support for the police? Repression by the government occurs across an array of contexts and political settings, making it one of the most common forms of violence against civilians. Existing studies focus on the effects of repression on political participation (Davenport et al. 2019; Ritter and Conrad 2016). 2 If state repression decreases support for police, it has implications for the ability of police to provide law and order and deter crime.
However, the political effects of repression are debated. On one hand, state repression is meant to punish political disloyalty, deter acts of dissent and induce obedience (Lichbach 1987; Lyall 2009; Ritter and Conrad 2016; Young 2019; Zhukov and Talibova 2018). On the other, state repression can incite political opposition, mobilizing future collective action (Balcells 2012; Curtice and Behlendorf 2021; Finkel 2015; Gurr 2015). The effectiveness of repression is shaped by governments’ repressive tactics: targeting clandestine activities might decrease dissent but using it against overt, collective challenges might escalate dissent (Sullivan 2016). Additionally, repression and dissent depend on a state’s infrastructural power. More developed infrastructural power might decrease reactive repression by police enabling police to preemptively limit dissent without shifting to excessive acts of state repression (Sullivan and Liu forthcoming). But we know less about the relationship between repression and public perception of the institution tasked with perpetrating it.
In many unconsolidated democracies and autocracies, security institutions like the police and military have two roles: first, they are responsible for providing security and law and order; and second, they repress dissent to maintain the political status quo. Internal security institutions used by political authorities to repress dissent and enforce the political status quo are also responsible for deterring crime, enforcing law and order and providing security more broadly. 3
For many countries, the police forces are the state agents most likely to use repression against civilians (Curtice and Behlendorf 2021; Davenport 2020). Examining how repression affects people’s support for the police is crucial to understanding the provision of law and order, especially when security institutions rely on cooperation from individuals. When political authorities rely on the police to repress dissent, it has implications beyond deterring dissent. As a tool used by political authorities to inflict fear and reduce dissent (Young 2019), repression likely increases people’s fear of the police. Relative to the military, police forces rely more heavily on people’s willingness to cooperate with them to be effective. The more people fear the police as repressive agents of the state, the less likely they are to support them. People who fear repression, for example, are likely to avoid interacting with security forces who employ it. Individuals refusing to voluntarily provide information, or even report crimes, makes routine acts of policing more difficult. People’s willingness to cooperate with police depends on public perceptions of the police as legitimate authorities.
However, the theoretical and empirical relationship between repression and its effects on public perception of police remains an open question. 4 Repression by police might have a uniform negative effect on the public’s perceptions of the police as legitimate authorities. Alternatively, people might view acts of repression through motivated reasoning, depending on whether they support the regime. People who support the regime might view repression as legitimate while people who oppose the regime see it as illegitimate. Then, repression might have a positive effect on assessments of incumbent supporters and a negative effect on evaluations by supporters.
I argue that when political authorities rely on the police to repress dissent, people are less likely to view the police as legitimate authorities. The extent to which people fear being repressed varies by whether they support or oppose the regime. People are more likely to fear the police in general; however, support for the ruling party should condition its effect. Incumbent supporters are less concerned about experiencing repression and consequently less likely to fear repression. However, opposition supporters are more likely to fear experiencing repression. The effects of repression on public perception should be strongest among those who are more likely to fear experiencing political violence from the state.
To test these claims, I empirically examine the effect of repression by the police on public perceptions of police in Uganda. I estimate the effect of selective repression by leveraging a unique research design opportunity that emerges from the social media tax protest led by Robert Kyagulanyi Ssentamu (also known as Bobi Wine) and his subsequent arrest by the Uganda Police Force while a nationally representative survey on police and security was being administered in Uganda. I find selective repression of these protesters by police decreased support for the police, even among those who did not directly experience the repression. Following the repression of Bobi Wine and the Social Media Tax protests, people viewed the police as less legitimate authorities. Additionally, these effects are largely driven by political loyalty and perceptions of normative alignment with the police. The effects are weakest among those who support the incumbent. 5
This paper makes several significant contributions to research on political violence (Kalyvas 2006; Lyall 2009), repression (Davenport et al. 2019; Ritter and Conrad 2016), and policing (Blair, Karim, and Morse 2019). First, these results demonstrate that the negative consequences of repression move beyond political behavior (Balcells 2012; Bautista 2015), rather repression by police undermines support for the state institutions responsible for providing security. The effects of repression on civilian-police interactions matter because police rely on individuals to be effective; higher support leads to more cooperation (i.e., better information being provided to the police), which leads to more effective security operations. Second, selective repression targeted at opposition political elites and protestors has individual-level effects even among those who were not engaged in political dissent. Selective violence might be an effective tool to induce political loyalty (Blaydes 2018; Kalyvas 2006), but it still adversely affects the way people view the police. Third, these results suggest that political authorities face a tradeoff: they can rely on the police to repress dissent but doing so undermines people’s trust in the police—even those who support the regime—and subsequently decreases the co-production of law and order and security.
