Across the globe, states have attempted to contain COVID-19 by restricting movement, closing schools and businesses, and banning large gatherings. Such measures have expanded the degree of sanctioned state intervention into civilians' lives. But existing theories of preventive and responsive repression cannot explain why some countries experienced surges in repression after states in Africa initiated COVID-19-related lockdowns. While responsive repression occurs when states quell protests or riots, “opportunistic repression” arises when states use crises to suppress the political opposition. An examination of the relationship between COVID-19 shutdown policies and state violence against civilians in Africa tests this theory of opportunistic repression. Findings reveal a large and statistically significant relationship between shutdowns and repression, which holds after conditioning for the spread and lethality of the disease within-country and over time. A subnational case study of repression in Uganda provides evidence that the increase in repression appears to be concentrated in opposition areas that showed less support for Yoweri Museveni in the 2016 elections. Opportunistic repression provides a better explanation than theories of preventive or responsive repression for why Uganda experienced a surge in repression in 2020 and in what areas. The results have implications for theories of repression, authoritarian survival, the politics of emergency, and security.

This article considers whether governments have used the COVID-19 pandemic to cement their authority through repression. A core feature of emergency responses is the suspension of the rule of law, which permits states greater latitude to take actions in order to protect the entire citizenry.1 We argue that crises create opportunities for governments to suppress the political opposition, which we refer to as “opportunistic repression.” We test this theory by examining the relationship between COVID-19 shutdown policies and state violence against civilians in Africa. We conduct a subnational case study of repression in Uganda to analyze whether patterns of repression after the shutdown mandate vary along partisan lines across different districts of the country. We find that theories of preventive and responsive repression cannot explain why Uganda experienced a surge in repression in 2020.

Understanding the effects of COVID-19 on governance is a pressing concern. The measures taken in response to the coronavirus pandemic constitute extraordinary government interventions into citizens' lives. Across the globe, states have attempted to contain COVID-19 by restricting movement, closing schools and businesses, and banning large gatherings. Such measures have significantly expanded the degree of sanctioned state intervention into civilians' lives. The measures associated with the COVID-19 pandemic resemble other expansions in state authority associated with crises and states of emergency.2 Joan Barceló et al. go so far as to describe many of the measures adopted in response to the pandemic to be “observationally equivalent” to government repression and find that the history of state repression influences the timing and likelihood of a lockdown.3

Several recent studies have examined the relationship between coronavirus-related shutdown policies and political violence,4 building on extant research regarding public health crises and political violence.5 But none of this research provides a framework for understanding how public health measures associated with the pandemic affect domestic repression. This gap reflects the uneven and intermittent treatment of crises by the literature examining political repression in general.

The study of crises and repression is complicated by the fact that emergencies often involve political violence such as terrorism, civil conflict, or interstate war. In these contexts, it is difficult to determine causation since an underlying trend of escalating conflict may drive a government to declare a state of emergency.6 To untangle the relationship between crises and political repression, we study an acute public health crisis in which it is plausible that underlying political violence trends are not driven by COVID-19-related lockdown policies.

We show that state violence against civilians increased across several African countries after states issued shutdown orders in response to the spread of the COVID-19 virus. We also find troubling evidence that, at least in some contexts like Uganda, this repression was partisan. Our results suggest that the coronavirus pandemic created a “window of opportunity” for governments to use legitimate public health interventions as cover to engage in physical repression.

We arrive at this conclusion through two steps. First, we exploit the within-country timing of COVID-19-related shutdown policies to estimate the relationship between emergency orders from governments and state violence against civilians across Africa. We find a substantively large and statistically significant relationship between shutdowns and repression, which holds after conditioning for the spread and lethality of the disease within-country and over time. This result suggests that lockdown measures related to the coronavirus pandemic expanded the ability of the state to intervene in citizens' lives. Moreover, the state often implemented these powers through force rather than more proportionate measures.

Second, to better understand the mechanism driving this relationship and to assess the targets of opportunistic repression, we use subnational variation in support for the Yoweri Museveni regime in Uganda (as measured by the district-level 2016 presidential vote share for Museveni) and the timing of the national shutdown order (in March 2020) to estimate how repression changed after the government clamped down on freedom of mobility and public gathering. We find that the increase in repression is concentrated in opposition areas that showed less support for Museveni in the 2016 elections. This finding complements our cross-national statistical analysis and suggests that states can leverage ostensibly legitimate expansions in the scope of state power produced by crises to repress political opponents.

Our study makes two crucial contributions, the first of which is theoretical. We introduce the concept “opportunistic repression,” which shifts the focus on the drivers of state repression from the activities and capabilities of the opposition to those of the state. A large theoretical literature on repression examines how governments choose an optimal level of state violence that balances the need to stymie immediate threats to its power with the risk of catalyzing more violence and resistance in response to a crackdown.7 This literature broadly groups repression into two categories: preventive and responsive.8 Preventive repression focuses on future challenges to the regime, whereas responsive repression targets dissent.9 Preventive repression is typically modeled as a static feature of autocratic governments.10 Responsive repression is that which emerges in direct response to a challenge to the state.11 As the name suggests, “opportunistic repression” accounts for the repression that emerges as a function of state opportunity rather than in response to actual or possible mobilization against the incumbent.

The key feature distinguishing opportunistic repression from other theories of repression is the focus on the changes in the tools at the incumbent's disposal. Although preventive repression is catalyzed by the threat of opposition activity and responsive repression results from active challenges, opportunistic repression emerges when there is an increase in the possible scope of repressive activity available to the state, rather than a change in the opposition's activities or behaviors. The theory of opportunistic repression contributes to the literature on how governments navigate and manipulate international norms while consolidating power.12 We show how repressive regimes may leverage changes in international norms during worldwide crises to repress citizens and secure their positions. Consequently, international critics are unlikely to advocate sanctions or other measures to stop governments from attempting to slow the spread of a novel virus. Scholars can use our framework to examine how other crises affect government repression.

Our second contribution is empirical. We examine state repression associated with COVID-19-related shutdown policies in Africa to provide evidence on the ways in which crises threaten citizens' health and their security. Existing evidence on disease and conflict is mixed: some studies find that pandemic-related public health measures increase civil conflict,13 while others show that shutdown policies reduced political violence in the early stages of the COVID-19 pandemic.14 This article provides important evidence about state violence against non-state actors. Although Tiberiu Dragu and Yonatan Lupu assert that there is an endogenous constraint on state violence during a crisis, we find that COVID-19 has been associated with increases in repression in Africa.15 Our cross-national findings underline that some African states used physical repression to enforce lockdown policies to prevent the spread of COVID-19, which constitutes an overstepping of loosened normative restrictions on state intervention. Furthermore, through our subnational case study of Uganda, we highlight the possibility that the brunt of such state violence is borne by those in opposition-controlled areas. This paper thus contributes to our understanding of the relationship between pandemics and conflict and various secondary effects of COVID-19 on citizens.

Across regime types, those in office seek to stay in power.16 Many incumbents use force to remain in office, and this repression can manifest in myriad forms. Officeholders sometimes use violence to target opposition rallies or civilians, to harass or detain journalists, or to intimidate political challenges to the status quo. Despite the diversity of ways in which repression can emerge, the literature has identified two broad categories of repression: responsive and preventive.

responsive repression: quelling direct challenges to the incumbent

Responsive repression asserts a sort of Newtonian logic to repression, whereby for every opposition action there is a repressive measure. Christian Davenport calls this relationship the “Law of Coercive Responsiveness.”17 Such repression emerges to quell large-scale dissent to demobilize civilians who are actively demanding change.18 As the name suggests, the catalyst for this repression is opposition activity.