Repression and Police Legitimacy
In many contexts, political authorities use members of the security apparatus to repress dissent. Several governments employ their police force to suppress and control opposition movements. The Chinese government relies on their police and non-state security institutions to repress Uighur culture and traditions (Ong 2015). Authoritarian governments in Burundi, Belarus, Ethiopia, Rwanda, Tanzania, and Zimbabwe have used their police to restrict public spaces associated with political dissent, intimidating opposition supporters during elections, arresting and even torturing political opposition leaders. Since 2013, police in Egypt have imprisoned thousands of Islamist opponents, as well as numerous liberal activists and journalists. In non-democracies where political authorities are not committed to the democratic transfer of power, people view actions taken by police through a broader political lens of the state. Acts of repression by members of the security apparatus should affect both people’s perceptions of those who ordered the repression and the institution responsible for implementing it.
However, most theoretical and empirical studies of repression focus on its political effects (i.e., how people view political authorities) rather than its effect on perception of those who implemented it. Existing studies have examined the relationship between repression and dissent (Davenport et al. 2019). 6 Others focus on the political effects of repression and exposure to violence (Blattman 2009; Bratton and Masunungure 2007; Gonzalez and Miguel 2015; Rozenas and Zhukov 2019; Zhukov and Talibova 2018). An important theoretical and empirical limitation of existing studies of the effects of state repression on political participation is the assumption that political authorities and the security apparatus act as a unitary actor, considering whether indiscriminate or mass repression affect political behavior of civilians. Consequently, we know less about how decisions by political authorities to employ various security institutions to repress dissent affects people’s support for those security institutions. This study examines the second dynamic: how repression by the police affects public perception of the police.
Exploring the relationship between the political authorities, the security apparatus, and the civilians to answer the question of “how individuals relate to the security apparatus” is foundational to theories of government and state development. The main purpose of government is to provide public goods. Of the many goods the state provides, such as healthcare and education, the provision of security may be one of the most important because without law and order society remains in relative anarchy (Hobbes 1946; Weber 1946; Wilson 1978). 7 Even in non-democracies, regimes see their ability to provide law and order as a cornerstone of legitimacy and economic development.
However, the co-production of domestic security depends on a set of three actors. The first set of actors are political authorities who are in charge of the government. These actors are tasked with legislating and executing laws to run the government and ensure development. As a function of their authority, they delegate the use of force to agents who are tasked with upholding and enforcing the laws (Blaydes 2018; Greitens 2016; Hassan 2017; Svolik 2012). The second set of actors are the security apparatus who are tasked with the capacity and directives to enforce the law. In the modern state, the police are the central actor responsible for providing law and order (Blair, Karim, and Morse 2019; Tyler 2006). The third set of actors are the civilians who live within the given territory of the state. They must decide whether to comply and cooperate with the security agents they interact with on a day-to-day basis. To the extent that political authorities relying on police to repress dissent might influence people’s perception of the legitimacy of police—determining who cooperates with police in the provision of law and order—the effects of state repression on public perception of police is of significant importance to our theories of governance and development.
Daily encounters between citizens and members of the police force are key to state development (Mazerolle et al. 2013; Wilson 1978). For many citizens, encounters with the police are the most likely interactions that people have with agents of the state (Lerman and Weaver 2014). As a consequence, the police and political authorities both rely on cooperation with the community to effectively accomplish their objectives (Skogan and Frydl 2004; Tyler 2006; Tyler and Fagan 2008). This cooperation involves citizens organizing neighborhood watches, taking note of suspicious activity, and reporting crimes. Information sharing and gathering is central to policing. Arriola et al. (forthcoming) argue that fragmentation of police forces can undermine information sharing and coordinated action, exacerbating conflict.
Yet, the structure and behavior of police forces also has implications for how people interact with the police. Critically, the police rely on information supplied by community members to prevent and solve crimes. They can receive this information only if citizens are willing to interact with them to provide it. People cooperate more with police when they trust them and view them as legitimate authorities (Nagin and Telep 2017; Tyler and Fagan 2008). Legitimacy is a function of whether or not individuals trust the police. 8 The legitimacy argument suggests that the “police can gain leverage for the co-production of security by inculcating the popular perception that their actions and decisions are legitimate” (Tyler and Fagan 2008, 235). The legitimacy-based framework linking legitimacy, compliance and cooperation generalizes across several policing contexts. 9 People who view the police as legitimate should be more likely to do what they say, obeying police directives and showing deference even if individuals do not agree.