In line with the Law of Coercive Responsiveness, different forms of domestic dissent are more likely to produce repression. Sabine Carey finds that guerilla warfare is likely to produce severe state repression, while other milder forms of dissent do not provoke crackdowns.19 Jennifer Earl et al. find that protests that constitute a greater threat to the state (i.e., those that either attract many participants or use “confrontational” tactics) are more likely to be policed than smaller or less aggressive demonstrations.20

Images of state security officials violently dispersing large protests and riots in the capitol city, or putting down attempted coups, vividly illustrate the prevalence of responsive repression. Yet, including only responsive repression is an incomplete assessment of political repression, and it portrays the state as being perpetually flat-footed in its response to domestic political challenges.

preventive repression: anticipating dissent

The theories of preventive or preemptive repression, in which incumbents may use preventive measures to raise the costs of dissent ex ante, emerged to explain those instances in which the incumbent represses to deter future or potential dissent.21 Preventive repression can be driven by a variety of calculations and conditions, including both the capacity of the opposition and looming opportunities for it to mobilize. Work on preemptive repression examines how autocrats consider local histories of dissent to plan their repression around electoral calendars,22 culturally relevant periods,23 and demographic profiling.24 Other work examines how governments detect and intervene in nascent opposition activities to prevent “overt” challenges.25 This focus often underscores the endogeneity of repression, as previous experiences with dissent and repression shape patterns of repression in the future. Countries that have experienced culturally significant challenges to their legitimacy may keep a “dissent calendar,” whereby states increase repression to prevent the political opposition from capitalizing on the symbolic relevance of those key dates.26 The “June Fourth Incident” in Tiananmen Square, for example, is associated with both responsive repression (on June 4, 1989, in response to the protest) and preventive repression (leading up to subsequent June 4ths).27 New developments may also prompt states to plan for preemptive repression against those whom they think might engage in dissent.28 For example, the discovery of oil fields may lead states to repress citizens in expectation of future conflict risk.29 Davenport's study of how states “license” repression underscores that domestic and international threats, more so than citizens' behavior, predict repression in the United States.30

opportunistic repression: manipulating crises to suppress opposition

The literature on repression has largely (and perhaps paradoxically) examined the capabilities and activities of domestic opposition rather than the context in which the state operates and the repressive capabilities of the government. It is important to consider what other factors influence patterns of political repression. Although there are material influences on a state's capacity to use coercion (e.g., the number of police or new technologies of surveillance or repression), this article considers the ideational constraints imposed on states that the international community considers to be “legitimate” or appropriate forms of intervention. Existing theories do not account for patterns of repression that such conditions produce. We build on Davenport's theory of “licensing repression,” which underlines that governments can use both domestic and international threats to legitimize their repression and finds that, throughout U.S. history, the presence of such threats were better predictors of repression than behavioral measures of dissent, such as demonstrations or antigovernment violence.31 We examine how states will respond when shifts in the international community's norms legitimize greater intervention into civilians' lives.

We conceptualize opportunistic repression as a form of repression that results from a shift in a state's operating environment. Opportunistic repression exists when incumbents make use of a widened scope of legitimized intervention into citizens' lives to engage in repression to maintain their positions in power. We theorize that emergencies and crises—defined in this article as situations in which there is a globally recognized threat that necessitates increased government intervention—will be associated with an increase in government repression.32 Specifically, we build from constructivist international relations theory to argue that states use the international community's increased tolerance for intervention into citizens' lives as a pretext to engage in political repression, particularly against those groups that oppose the incumbent.33

Some incumbents seek to retain and consolidate power by using the coercive power of the state against perceived threats (i.e., opposition groups, social movements, and rival political candidates). Repression is one tool that states can use to achieve these interests, but a variety of domestic and international factors constrain how states can use repression, including fear of public backlash and international condemnation, or even international intervention to prevent such activities. Crises and emergencies can produce rapid changes in what degree of state intervention (and what forms) is tolerated. Article 4, section 1 of the International Covenant on Civil and Political Rights clearly states that, “[…] in time of public emergency which threatens the life of the nation and the existence of which is officially proclaimed, the States Parties to the present Covenant may take measures derogating from their obligations under the present Covenant to the extent strictly required by the exigencies of the situation, provided that such measures are not inconsistent with their other obligations under international law and do not involve discrimination solely on the ground of race, color, sex, language, religion or social origin.“34 In some instances, increased intervention during a crisis is a legal expansion of state authority; a crisis or the declaration of a state of emergency is legible to the international community as a situation requiring unusually invasive state activity.35 In short, emergencies provide more opportunities for the state to legitimately intervene in citizens' lives, which permits incumbents to secure their positions by repressing the opposition. Wesley Widmaier et al. emphasize that ”no less than anarchy itself, wars and crises cannot be reduced to material forces or socialization, but are what agents make of them.“36 We adopt the framework of international legitimation to examine both how crises can shift what is deemed to be appropriate interventions into citizens' lives and how this shapes the pattern of incumbent repression.

Building on previous studies of how international norms and constraints shape domestic governance, our concept of opportunistic repression helps shed light on how states respond during crises. Although studies have considered how the international environment affects state repression, they often focus on how patterns of repression do (and do not) change when states voluntarily join human rights treaties and conventions,37 or when the international community scrutinizes certain forms of repression.38 Other studies have underlined how the international community can influence the manner in which an incumbent “frames” or justifies repression.39 Additional work examines how international factors, such as threats related to war contagion40 and transnational terrorist networks,41 may increase state repression.42 Recent studies have also underscored how processes of “authoritarian learning”43 and “authoritarian diffusion”44 shape the tactics that governments use to respond to challenges to their rule. Davenport's examination of “tyrannical peace” and how the end of the Cold War affected domestic repression demonstrates that different types of governments respond differently to shifts in the international system, which suggests that there is a complex relationship between regime type, international influences, and domestic repression.45 Furthermore, Mirjam Edel and Maria Josua contend that authoritarian governments frame repressive acts differently for domestic and international audiences to gain legitimation from both groups.46 Although these studies suggest that the international community influences how domestic political repression manifests, the threat of international sanctions or disapproval may be insufficient to entirely eradicate repression.

Our theory of opportunistic repression helps illuminate the understudied phenomenon of how repression shifts in response to crises. Of the few academic studies that have considered how crises affect state repression, most have primarily focused on domestic audiences or factors. A game theoretic model from Tiberiu Dragu and Xiaochen Fan suggests that the preferences of security forces will act as an “endogenous constraint on the abuse of emergency powers.”47 They find that emergencies are associated with no change in the degree of repression, all things held equal. Yet, Dragu and Fan's finding stands in stark contrast to other studies that point to how states have leveraged crises to justify political repression.48 In their study about COVID-19-related lockdowns, Barceló et al. find that previous state repression predicted the timing and likelihood of a state adopting a lockdown measure.49

Thus, despite decades of research into the dynamics of state repression, it is still unclear whether governments increase political repression during certain types of crises. Crises may not always trigger political repression, but if they do, the characteristics that influence this process remain undertheorized. Existing theories of responsive and preventive repression do not offer clear guidance about how states will respond to crises that produce shifts in the acceptable degree of intervention in citizens' lives, absent a clear connection between the crisis and the domestic opposition's capacity to engage in dissent or physical displays of their opposition. Determining if, how, and against whom states will use their expanded powers is a critical task for political science.