One reason why people are unlikely to cooperate with police is because they fear the outcome of their interactions. 10 Individuals who have negative interactions with the police are less likely to trust them, especially if these individuals are from minoritized or politically marginalized groups. Existing work within criminology, for example, often focuses on apolitical forms of violence that police officers employ against individuals. Individuals see the police as less legitimate and are less likely to cooperate when the police behave in ways that are considered procedurally unfair or when the police are normatively misaligned with the community (Mazerolle et al. 2013). 11 Violations to procedural fairness or normative alignment undermine police legitimacy, precisely because they increase fear and mistrust in the police (Skogan and Frydl 2004; Tyler 2003, 2004). 12 When police act in ways that people believe are procedurally unfair, people will be more likely to fear encountering them. Similarly, when police act in ways that undermine their relationship to the communities they are policing, community members are more likely to believe that the police will engage in excessive force.
However, one important dimension less explored in the criminology literature is the way that political authorities rely on the police to repress dissent and the effect this might have on people’s perception of the police. This study focuses explicitly on politicized police action—coercive action taken by the police on behalf of political authorities to repress political opponents.
Coercive violence by the state often generates fear, which shapes people’s willingness to engage in dissent (Young 2019). Fear of experiencing abuse likely determines whether people view the police as legitimate authorities. Fear and mistrust of the police should decrease people’s willingness to interact with police, undermining the police’s ability to gather information, investigate crimes, and deter crimes. Political authorities using violence to repress dissent may be perceived as inherently illegitimate to citizens, even those who support the regime (Lupu and Wallace 2019).
13
Coercion against protestors can make people doubt their own loyalty to the ruling regime and question whether the security forces serve the interests of the citizens or the regimes, including their ability and willingness to protect them (Lupu and Wallace 2019). This is especially the case with escalating violence that becomes more indiscriminate. During indiscriminate repression citizens “can no longer assure themselves of immunity from repression by simply remaining politically inert” (Mason and Krane 1989, 176). If people see human rights violations by the police, we should observe police abuse negatively shaping whether people believe the police treat individuals with dignity and respect and their expectations about whether the police make fair and impartial decisions.
Partisanship, Fear, and Motivated Reasoning
There are several reasons why the effects of repression on public perceptions might vary depending on the attitudes and beliefs of individuals. The use of fear to control dissent is a go-to strategy for political authorities in non-democracies. However, mistrust of the government is unlikely to be normally distributed in society. The more people fear and mistrust the police, the more likely they are to see the police as illegitimate authorities. Accordingly, state repression is likely to have heterogeneous effects on individuals’ perceptions of the police depending on whether they support or oppose the political authorities. First, people might have a stronger negative reaction to repression, if they fear repression more (i.e., they political oppose the incumbent government). Second, people might interpret repression through a motivated reasoning lens, justifying or supporting repression if it supports the political status quo. This section explores these two possibilities.
Partisanship and Fear
Individuals’ loyalty to political authorities might affect how they respond to repression by police. Based on the theoretical framework above, I expect that political loyalty to the regime moderates the effect of repression on perceptions of the police based on whether or not people believe they are likely to experience repression. People who support the political authorities by attending pro-regime rallies, campaigning for candidates, and joining the ruling party should be less likely to fear repression. Alternatively, people who are engaged in collective action opposing the political status quo are more likely to fear experiencing future repression. Consequently, people who identify as regime supporters should be less likely to fear repression relative to non-supporters.
Partisanship and Motivated Reasoning
An alternative perspective based on motivated reasoning might predict that people interpret repressive action based on their support for or opposition to the political authorities who ordered the repression. The basic idea of motivated reasoning is that individuals often form opinions based on their existing values, beliefs, identities, and attitudes. One possibility is that people might view actions done by the police through a partisan lens. 14
Following the logic of motivated reasoning, people might interpret actions taken by the police as legitimate or not depending on whether they achieve or challenge their political preferences. Beliefs about whether or not repressive action taken by the police is legitimate depends on a partisan lens. Individuals supporting the ruling political authorities might view political opponents engaging in collective action as social deviants disturbing social order. People who are loyal to the political authorities should support state repression by the police, as those actions might be perceived as necessary to maintain the political status quo. From this perspective, police using force to deter dissent should increase support for the police among some individuals who politically align with the authorities.