Our theory of opportunistic repression explains how crises and emergencies affect governments' patterns of repression. In our formulation, crises that produce shifts in internationally acceptable levels of state intervention provide the backdrop for opportunistic repression. As table 1 emphasizes, the crises that can catalyze opportunistic repression do not necessarily increase antigovernment sentiment or affect the mobilization or capacity of domestic opposition forces. An important difference between our theory of opportunistic repression and theories of preventive and responsive repression is that opportunistic repression results from a shift in the operating environment of the incumbent rather than increases in the capacity or activity of the opposition.

Table 1.
Types of Repression and Expectations during a Pandemic
Type of repressionCatalyst of repressionExampleExpectations during COVID-19
responsive repression informed by domestic political conditions; opposition challenge or active, ongoing threat to regime An opposition leader holds a rally, which is met with violence by state forces. reduced repression 
preventative repression informed by domestic political history; expectation of opposition challenge, anticipated upcoming threat to regime A state increases patrolling and repression prior to a politically significant anniversary of unrest. no change in repression 
opportunistic repression exogenous shock to the state's operating environment (temporarily) expands permissible levels of state repression A sudden disaster or new threat legitimizes expanded scope of state intervention into citizens' lives. increased repression 
Type of repressionCatalyst of repressionExampleExpectations during COVID-19
responsive repression informed by domestic political conditions; opposition challenge or active, ongoing threat to regime An opposition leader holds a rally, which is met with violence by state forces. reduced repression 
preventative repression informed by domestic political history; expectation of opposition challenge, anticipated upcoming threat to regime A state increases patrolling and repression prior to a politically significant anniversary of unrest. no change in repression 
opportunistic repression exogenous shock to the state's operating environment (temporarily) expands permissible levels of state repression A sudden disaster or new threat legitimizes expanded scope of state intervention into citizens' lives. increased repression 

NOTE: The theory of responsive repression predicts that pandemic-related lockdown measures will reduce the opposition's capacity to organize, whereas the theory of preventative repression predicts that such measures will not affect repression levels. Opportunistic repression, in contrast, predicts that repression levels will increase because COVID-19-related shifts in international norms legitimize state intervention.

Though we test our theory of opportunistic repression on a public health crisis, there are many examples of states leveraging the international community's concern regarding transnational terrorist networks to provide cover for domestic repression. For example, China used repressive tactics against the Uighur community after accusing it of being associated with transnational jihadist groups.50 Max Bergmann and Alexandra Schmitt tie the leeway that “stable” states engaging in repression have to the “‘building partnership capacity’ strategy” that the United States adopted following September 11, 2001, “which called for increasing the capabilities of developing states to better police and patrol their neighborhoods and to close off space for insurgent groups.”51 As a result of this approach, “U.S. aid was often provided to nondemocratic states or partners that violated human rights but were considered critical partners in the ‘war on terror,’” to the point that “[a]lmost all U.S. security aid provided year over year is driven by a strategic rationale that is centered on building better counterterrorism partners.”52 Some incumbent regimes have adopted this frame, both to foster closer ties to Western governments like the United States and to exploit counterterrorism norms in order to engage in repression. Edel and Josua note that “the diffusion of ‘war on terror’ rhetoric and practices on the global level” has altered “international learning processes,” enabling the global spread of political repression.53 In short, shifts in what the international community considers legitimate grounds for intervention can lower the costs of repression for incumbents by reducing the likelihood of international sanction or censure for behavior that is framed within the concerns of the international community.

Additionally, natural disasters and public health crises may also shift the norms of the international community. States of emergency, declared for whatever reason, are often accompanied by the suspension of certain rights or liberties that are tolerated by the international community (within certain boundaries and in response to certain threats). A crisis may thus represent an opportunity for the incumbent to use the new state powers to engage in political repression. We posit that extant theories of state repression generally overlook the role that international norms play in shaping the scope, tactics, and justification of political repression. By examining the behavior and capabilities of the state rather than the domestic opposition, our theory contributes to a more holistic understanding of political repression dynamics and the relationship between crises and political repression.54

Though some may suggest that repression following the adoption of lockdown measures is a form of preventive repression, this reflects a misunderstanding of how COVID-19 has shifted the capabilities of domestic opposition. In many ways, lockdown measures hamstrung opposition forces by restricting their ability to credibly threaten mass protest or other forms of dissent and mobilizations against human rights violations. For example, Melissa Pavlik shows that demonstration activity fell by about one-third globally during the pandemic compared with pre-pandemic levels.55 Some burgeoning or resurgent protest movements, like those in Iraq and Lebanon, were stymied following lockdown orders. Though it is possible to organize an opposition rally with social distancing measures (as protests in Israel, Pakistan, Brazil, and the Black Lives Matter movement in the United States demonstrate), and there are certainly other ways for domestic opposition groups to mobilize against the government, we find that pandemic-related restrictions increased collective action problems.56 Previous theories of state repression would posit that states would reduce their repressive activity in response to the weakened capacity and reduced activity of the opposition.

To test our theory, we consider how states respond to a plausibly exogenous event (i.e., the COVID-19 pandemic) that produces an internationally sanctioned increase in the state's scope of intervention but no increase in the opposition's capabilities to challenge the incumbent's hold on power. The pandemic represents a unique opportunity for increased repression. Lockdown procedures, increased scrutiny of civilians' movements and interpersonal interactions, and COVID-19 testing and tracing represent significant interventions into civilian life. Although the international community may have previously regarded such activities with suspicion, it has endorsed many of these policies within the context of the pandemic. Consider, for example, the World Health Organization's (WHO) 2019 guidelines for responding to pandemics, which recommend measures such as school and workplace closures, avoiding crowding, and international travel restrictions. As Barceló et al. note, the lockdown measures are “observationally equivalent” to repression, and the authors compare the policies that states adopted to prevent or slow the spread of COVID-19 with curfews and restrictions on movement during clear-cut instances of political repression.57 Under the conditions of COVID-19, however, the government's response to the crisis can be plausibly justified to international actors who would otherwise sanction or punish rights violations. Despite language from the WHO regarding respect for human rights during health emergencies, the body lacks the capability to enforce these regulations.58 According to Human Rights Watch, “governments in at least 24 countries have enacted vague laws and measures that criminalize spreading alleged misinformation or other coverage of Covid-19, or of other public health crises, which the authorities claim threaten the public's well-being. Governments can easily use imprecise laws as tools of repression. At least five countries have also criminalized the publication of alleged misinformation on a range of other topics, including public health.”59 We argue that the international backing of such interventions, and their relative silence regarding early reports of abuse of these new powers, provides latitude for incumbents to intervene more severely in some communities than in others.

Importantly, the WHO guidelines do not authorize physical integrity rights violations, even while sanctioning greater regulations of citizens' behaviors. When public health measures are implemented appropriately, proportionately, and within the bounds of the rule of law, we do not consider them to constitute state repression. For many of these public health policies, nonviolent and proportionate enforcement mechanisms (such as fines) for noncompliance are feasible options.60 As UN High Commissioner on Human Rights Michelle Bachelet stated in April 2020, “[E]mergency powers should not be a weapon governments can wield to quash dissent, control the population, and even perpetuate their time in power. They should be used to cope effectively with the pandemic—nothing more, nothing less.”61 These statements implicitly acknowledge that loosened international norms regarding intervention in civilians' lives may tempt governments to repress the domestic opposition. In this article, we consider the use of physical violence to enforce COVID-19 lockdown measures to be instances of government repression. The following section describes, briefly, how some governments have enforced lockdown measures through unjustifiable means.

how incumbents use international norms to repress opposition

A number of countries have abused COVID-19 lockdown provisions since the onset of the pandemic. According to Amnesty International, “[A]cross Europe, there were several cases in Belgium, France, Greece, Italy, Romania, and Spain in which law enforcement officials resorted to the unlawful use of force to impose lockdown measures on people who did not offer any resistance or constitute a significant threat.”62 In some instances, this repression has fallen along preexisting fault lines. In India, for example, there are reports of the government using COVID-19 as an excuse to engage in repression in Kashmir.63 Kim Yi Dionne and Fulya Felicity Turkmen also found a surge in “othering” of marginalized communities in response to the COVID-19 pandemic, reflecting a broader pattern of xenophobia and discrimination in response to pandemics.64

Although the United Nations has condemned the “toxic lockdown culture” that has emerged in some places, the international community has done little to sanction incumbents who have overstepped the bounds of appropriate lockdown enforcement.65 The guidance that the United Nations released in April 2020 states that “law enforcement officials may use force only when strictly necessary and to the extent required for the performance of their duty and only when less harmful measures have proven to be clearly ineffective,” and calls for prompt investigations into allegations of improper enforcement.66 Yet, more than a year into the pandemic, there was evidence that some states had abused their powers and were not being held accountable for these violations. Human Rights Watch estimates that “authorities in at least 51 countries have used laws and regulations adopted to prevent the spread of Covid-19, as well as counterterrorism and other measures predating the pandemic, to arbitrarily arrest, detain, and prosecute critics of government responses to the coronavirus, or of policies unrelated to the pandemic, resulting in fines and imprisonment. Those targeted include journalists, bloggers and others posting online, opposition figures and activists, protesters, academics, healthcare workers, students, lawyers, cartoonists, and artists.“67 Their full investigation revealed ”violations of the rights to freedom of expression and assembly in 83 countries.“68

This brief overview illustrates how the COVID-19 pandemic expanded the scope of what forms of intervention were considered legitimate and appropriate by the international community and how some governments translated their enhanced authorities to engage in political repression.

Our theory predicts that the onset of an emergency will give states the opportunity to repress dissidents using crisis as a pretext. We study the relationship between the COVID-19 pandemic and repression among African states to test our argument, according to which the onset of the emergency should increase repression, especially in areas with more opposition members. Our analysis has two steps: first, we show that repression generally increased after the onset of the pandemic. Second, we use Uganda as a case study to show how repression changes in opposition areas compared with incumbent strongholds.

For this study, we limit our examination of the relationship between pandemic-related lockdowns and repression to African countries. This choice reflects data constraints as well as the region's history and characteristics that make it an especially relevant area to study. Many African governments have previous experience responding to public health crises such as the HIV/AIDS and the Ebola epidemics, which suggests that these governments would be institutionally well-equipped to navigate the COVID-19 pandemic and implement effective interventions.69 Additionally, the first reported case of COVID-19 in Africa was in February 2020, giving African heads of state the opportunity to learn from other governments' responses to the virus.70 The African context also provides the opportunity to explore the impact of COVID-19 response measures across a wide variety of institutional contexts and regime types.

Our outcome data for political violence and repression come from the Armed Conflict Location & Event Data Project (ACLED).71 For more than two decades, ACLED has been coding data for African countries, which were the original focus of the dataset. Apart from its robust coverage of Africa, ACLED data have several unique features.72 First, ACLED data are precisely measured in time and space: ACLED measures the timing (to the day) and specific location of protests and political violence events, and records the fatalities, conflict actors involved, and a description of the event. These details allow us to conduct fine-grained analyses of conflict at the subnational level. Second, ACLED's coding methodology relies on a team of researchers who code events in more than twenty different languages using local, regional, national, and international media sources.73 This robust data collection methodology substantially increases the number of political violence events captured compared with other datasets, especially in rural and remote areas. ACLED codes a wider variety of violent and nonviolent political confrontations than other datasets, which allows us to more accurately measure how patterns of protest and violence change over time. Moreover, other prominent observational datasets of political violence (e.g., Uppsala Conflict Data Program [UCDP] and Global Terrorism Database [GTD]) are inappropriate for our analyses because the universe of events and modes of collection are different than ACLED's.74 Notably, many datasets have strict limits on the type of events that they include in their data. For example, GTD only codes events that qualify as “terrorism,” whereas UCDP includes only events that occur within a predefined contest between conflict dyads, which results in an arbitrary number (i.e., twenty-five or more fatalities each year). The universe of events coded by these datasets is therefore not conducive to the types of repression that we study.

ACLED data have important limitations, however, many of which are inherent to observational data generally and to cross-national, observational data of political violence specifically. ACLED probably does not capture a complete universe of political violence events, and those events that are missing are nonrandom. To the extent that missing data would be correlated with the onset of shutdowns, we believe that bias would be downward because COVID-19 caused an underreporting of repression events. First, it is more likely that the operations of the free press are curtailed during emergencies. Second, social distancing and isolation would have made transmitting information more costly than before shutdowns. For missing data to bias our results upward, repression events would have to have been underreported before the pandemic and overreported after shutdowns began, without any corresponding changes in the levels of state violence. Given the conditions on the ground in several states across the continent during lockdowns, this possibility is unlikely.

To measure repression, we rely on a conservative measure that captures onesided state violence against civilians. For this article, we define civilians as unarmed individuals who are not partaking in collective action, such as protesting or rioting. Importantly, although we believe that state dispersal of protesters or rioters or violence against citizens who are engaging in public demonstrations constitutes oppression, such instances do not capture our main phenomenon of interest because they are examples of responsive repression.

Although ACLED covers all fifty-four African countries, we restrict our analysis to the forty-nine African states that are also covered by the Oxford COVID-19 Government Response Tracker, our data source for COVID-19 policies. We use data from January 1, 2017, until June 30, 2020. This period goes back far enough to capture potential pre-COVID-19 trends but not too far such that the control period becomes unlike the shutdown period.75

We measure exposure to COVID-19 restrictions using the Oxford COVID-19 Government Response Tracker.76 The database records the country-date of COVID-19-related policies, deaths, and cases. For this article, we are interested in the measurement of government shutdown policies rather than economic relief packages or information campaigns. Specifically, we measure whether a government has enacted at least one policy that shuts down public transit, closes workplaces and schools, cancels public gatherings, implements stay-at-home orders, or limits internal movements. It is most likely, we argue, that governments will use coercion to enforce shutdown policies. We use measures of deaths and cases to capture the lethality and spread of the virus over time and space, but it is important to note that the exact numbers are likely to contain some measurement error. Nonetheless, the information on case numbers and deaths represents the best available data sources to capture the prevalence of COVID-19.

We aggregate our data by the country-month to construct a balanced panel dataset of African states from January 2017 until June 2020. We first study the cross-national impact of COVID-19 shutdown policies on repression. In March 2020, nearly all African countries issued some form of shutdown order, including restricting public gatherings, closing schools and places of work, shutting down public transportation, canceling public events, as well as issuing stay-at-home orders and policies constraining internal movements. Indeed, of those African countries with data, only Burundi and Chad did not shut down in March 2020. Oxford data does not track the following states: Comoros, Equatorial Guinea, Guinea-Bissau, and São Tomé and Príncipe. We exclude these states from our analysis because there is no variation in their enacted shutdown policies. Including them, however, does not substantively change our results because our estimation procedure computes the average change in repression within states that enact lockdowns. A list of all countries in the sample is included in online appendix A.