Alternatively, people who do not politically identify with the political authorities should be less likely to justify repression, seeing any act of repression by the police as a clear violation of the role of police in society. Individuals who do not identify as loyal to the incumbent regime are relatively more likely to support challenges to the status quo, seeing repression as unjustifiable. People who oppose the regime are more likely to see political violence by the police, even selective repression, as a signal that the police are aligned with the political authorities rather than the community they are meant to be protecting. Even if they do not directly fear repression, opposition supporters are likely to see repression by the police as illegitimate.
Context: The Social Media Tax Protests and Selective Repression
I test my theory of selective repression on perceptions of police legitimacy with evidence from the Social Media Tax Protests in Uganda on July 11, 2018. To understand the selective repression of the Social Media Tax Protests in Uganda, it is necessary to understand the broader political context. Yoweri Museveni has maintained control of Uganda since 1986 alongside his incumbent party, the National Resistance Movement. 15 Although Museveni won the last three elections with an average vote-share of 60.27 percent, elections in Uganda were generally panned by international and domestic observers as lacking electoral credibility. Human rights are severely restricted and in many cases violated. 16
Politics as usual shifted in Uganda in 2017 when Robert Kyagulanyi Ssentamu, also known as Bobi Wine and the Ghetto President, announced his plans to compete in the upcoming by-election for Kyaddondo East Constituency. Well-known throughout Uganda as a famous musician, performer, and actor, Bobi Wine’s decision to campaign for parliament was not a surprise. His music became increasingly political following the controversial 2016 presidential election. Wine used his music and social media platform as a venue to call for political reform. However, as a political neophyte, his campaign was met with broad dismissal by much of the political establishment in Kampala and few took his candidacy seriously. The Kyaddondo East Constituency had two well-established candidates competing for it (Sitenda Sebalu of the incumbent ruling National Resistance Movement party and Apollo Kantinti of the dominant opposition Forum for Democratic Change party) in addition to two additional independent candidates. Wine’s electoral campaign strategy included combining a rigorous door-to-door walking campaign and substantial social media presence relative to many of the established candidates. Wine leveraged his social media presence and door-to-door walking campaign to a landslide victory. 17 Wine’s rise to political office threatened the thirty-three-year-old government led by President Yoweri Museveni and the National Resistance Movement. His social media presence, ability to energize crowds, and commitment to political reform contrasted starkly with the incumbent administration.
In March 2018, in part in response to Wine’s electoral success, the president of Uganda gave a directive that all social media platforms would be taxed. Broadly seen as a mechanism used by political authorities to repress political freedom, the new Over-The-Top Tax was designed to to raise resources “to cope with the consequences” of social media users’ “opinions, prejudices, [and] insults.” 18 For many in Uganda, WhatsApp, Twitter, and Facebook among other social media platforms were a major source of news and political information. The heightened exposure to information had reportedly led Ugandans to become more critical about political conditions in the country. 19 The Over-The-Top tax on WhatsApp, Twitter, and Facebook, among other sites, required users to pay a daily fee of 200 Ugandan Shillings (USD 0.05). This was not the first time political authorities attempted to control the flow of information via social media platforms. In 2016, the president ordered all social media sites to be shut down during the elections to control the flow of information. As a mechanism of social control, the Social Media Tax was implemented on July 1, 2018. 20
On July 11, 2018 a group of political activists organized by Bobi Wine gathered in Kampala to protest the Social Media Tax. Political authorities directed the Uganda Police Force to use force to break up the protest. The police used tear gas and live bullets to break up the protest. At least three protestors were arrested and remanded to Luzira Prison. Following the clash with the police, the Directorate of Public Prosecutions issued charges of assault and theft against several protesters including Bobi Wine, his bodyguard (Eddy Ssebufu), and two journalists (Raymond Mujuni and Joel Ssenyonyi). 21 Wine was detained and interrogated for more than 8 hours at the Central Police Station in Kampala. The coercive force used against the protestors was selective repression meant to deter collective action against the new social media tax. The security apparatus responsible for implementing it was the national police force. The police justified their actions by saying that Wine and the protestors had not gained permission to publicly assemble and charging protesters with assaulting members of the police and theft of police property including handcuffs. The selective state repression by the police of the Social Media Tax Protest is the type of violence that is likely to undermine public perceptions of police legitimacy depending on individuals’ political loyalty. In the next section, I explain the research design and data I employ to test my theory by leveraging this incident of selective repression.
Research Design
This study uses an unexpected event during surveys design approach to causal inference to estimate the effect of selective repression of the social media tax protests on individuals’ perceptions of police legitimacy (Muñoz, Falcó-Gimeno, and Hernández 2019). Next, I summarize my identification strategy, data, measures, and model specifications.