We use an interrupted time series (ITS) design to study the relationship between shutdown enactment and state repression. Our design exploits the exogenous timing of the policies within states to estimate the change in repression resulting from state-ordered shutdowns. The infectious nature of the virus forced states to quickly enact restrictive measures that limited citizens' freedom of movement in order to prevent rapid spread of COVID-19. The timing of shutdowns is plausibly unrelated to underlying political violence trends within countries because government policies did not control or influence the development of the COVID-19 emergency. Once enacted, however, we find that lockdown orders licensed repression. Under this assumption, we estimate the relationship between shutdowns and repression using an ITS methodology. The goal of ITS is to identify changes in levels of an outcome at the time of some event that is expected to interrupt the trend of the outcome over time. In our study, the time series is repression within a country, and the interruption to the trend is the onset of COVID-19-related lockdowns. Our core identifying assumption is that the level of repression within a country before the onset of shutdown policies is a valid counterfactual for what repression would have looked like in the absence of lockdowns. In other words, we expect that if the COVID-19 crisis had not occurred, repression trends in the country would have remained consistent. Under this assumption, our estimates represent the change in repression caused by the COVID-19 emergency. We describe the estimating equation below, and the econometric specification appears in online appendix C.

The outcome of interest is our narrowly defined concept of repression, measured using ACLED data regarding state violence toward civilians (interaction code 17 in the ACLED dataset). Our baseline estimates measure the number of events. We include binary coding because whether repression occurred in a country-month is less prone to measurement error than the exact count of events.

Our exposure variable of interest is a binary indicator for the month in which shutdown policies started. This variable is coded as 1 after states implement shutdown policies and 0 otherwise. Exposure to shutdown is a deterministic function of time. Once lockdowns begin, states are scored as being exposed to a shutdown policy, meaning that the status of being shut down is correlated with time itself. Therefore, a natural confounding variable is the underlying time trend in repression. To account for this potential confounder, we include a linear time trend and its interaction with shutdown to allow for a slope change in the repression time trend post-shutdown.

We are interested in the discrete change in repression levels that occurs during the month of a shutdown within a country. The sudden change represents the government's repression that is made possible by the emergency. Our empirical approach identifies a change in repression at the onset of shutdowns. If repression discontinuously increases on average within countries after lockdown, then state violence is responsive to exogenous emergencies.

Critically, we claim that lockdown policies are a type of public health measure that may give the state license to repress under the guise of public safety. This article does not argue that states always enact lockdown measures with the intention of repressing civilians. The choice to shut down the country was largely forced onto states because remaining open during March 2020 would have risked unmitigated spread of COVID-19 and, relatedly, international condemnation. Therefore, the onset of shutdowns represents an exogenous change in the operating environment.

To account for unobservable country traits, we include country fixed effects. Fixed effects adjust for unmeasured time-invariant confounds that could explain shutdown policies and repression in a country by removing the average level of repression and shutdown probability within a state. For example, country fixed effects control for factors such as regime type, colonial heritage, and legal system, all of which may influence patterns of state violence. Our estimation strategy compares countries both before and after they enact shutdown policies. Rather than identifying “early” or “late” adopters of shutdowns, our research design exploits variation in shutdown policies by comparing countries to themselves before and after shutdowns. Our aggregated approach (compared to the daily level of analysis), therefore, avoids the pitfalls of comparing repression trends among those countries that shut down only days before other countries. We report standard errors clustered by country.

We adjust our estimates to account for three threats to causal inference. First, other time-varying factors could explain the relationship between the timing of shutdowns and repression events. One obvious time-varying confound would be the prevalence and lethality of COVID-19 itself, both of which can influence how governments and citizens behave. Our assumption that the pre-trend of repression is a valid counterfactual for repression after shutdowns may depend on covariates. We include one-month lags of COVID-19 case levels and lethality from the Oxford COVID-19 Government Response Tracker.77 Further, we adjust for the lagged dependent variable to account for serial dependence in the outcome variable.

The second inferential threat is related to the timing of COVID-19, and how it overlaps with the change in weather from February to March. Exposure to the emergency potentially overlaps with a change between the dry and wet seasons, which is possibly associated with increased levels of opposition activism that could explain an increase in repression. To account for this possibility, online appendix A shows that our results are unchanged when we restrict our analysis to March–June 2019 as the counterfactual period.

The third inferential threat to our analysis is that our approach could mistake a change in levels of repression after shutdowns with a nonlinear relationship between time and repression. If the relationship between time and repression within countries follows a nonlinear pattern, what appears to be a jump in repression after shutdowns could instead reflect a disruption in the overall trend of violence. To adjust for this possibility, we include a squared term for time to account for nonlinearity.

The intuition of our design is visualized in figure 1a. The horizontal axis is a timeline that shows the months until COVID-19 lockdown, normalized to 0 for March 2020, and the vertical axis is the count of repression events recorded by ACLED. The vertical line shows the month COVID-19 lockdowns began. During COVID-19 lockdowns, repression levels in South Africa, Nigeria, Kenya, and Uganda had not been this high in years. Our selection to visualize these cases are illustrative. Indeed, as figure 1b shows, the general trend among African countries in our sample is that repression peaked in 2020 as lockdown stringency increased.

Monthly Repression Trend before and after COVID-19

Figure 1a.
Monthly Repression Trend before and after COVID-19

NOTE: Vertical axis plots the count of repression events from ACLED over time. Vertical line indicates the month of each country's respective shutdown. Vertical axis in figure 1a scaled differently than in figure 1b to illustrate trends. Illustrative countries selected on the basis of qualitative evidence surrounding human rights abuses amid the governments' COVID-19 responses.

Figure 1a.
Monthly Repression Trend before and after COVID-19

NOTE: Vertical axis plots the count of repression events from ACLED over time. Vertical line indicates the month of each country's respective shutdown. Vertical axis in figure 1a scaled differently than in figure 1b to illustrate trends. Illustrative countries selected on the basis of qualitative evidence surrounding human rights abuses amid the governments' COVID-19 responses.

Close modal

Monthly Repression Trend before and after COVID-19 in Africa

Figure 1b.
Monthly Repression Trend before and after COVID-19 in Africa

NOTE: Repression trends from ACLED data (thick line) and Stringency Index from Oxford COVID-19 Government Response Tracker (dashed line), rescaled for presentation. Data aggregated to daily level. The figure shows an increase in repression coinciding with an increase in response stringency. This finding is consistent with our theory of opportunistic repression, which predicts that emergencies license repression.

Figure 1b.
Monthly Repression Trend before and after COVID-19 in Africa

NOTE: Repression trends from ACLED data (thick line) and Stringency Index from Oxford COVID-19 Government Response Tracker (dashed line), rescaled for presentation. Data aggregated to daily level. The figure shows an increase in repression coinciding with an increase in response stringency. This finding is consistent with our theory of opportunistic repression, which predicts that emergencies license repression.

Close modal

We show baseline results in figure 2. Throughout the empirical results section, we present our findings by plotting the average difference in the outcome of interest post-lockdown as point estimates, and present 90 and 95 percent confidence intervals constructed from standard errors that allow serial correlation within countries around these estimates. The point estimates (i.e., circles, triangles, and squares in the figure) represent the relationship between lockdowns and the outcome(s) of interest, and the lines surrounding them characterize the uncertainty around each estimate.

Baseline Interrupted Time Series Results for Repression (Africa Sample)

Figure 2.
Baseline Interrupted Time Series Results for Repression (Africa Sample)

NOTE: ITS results from three different models. The baseline model includes an indicator for post–March 2020, a linear time trend, post–March 2020 interacted with the time trend, and country fixed effects. The covariate model includes lagged repression, lagged cases, lagged fatality rates. The quadratic model includes the time trend and its squared term interacted with the post indicator. Robust errors clustered at the country level used to construct 90 and 95 percent confidence intervals. See online appendix C for estimating equations.

Figure 2.
Baseline Interrupted Time Series Results for Repression (Africa Sample)

NOTE: ITS results from three different models. The baseline model includes an indicator for post–March 2020, a linear time trend, post–March 2020 interacted with the time trend, and country fixed effects. The covariate model includes lagged repression, lagged cases, lagged fatality rates. The quadratic model includes the time trend and its squared term interacted with the post indicator. Robust errors clustered at the country level used to construct 90 and 95 percent confidence intervals. See online appendix C for estimating equations.

Close modal

Panel A reports the outcome variable measured in levels and panel B reports a binary outcome. By levels, we mean the count of events that occur within a country (e.g., four repression events in a month), and by binary we mean a 1/0 measure for whether any repressive event was reported. We choose to report both sets of estimates because the occurrence of a repression event is less susceptible to measurement error than the exact number of repression events. We test three sets of models: a baseline equation including only country fixed effects, the time trend, an indicator for post-COVID, and the interaction of the post indicator with the time trend; a second model adds lagged repression, cases, and case lethality to the baseline; and a third model adds a squared term for time. Our results are robust to the addition of different covariates; the estimates remain very similar across specifications.

Our results for the levels estimates suggest that nearly two additional repression events occurred in the average country after COVID-19 shutdown orders. This increase corresponds to 0.63 of the pre-COVID standard deviation, which represents a substantively large spike in state violence. The average number of repression events per month before March 2020 was 1.63. A simple calculation suggests that there were nearly seventy-nine more repression events across the forty-nine African countries in our sample after shutdowns. These results are statistically significant at the 95 percent level, indicating high confidence that repression rose following the implementation of COVID-19 shutdown orders.

The binary (panel B) measures show that the probability of a repression event in a country increases by 17 percent in response to shutdown policies. The average probability of a repression event pre-COVID is 31 percent with a 0.46 standard deviation. Substantively, the binary outcome suggests a similar, and substantively large, impact of the COVID-19 emergency on state repression. Results are unchanged when adjusting for seasonality and when using a continuous measure for policy intensity (see online appendix A).

Next, we consider whether the change in the repression trend can be attributed to citizen dissent, as measured through public demonstrations. To reiterate, the increase in repression captured in figure 2 reflects instances in which the state engaged with unarmed civilians who were not participating in acts of public dissent. For the theories of responsive or preemptive repression to have explanatory power in this context, in the aftermath of lockdown policies one would expect to observe increased civilian demonstrations. An increase in such demonstration activity could catalyze the state to respond to immediate protest activities and would suggest that the crisis had advanced the capabilities of the domestic opposition to dissent. We use two different ACLED measures to test these alternative explanations: protests, during which citizens peacefully dissent; and riots, during which citizens violently dissent. We employ the terms protesters and rioters in this article for two reasons. First, these terms reflect the coding schema adopted by ACLED. For both protest and riot events, we code their occurrence using the “Event Type” variable in ACLED. This includes protest and riot events that are allowed to progress without interference, as well as those events in which protestors or rioters engaged with another political actor (e.g., the state, rebel groups, or political militias). Second, in this article, we are explicitly testing whether repression of civilians is in response to public displays of dissent. Our use of the terms protesters and rioters is not intended to suggest that those engaging in public protest (whether violent or nonviolent) are not civilians. We are not suggesting that by engaging in dissent protesters or rioters have forfeited some of their civil rights, nor are we condoning state repression of these demonstration activities.78 We display the results in figure 3. Panel A shows that the level of protests declines after shutdowns. Substantively, the decline in protests is not very large compared with the increase in repression. Relative to the pre-COVID period, the number of protests decreased by 8, or 0.34 of the pre-COVID standard deviation, which is nearly half of the standardized effect of shutdowns on repression. Consistent with this idea, panel B shows that the probability of protest is unchanged after March 2020. We find no increase in the level or probability of violent collective action in the form of riots (panels C and D).

Interrupted Time Series Results for Protests and Riots (Africa Sample)

Figure 3.
Interrupted Time Series Results for Protests and Riots (Africa Sample)

NOTE: ITS results from three different models. Baseline model includes an indicator for post–March 2020, a linear time trend, post–March 2020 interacted with the time trend, and country fixed effects. Covariate model includes lagged repression, lagged cases, lagged fatality rates. Quadratic model includes the time trend and its squared term interacted with the post indicator. Robust errors clustered at the country level used to construct 90 and 95 percent confidence intervals. See online appendix C for estimating equations.

Figure 3.
Interrupted Time Series Results for Protests and Riots (Africa Sample)

NOTE: ITS results from three different models. Baseline model includes an indicator for post–March 2020, a linear time trend, post–March 2020 interacted with the time trend, and country fixed effects. Covariate model includes lagged repression, lagged cases, lagged fatality rates. Quadratic model includes the time trend and its squared term interacted with the post indicator. Robust errors clustered at the country level used to construct 90 and 95 percent confidence intervals. See online appendix C for estimating equations.

Close modal

Although our cross-national test provides suggestive evidence that supports our theory of opportunistic repression, these results are incomplete. Measuring repression at the country level obscures rich subnational variation in state behavior, which can provide insight into who bears the brunt of opportunistic repression. The locations of repression events could reveal that the state has taken advantage of the pandemic to engage in targeted repression. Namely, if the state violence differential increases in localities that are less aligned with the regime, it may suggest that the increase in state violence after shutdown is differentially borne by the political opposition.

We test for this possibility by conducting an in-depth study of political repression in Uganda. Facing what promised to be a tight presidential election in 2021, particularly given the rise of the charismatic young musician-turned-opposition-leader Robert Kyagulanyi (also known as Bobi Wine), Uganda's sitting president Museveni may have faced incentives to use the state's coercive apparatus, the Ugandan Police Force in particular, to consolidate authority in opposition areas.79 After gaining power in 1986, Museveni and his ruling political party, the National Resistance Movement (NRM), have relied on repression to establish and maintain control.80 Acts of repression involve targeting opposition areas or members of ethnic groups affiliated with political opponents.81 Despite the excessive use of force against civilians by Ugandan security forces, the regime often justifies these acts as maintaining law and order or fighting terrorism.82

Qualitative evidence suggests that opportunistic repression is at play in several aspects of the COVID-19 response in Uganda.83 Politicized enforcement of bans on rallies, the design of COVID-19 relief programs, and the relationship between the state's response to the pandemic and the possibility of elections in 2021 all suggest that COVID-19 has enabled opportunistic repression in Uganda. Early in the pandemic, for example, Human Rights Watch reported that the Ugandan police shot two men on March 26 for “riding a motorcycle taxi in Mukono, outside of Kampala, despite the ban on motorcycle transport with multiple passengers.”84 That same day, members of the Local Defense Unit “used wires and sticks to beat people, including vendors selling fruit and vegetables and motorcycle riders, in downtown Kampala in an apparent attempt to punish non-compliance with the measures to close non-food markets.”85 In July 2020, the Kampala Metropolitan Deputy Police spokesperson justified using force against those rallying as a part of the People Power opposition movement, saying that participants “were blocking roads, in and around Kamwokya, which is also not allowed in the current situation where we are supposed to keep distance. So, we had to use a bit of tear gas to disperse them.”86 A protest led by activist Stella Nyanzi over the lockdown's disproportionate effect on the country's poor was also dispersed. The Kampala Metropolitan Deputy Police spokesperson claimed that Nyanzi “is exploiting the COVID-19 situation to advance her political motives.”87 Because of the incumbent's media influence, opposition parties and social movements cannot easily transition to remote campaigns.88 Even when the media tried to interview opposition candidates, the state physically blocked them. Bobi Wine, for example, was physically blocked by the army when trying to enter a radio station.89