Identification Strategy and Data
I employ a unique research design opportunity that comes from the unexpected police clashes with the social media tax protesters while I was fielding a nationally representative survey on security and policing was being administered across Uganda. I consider participants surveyed before the selective repression of the protesters on July 11 as assigned to the control group and respondents interviewed on or after July 11 as assigned to the treatment group. 22 The timing of when respondents would be interviewed was determined at random without any knowledge of the social media tax protests occurring. Importantly, the timing of the selective repressive event did not influence the administration of the survey.
Data collection was conducted between 29 June and 20 July 2018, in 194 parishes located in 180 sub-counties within 127 counties, 100 districts and all 4 regions in Uganda. Table 1 shows the geographical distribution of the sample. The survey was embedded in a round of Twaweza’s Sauti za Wananchi project with assistance from Ipsos. 23 Twaweza is a highly respected research firm working throughout east Africa. Sauti za Wananchi is Africa’s first nationally representative mobile phone survey.
Overview of Multistage Sampling.
Note: Data on administrative units from the 2016 Uganda Electoral Commission Zoning.
Twaweza’s research team employed a multi-stage stratified sampling approach to achieve a representative sample of the total population of Ugandans who are eighteen years and older. The sample frame is based on the 2014 Uganda Population and Housing Census. 24 There was a high participation rate in the study (1,920 of 2,000 respondents participated).
Measurement
I employ two measures of police support as my main dependent variables. Respondents were asked whether they agreed with two statements regarding the police: (1) “The police in your community are legitimate authorities and you should do what they tell you to do.” (2) “You should do what the police tell you even if you do not understand or agree with the reasons.” These two measures refer to the legitimacy of police and whether people feel obligated to cooperate with them. 25 The first captures whether respondents perceive the police as legitimate and their obligation to cooperate with them. The second measure focuses on whether individuals feel obligated to comply with police, regardless of whether they agree with the directive. Responses to these direct questions were measured with a five-point ordinal scale from “strongly agree,” “agree,” “neither,” “disagree,” to “strongly disagree.” In addition to the main dependent variables, I employ four measures capturing whether people view the police as procedurally fair or normatively aligned with the community as alternative dependent variables.
Incumbent support is a binary indicator coded as 1 if the respondent stated that the National Resistance Movement is the political party they feel closest to and 0 otherwise. 26 Additionally, I consider an alternative measure: “If the election were held today, which political party will you support or vote for.” Similarly, if respondents said they would vote for the NRM, I coded this second measure as Incumbent vote and 0 otherwise. This second set of analyses using Incumbent vote are reported in the Online Appendix. 27 Control variables include participants’ self-reported age, gender, education level, and income status. Additionally, I control for whether the respondent lives in an urban or rural setting. Since exposure to crime, number of police officers deployed, and other potential confounders like prior state repression could affect people’s perception of police, I control for parish level fixed effects. 28
Model
I use the following OLS specifications to test hypothesis 1:
The respondent is referenced with the subscript i within the parish p. The
Similarly, I employ the following OLS specifications to examine the conditional effects of partisanship on repression by the police. To investigate heterogeneity in the treatment effect across political support for the regime in the other hypotheses, I interact
The difference between the equations is the addition of the interaction between
To identify valid causal estimates by comparing respondents surveyed before and after the event requires two key assumptions. First, I assume excludability: any difference between respondents surveyed before or after the event is the only because of the event. Specifically, the timing of the survey t affects the outcome variable
In expectation, we should not observe systematic differences in how people view the police based on how the data generating process unfolded. Covariate balance analyses show that there were not systematic differences between respondents interviewed before the selective repression of protesters and those interviewed after. 31 Consequently, I can estimate the effect of selective repression on public perceptions of police by exploring the variation between those in the control and treatment groups: those interviewed before the event compared to those interviewed following it.
Results
Table 2 shows the effects of the social media protest and arrests on whether people view the police as legitimate authorities (Model 1 and Model 3) and whether people think they should obey the police (Model 2 and Model 4). I hypothesized that repression by the police should decrease support for the police, negatively affecting public perceptions of police. A second observable implication of my argument is that the magnitude of the effect should be stronger among those who oppose the regime—those who have more to fear from police engaging in repression on behalf of political authorities.
Effect of Social Media Tax Protest/Arrests on Public Perception of Police Legitimacy and Obligation to Comply: Results from Parish Fixed-effects OLS Models.
Note: Parish fixed effects and control variables indicating education, age, gender, economic status, and urban/rural are included in all models. Robust standard errors are clustered by respondent in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1.