Pairing partisan sticks with carrots, the public health response to the crisis was also designed to bolster the reputations of Museveni and the ruling NRM party leading up to the 2021 elections. Museveni banned food distributions—a common electoral strategy—following the COVID-19 lockdown, and when independent MP Francis Zaake was caught distributing food to his constituents, he was arrested and tortured while in police custody.90 Forbearance is reserved for members of Museveni's party; politicians affiliated with the NRM have continued to engage in food distributions. Furthermore, pandemic-related “task forces” throughout the country are headed by either Museveni appointees or those who report directly to the president.91 Innocent Anguyo suggests that these task forces designed food distribution as part of the NRM's political strategy in the period leading up to the election, and that distributions were determined by political factors, rather than designed to target those with the greatest need. COVID-19 seems to have heightened rather than mitigated the pre-pandemic habit of pairing patronage with coercion in urban areas ahead of elections.92

Despite the abuses that occurred during the COVID-19 lockdown, international criticism (aside from reports from human rights watchdogs) has been tepid. A Reuters report in August 2020 quoted the head of the Africa Centres for Disease Control and Prevention as saying, “Uganda knew where to conduct their surveillance. … The lesson from them is you should know your pandemic.”93 Other reports noted that Uganda was “widely lauded” for the stringency of its public health measures.94

Finally, Museveni's discussion about the feasibility of the election also suggests that the NRM politicized the pandemic to keep its grip on power. The president, who has been in power since 1986, suggested that holding elections in 2021 would be “madness,” which can be interpreted as a cloaked bid to expand his time in power.95

We test for politicized enforcement of the shutdown using a difference-indifferences (DD) design, which compares the difference in repression before and after lockdown in districts that are either less or more supportive of Museveni. To measure changes in repression in both opposition areas and incumbent strongholds, we use the month prior to lockdown as the reference period. Estimating changes in repression for months prior to and after lockdown allows us to identify both the pre-COVID-19 levels of repression across districts and the dynamics of repression after lockdown.

Figure 4 visualizes the intuition of our design. Panel A shows repression trends in Ugandan districts above and below the median level of support for Museveni in the 2016 election. The levels of repression look similar until March 2020, when the government substantially increased the stringency of its COVID-19 response, including bans on public transportation and restrictions on gatherings. After the policy change, the level of repression in districts that voted against Museveni at a higher margin in 2016 experienced a disproportionate increase in repression.96

Repression Trends in Uganda by District Support for Museveni

Figure 4.
Repression Trends in Uganda by District Support for Museveni

NOTE: Descriptive plot showing the repression trends in areas more and less opposed to Museveni (panel A) and the COVID-19 Policy Stringency (panel B) over time. Horizontal axis is the months until March 2020 (i.e., the beginning of shutdowns) and is marked by the vertical line labeled “shutdown begins.” Vertical axis is the count of repression events in panel A and the average COVID-19 Policy Stringency in panel B. “Opposed” is a binary indicator scored 1 when district vote share in favor of Museveni falls below the median, and 0 otherwise.

Figure 4.
Repression Trends in Uganda by District Support for Museveni

NOTE: Descriptive plot showing the repression trends in areas more and less opposed to Museveni (panel A) and the COVID-19 Policy Stringency (panel B) over time. Horizontal axis is the months until March 2020 (i.e., the beginning of shutdowns) and is marked by the vertical line labeled “shutdown begins.” Vertical axis is the count of repression events in panel A and the average COVID-19 Policy Stringency in panel B. “Opposed” is a binary indicator scored 1 when district vote share in favor of Museveni falls below the median, and 0 otherwise.

Close modal

case study results

We graphically display our results in figure 5. We use February 2020, the month prior to the shutdown, as the reference period. Each coefficient plotted is a DD estimate between opposed and aligned areas from this period, and the estimates represent the change in repression between opposed and aligned areas compared with February 2020. If the design assumptions are valid, we would not find a large difference in repression between opposed and aligned areas prior to lockdown. Meanwhile, after lockdown, we would expect a differential increase in repression in areas that opposed Museveni.

Difference-in-Differences Results for Repression in Opposition Districts in Uganda

Figure 5.
Difference-in-Differences Results for Repression in Opposition Districts in Uganda

NOTE: Event study plotting coefficients from regression of repression (binary) on opposition vote share interacted with time dummies (February 2020 as the reference) with time and district fixed effects. Each coefficient is a difference-in-differences estimate, computing the change in repression in opposed versus aligned areas from February 2020. Shaded region represents the post-lockdown period. Vertical dashed line shows the reference period, normalized to zero (since the change in February 2020 from February 2020 is 0). Robust standard errors clustered at the district level used to construct 95 percent confidence intervals. See online appendix C for estimating equations.

Figure 5.
Difference-in-Differences Results for Repression in Opposition Districts in Uganda

NOTE: Event study plotting coefficients from regression of repression (binary) on opposition vote share interacted with time dummies (February 2020 as the reference) with time and district fixed effects. Each coefficient is a difference-in-differences estimate, computing the change in repression in opposed versus aligned areas from February 2020. Shaded region represents the post-lockdown period. Vertical dashed line shows the reference period, normalized to zero (since the change in February 2020 from February 2020 is 0). Robust standard errors clustered at the district level used to construct 95 percent confidence intervals. See online appendix C for estimating equations.

Close modal

Pre-policy coefficients are plotted in the white space of the plot and show no evidence of a pre-trend in violence against civilians leading up to March 2020. In March 2020, however, violence against civilians from state forces significantly increased (shaded region). A standard deviation increase in opposition vote share corresponds to a 6.6 percentage point increase in the probability of state violence against civilians in March. The upper right corner shows the DD point estimate replacing the leads and lags with a binary indicator for post-shutdown. Results are consistent with the event study: a standard deviation increase in opposition vote share increases the probability of state violence by 5.4 percentage points on average after the government-imposed restrictions.

The DD estimates attenuate two months after lockdown and return to their February 2020 level. The result suggests that the increase in repression was transitory: the opportunity to repress more during lockdown did not result in a permanent shift toward more repression overall. Instead, repression spiked in the short term and began to subside.

alternative explanations for increased repression in opposition areas

The increase in repression that we detect may be because of other underlying factors that are correlated with regime support. For example, if Museveni's vote share is a function of underlying demographic or developmental factors, ethnicity or economic inequality may bias our estimates.

We include pre-treatment covariates interacted with year fixed effects to address potential confounding differences between districts. Specifically, we include gross domestic product proxied through nighttime lights to capture development,97 ethnic fractionalization to measure group heterogeneity,98 and population density to capture more urbanized areas. Results are shown in figure 6. The baseline estimate with time and district fixed effects is represented by the circle at the far left of the plot; the triangle-shaped estimate in the center of the plot includes the aforementioned controls. The size of the estimate attenuates but remains relatively stable and statistically precise. These models underline that the probability of a repressive event increases by roughly 30 percent in opposition districts in the aftermath of a shutdown.