As theorized, police employing state repression, even selective repression against protestors, negatively affects public support for the police. Model 1 and Model 3 show the unconditional effects of selective repression of the Social Media Tax Protests on support for the police. Results from Model 1 show that people interviewed on or after the July 11 had lower perceptions of the police as legitimate authorities (
From a motivated reasoning perspective based on supporting or opposing the regime, it is surprising that these results hold across all levels of regime support. A theory of motivated reasoning and partisanship would expect that people who support the incumbent would support repression while people who oppose the regime would view such acts as illegitimate, hypotheses 3 and 4, respectively. Alternatively, I argued that political support for the regime should have a conditional effect on repression. People who support the regime fear experiencing repression less relative to those who oppose the regime; however, they are still concerned about the incumbent using the police to repress others. If this is the case, we should observe a weaker effect of repression on incumbent supporters.
Results from Equation 2 demonstrate the effects of selective repression on perceptions of police legitimacy are conditioned by political loyalty. Following best practices suggested by Brambor, Clark, and Golder (2006), I calculate the marginal effect of the repressive event.
32
The marginal effect of the repressive event is negative and statistically significant among opposition supporters,
The results of the interaction models show the effects of selective violence are conditioned by political loyalty to the regime. In Model 2, the effect of the repressive event is negative for people in the opposition and null for regime supporters; however, in Model 4, the effect is negative for both groups though lower in magnitude among regime supporters. This suggests that an explanation based on motivated reasoning and partisanship does not fully explain how civilians interpret repression by the police. Even controlling for parish-level fixed effects, the results of Model 4 demonstrate that selective repression of the Social Media Tax Protesters decreased support for the police among opposition and incumbent supporters.
Procedural Fairness and Normative Alignment
Does repression affect other perceptual measures of support for the police? Public perceptions of procedural fairness and normative alignment are key inputs of whether people support and trust the police (Knight and Schwartzberg 2019; Trinkner, Jackson, and Tyler 2017). I use two questions that capture respondents’ perception of the procedural fairness of police to consider the effects of repression on these perceptions of police. The first statement measures procedural fairness in interpersonal treatment: “The police treat people with dignity and respect.” The second statement captures procedural fairness in the decision making of the police officers: “The police make fair and impartial decisions in the cases they deal with.” In turn, I examine two statements that focus on whether respondents see the police as normatively aligned with the communities they are policing: “the police stand up for values that are important to you;” and, “the police usually act in ways consistent with your own ideas about what is right and wrong.”
Table 3 shows the effect of selective repression on public perception of procedural fairness of police in interpersonal treatment. The odd columns show the results of Equation 1 and the even columns report the results of Equation 2 (including the interaction term between
Effect of Social Media Tax Protest/Arrests on Public Perception of Procedural Fairness of Police: Results from Parish Fixed-effects OLS Models.
Note: Parish fixed effects and control variables indicating education, age, gender, economic status, and urban/rural are included in all models. Robust standard errors are clustered by respondent in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1.
In each of these models, the interaction between the repression and incumbent support remains in the hypothesized direction; however, the interaction is only statistically significant in Model 8.
33
In Model 6, the marginal effect of the repressive event among opposition supporters is
Next, I examine how selective repression affects public perception of the normative alignment of police operationalized through statements that measure shared values between respondents and the police. Table 4 considers the effect of social media tax protest/arrests on public perception that the police stand up for values that are important to the respondent (Models 9 and 10) and the police act in ways that the respondent considers to be right (Models 11 and 12). Again, odd columns show the results of Equation 1 and even columns show results from Equation 2 with the interaction term. These results provide additional support for my main hypotheses: selective repression i) decreases support for the police; and ii) the effects of state repression on perceptions of police legitimacy decrease in magnitude among those who support the regime.
Effect of Social Media Tax Protest/Arrests on Public Perception of Normative Alignment of Police: Results from Parish Fixed-effects OLS Models.
Note: Parish fixed effects and control variables indicating education, age, gender, economic status, and urban/rural are included in all models. Robust standard errors are clustered by respondent in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1.
In Model 10, we observe a decrease in the perception that the police stand up for values that are important to those individuals who do not support the ruling political party,
Pre-Existing Time Trends
Although the results above provide robust evidence, there are potential threats to inference. For example, one possibility is that pre-existing time trends that are unrelated to the event of interest might bias the finding. This section considers this possibility by employing a series of placebo treatments to test for plausible existing time trends that might bias the above results. If there are pre-existing time trends, we should expect an arbitrary cutoff point to affect the outcome of interest. However, assuming no pre-existing time trends then in expectation, an arbitrary point to the left of the cutoff point should not affect the outcome of interest. I construct five placebo treatments to test for time trends.
In these tests, I use the control group subsample, setting aside all respondents assigned to the treatment group (Imbens and Lemieux 2008; Muñoz, Falcó-Gimeno, and Hernández 2019). Data collection for the control group ran from June 29 to July 10. Rather than using the empirical median of the control group subsample to split the sample, as suggested by Imbens and Lemieux (2008), Muñoz, Falcó-Gimeno, and Hernández (2019), I create five placebo treatments in two-day windows (7/1, 7/3, 7/5, 7/7, and 7/9). Figure 1 shows the results of twenty regression models to examine whether the placebos have any effect on the two main measures of support for the police. The results show no evidence of consistent time trends occurring prior to the repressive event. 34

Results from twenty regression models of the date on which peoples attitudes changed.
Falsification Test: Effects of the Event on Other Outcome Variables
Another potential threat to inference is a violation of the excludability assumption by the occurrence of simultaneous events. One approach to consider this is to run a falsification test that examines the effect of the event on outcome variables that should not be affected by the repressive treatment event (Muñoz, Falcó-Gimeno, and Hernández 2019). To ensure that another event
Theoretically, a related series of events like an increase in crime, other security threat, or even frustration about the protester (as opposed to the state response) might affect both people’s trust in the police and their trust in others. Alternatively, the selective repressive event by police should not have an affect on the way people view other individual members of the society, even though it does affect public perception of the police and the broader political system. I construct a falsification test to rule out this possibility.
To do so, I employ a Relational Justice Schema Index used by Pickett, Nix, and Roche (2018). The relational justice schema maps to individuals’ “beliefs about the degree of procedural justice (as distinct from injustice) exhibited by others in society during interpersonal interactions” (Pickett, Nix, and Roche 2018, 99). These measures focus on interactions between various members of the public rather than on interactions between the public and the police. Broadly, this schema measures the extent to which people in society “generally exhibit procedural justice in their dealings with others…whether they tend to be respectful, fair, and unbiased” (Pickett, Nix, and Roche 2018, 99). Participants were asked to think about how people in society generally treat one another, and then responded whether they agreed or disagreed with three statements. (1) Most people are polite when dealing with others. (2) Most people treat other people fairly. (3) Most people treat other people with dignity and respect.
Table 5 shows the results of the falsification tests. I estimate the baseline OLS models from Equation 1 (odd columns) and Equation 2 (even columns). The dependent variables are participants’ responses to the justice schema instruments. Most people are polite when dealing with others (Models 13 and 14). Most people treat other people fairly (Models 15 and 16). Most people treat other people with dignity and respect (Model 17 and 18). Across each of the baseline models from Equation 1, the treatment coefficients (
Interestingly, Model 16 shows non-incumbent supporters express a lower assessment about whether people treat others fairly following the repressive event (
The falsification tests are neither necessary nor sufficient to demonstrate that the estimated effects of the selective repressive event on public perception of police is not caused by another event. However, they do increase our confidence that the exclusion restriction assumption is credible. In short, the link appears to be the state repression by the police against the protesters that affects public perception of the police and not mistrust of others in general or another event.
Effect of Social Media Tax Protest/Arrests on Public Perception of Other Members of Society: Results from Parish Fixed-effects OLS Models.
Note: Parish fixed effects and control variables indicating education, age, gender, economic status, and urban/rural are included in all models. Robust standard errors are clustered by respondent in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1.
Discussion and Conclusion
What are the effects of state repression by police on public perceptions of the legitimacy of the police? Using an unexpected event during surveys design approach to causal inference, I show that even selective repression by the police negatively affects how individuals view the police. I theorize that fear of the police as agents of repression is the main theoretical mechanism by which repression decreases support for police. When political authorities rely on the police to repress dissent, it increases people’s fear that the police function as agents of repression and undermines people’s trust in the police as legitimate authorities who provide law and order.
Surprisingly, repression decreases support for the police across the levels of support for the incumbent regime. 35 A theory of motivated reasoning does not completely capture changes in people’s attitudes toward the police following repression. Political loyalty to the political authorities, operationalized by support for the ruling party, does condition the effect of state repression by police on individuals’ perceptions of police abuse. I demonstrate the effects of selective state repression are strongest among those who do not identify as supporting the political party. Yet, even respondents who identify with the ruling party view the police as more illegitimate authorities and say they are less likely to comply with police directives after the police engage in repression.
This study makes several theoretical and empirical contributions to a growing literature on the politics of policing. Existing studies of state repression, by and large, focus on its effects on political participation either voter turnout or protests. Answering how repression affects political participation matters because even non-democracies employ protests and elections as important sources of information, if not a way to foster legitimacy (Brownlee 2007; Gandhi 2008; Magaloni 2006). However, one limitation is that this work primarily focuses on the costs of repression based on the negative effects it has politically (i.e, future protests or more support for the opposition at the ballot box).
By examining the effects of state repression on how the public views the legitimacy of the actors responsible for implementing it, I identify another cost for political authorities and society more generally. When political authorities rely on state security institutions, it undermines people’s willingness to cooperate with them. Even if fear associated with selective repression might induce political loyalty and deter dissent (Young 2019), relying on official members of the state security apparatus for coercive force against political challengers adversely affects people’s perceptions of the legitimacy of these institutions. This study demonstrates that even selective repression by the police undermines public perceptions of the police and people’s confidence that the police exist to protect and serve the interests of the community. This raises several implications for the relationship between repression and the politics of policing and crime.
Next, I join previous work that relaxes the unitary actor assumption within the state repression literature (DeMeritt 2015; Greitens 2016; Hassan 2017; Svolik 2012). However, rather than focusing only on the principal-agent problems, I argue the co-production of security depends on the relationship between political authorities, members of the security apparatus, and citizens. This has implications for the state’s ability to provide law and order. By examining the relationship between political authorities, the police, and citizens, I explore important variation in how people might view repression depending on their political preferences. In the context of this study, incumbent supporters did not support the use of repression by the police (at least selective repression did not have a positive effect on their assessment of the police). Future work might consider the strategic role that citizens play not only in the co-production of security but also the co-production of repression.
Finally, this analysis raises several policy implications and questions about the challenges of rebuilding trust in security institutions in post-conflict and fragile environments (Blair, Karim, and Morse 2019). If the recent work on the lingering political effects of state repression are any indicator, the negative effects of state repression on people’s perception of the legitimacy of their security providers are likely to persist. However, these effects might depend on whether people believe a person or population is deserving of repression or retaliation (Kao and Redlich Revkin 2019). Moreover, these results might be even less likely in areas with higher levels of conflict and repression, especially if people have no expectation of fair or unbiased treatment at the hands of the police. Future studies could examine the lingering effects of repression on public perceptions of the police and the potential for a “floor effect” when people no longer trust the state to provide security. One future avenue of research could examine challenges of police to solicit information to deter crime among politically marginalized communities, especially among those who have experienced past exposure to state abuse at the hands of law enforcement officers. This study provides a rare window into the effects of repressive action of a regime that is increasingly relying on the security apparatus to maintain power. Overall, this study demonstrates a cost of repression to political authorities and society. When political authorities rely on the police as agents of repression, people view the police as more illegitimate authorities.
Supplemental Material
Supplemental Material, sj-pdf-1-jcr-10.1177_00220027211013097 - How Repression Affects Public Perceptions of Police: Evidence from a Natural Experiment in Uganda
Supplemental Material, sj-pdf-1-jcr-10.1177_00220027211013097 for How Repression Affects Public Perceptions of Police: Evidence from a Natural Experiment in Uganda by Travis Curtice in Journal of Conflict Resolution
Supplemental Material
Supplemental Material, sj-zip-1-jcr-10.1177_00220027211013097 - How Repression Affects Public Perceptions of Police: Evidence from a Natural Experiment in Uganda
Supplemental Material, sj-zip-1-jcr-10.1177_00220027211013097 for How Repression Affects Public Perceptions of Police: Evidence from a Natural Experiment in Uganda by Travis Curtice in Journal of Conflict Resolution
Footnotes
Author's Note
Replication files, supplementary material and online appendix will be available on the author’s website.
Acknowledgments
Thanks to Jennifer Gandhi, David Davis, Danielle Jung, Jason Lyall, Ben Valentino, Jeffrey Friedman, Courtenay Conrad, Kristine Eck, Charles Crabtree, Jonathan Mummolo, and Dartmouth’s IR Dickey Center Seminar Series, the Emory Comparative Politics Reading Group and Empirical Study of Conflict (ESOC) Annual Meeting Virtual Session on Crime and Policing for helpful comments and suggestions on previous drafts of this paper. Mildmay Uganda Research Ethics Committee (MUREC), an ethics review committee accredited by the Uganda National Council of Science and Technology approved the study (#REC REF 0204-2017) and the Uganda National Council of Science and Technology (REF SS4302). Additionally, Emory University’s Institutional Review Board declared the study exempt under 45 CFR 46.101 (b)(2) REF IRB00104491.
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 research received funding by the Institute of Developing Nations, Laney Graduate School at Emory University, the Andrew W. Mellon Foundation (Mellon PhD Intervention), National Science Foundation (#1343123), and the Peace Scholar Fellowship at the United States Institute of Peace.
Supplemental Material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