Difference-in-Differences (Covariate Adjustment)

Figure 6.
Difference-in-Differences (Covariate Adjustment)

NOTE: Difference-in-differences results including covariate adjustment (year fixed effects interacted with pre-COVID district traits and region, respectively). Robust standard errors clustered at the district level used to construct 90 and 95 percent confidence intervals. See online appendix C for estimating equations.

Figure 6.
Difference-in-Differences (Covariate Adjustment)

NOTE: Difference-in-differences results including covariate adjustment (year fixed effects interacted with pre-COVID district traits and region, respectively). Robust standard errors clustered at the district level used to construct 90 and 95 percent confidence intervals. See online appendix C for estimating equations.

Close modal

The square estimate on the far right of figure 6 includes region by time fixed effects and pre-treatment covariates. Region by time fixed effects compare districts with different levels of support for Museveni that are in the same major region of Uganda in the same month. This approach helps guard against the possibility that the results are driven by regional trends, since districts that are more proximate to one another are more likely to share unobserved traits than districts in the opposite end of the country. The estimates are substantively similar.

responsive repression

The theory of responsive repression would predict that the patterns that we demonstrate are the result of the Ugandan government politically repressing those areas that are engaging in more dissent. For instance, areas that are less trusting of the central government may engage in more collective dissent against lockdown orders. If this were the case, then the repression that we observe would be responsive rather than opportunistic or preventive.

We test for this possibility using ACLED data on public demonstrations (either protest or riot events). These events capture the demand for responsive repression because they entail collective action from citizens. If we detect a similar trend for riots and protests as we do for state violence against civilians, it may suggest a pattern of responsive repression. Alternatively, if the results show no difference in protests and riots in opposition areas, they would suggest that the state is unilaterally using force irrespective of observed citizen dissent.

We plot our results in figure 7. We do not find evidence that state repression against protesters or rioters differed in opposition and incumbent districts. The DD estimate for protests is imprecisely estimated at 0.03, with −0.029 for riots. The negligible size of the point estimates combined with their statistical insignificance—even after we include covariates—suggests that incentives for responsive repression remained the same in opposed and aligned districts before and after the shutdown.

Riots and Protests Test

Figure 7.
Riots and Protests Test

NOTE: See online appendix C for estimating equations.

Figure 7.
Riots and Protests Test

NOTE: See online appendix C for estimating equations.

Close modal

These trends provide evidence against the alternative hypothesis that citizen malcontent or distrust in government caused repression to increase. If repression differentially increased in districts that were reluctant to comply with government mandates, one would expect to find more protests or riots decrying stay-at-home orders. Yet, we do not find any evidence that citizen dissident behavior differs over time among districts that mostly supported Museveni in the last election. Our theoretical framework provides an explanation for why this is the case: COVID-19 provided states with opportunities to repress their citizens at a lower cost. Furthermore, it is worth emphasizing that even if there were higher rates of noncompliance with COVID-19-related shutdown policies in opposition-controlled areas, this would not justify the use of force to compel adherence.99

preventive repression

Some scholars may suggest that the repression in Uganda during the COVID-19 pandemic is preemptive, reflecting the Ugandan government's anticipation of future dissent from opposition neighborhoods. Although it is nearly impossible to attribute intention in this instance, theories of preventive repression do not explain several empirical and theoretical characteristics of the Uganda case. Unlike other instances of preventive repression, the COVID-19 lockdown did not meaningfully increase the opposition's capacity to dissent, which is reflected in the pattern of public demonstrations. Rather, the costs of collective action increased because social distancing protocols and lockdown dynamics complicated in-person meetings, which are important features of opposition organizing. The COVID-19 lockdown reduced the capacity for the opposition in several concrete ways, including reducing opposition figures' ability to distribute food to citizens, a common approach to engaging with Ugandan citizens.100

Although it is possible that in the months following the lockdown the Ugandan government engaged in preventive repression in anticipation of criticism of its COVID-19 response, there are two reasons to question this assumption. First, Uganda has been applauded for adopting many pandemic-related international recommendations and for its response to the crisis. Thus, we argue that it is not credible to link the spread of the virus to the government's behavior, which could be the source of domestic dissent.101 Although Human Rights Watch has criticized the Ugandan government for “weaponiz[ing]” COVID-19-related regulations to repress the political opposition, attributing such repression to the government's fear of criticism over its repression is circular logic.102 Second, the relatively brief increase in the number of repression events suggests that this shift is attributable to lockdown policies rather than a more general pattern of state preelection violence.

If the government was not confronting a domestic opposition with meaningfully increased capacity to dissent and still engaged in repressive behavior, this would suggest the need to revisit both preventive and responsive theories of repression. We argue that our theory of opportunistic repression provides a more holistic and nuanced understanding of repression in the Uganda case because it shows how repression may arise opportunistically for governments that are facing moments of crisis.

In this article, we argue that the patterns exhibited in the aftermath of COVID-19-related lockdowns suggest that states are opportunistically engaging in repression against their civilians. The loosened norms regarding the acceptable level of state intervention into civilians' lives in response to a public health crisis were leveraged by incumbents to secure their positions.

We find that, across Africa, COVID-19 lockdowns were associated with increases in physical repression by the state. A case study of political repression in Uganda revealed that the brunt of the pandemic-era repression in that country was borne by opposition areas. These findings have implications for both the abstract study of political repression, as well as for the tangible and ongoing efforts to protect human rights and demand accountability for violations.

First, we have contributed to the study of state repression by introducing the concept of opportunistic repression. In contrast to existing theories of repression that focus on the activities or capacity of the domestic opposition, opportunistic repression refers to instances in which shifts in the state's ability to legitimately intervene in citizens' lives are manipulated to repress the domestic opposition. Analyzing how governments deploy these new authorities and against whom is a pressing concern.

Our theory of opportunistic repression underlines how crises generate threats that are both immediate (i.e., to citizens' security and health) and secondary (by generating opportunities for government repression). This finding has implications for how the international community helps design and support interventions for states in crises, whether the threat emerges from a natural disaster, a public health crisis, or terrorism. Politics are not paused in a time of crisis. Although a comforting fiction, we argue that assuming that politics are paused is not a suitable foundation for approaching crisis response. Opportunistic repression provides a helpful analytical lens because it shows how changes in international norms brought about by emergencies may permit states to repress their citizens. Policymakers will hopefully use this insight to produce safeguards to deter such repression. Furthermore, more effort must be placed into holding accountable those states that clearly and repeatedly violate the international community's guidance on how to enforce public health measures in a manner that respects citizens' rights.

Second, our empirical investigation on a pressing topic—the impacts of the COVID-19 pandemic on civilian well-being—makes an important contribution to the literature. Understanding the fallout from COVID-19 in the short term is important for the aid agencies and human rights organizations that are developing programs to respond to the lingering effects of the pandemic. The pandemic's long-term political effects remain unknown, however. For example, do our findings that repression in Uganda began to fade two months after the lockdown indicate that opportunistic repression is short-lived, or that states need only a brief window of opportunity to consolidate their authority? The legacies of repression are a burgeoning field of study, which suggests that there are long-term effects for the ways in which repression affects political preferences.103 Repression associated with the COVID-19 crisis could well be a critical juncture that alters how the international community views abuses by states, and it remains unclear whether this particular emergency represents a permanent or transitory shock to state coercion and autocratic power. Additional research on the persistence of power grabs during emergencies, potentially using historical case studies, could provide important insights for scholars and practitioners as they weigh the more long-term consequences of repression in the wake of COVID-19.

The authors contributed equally to this article. They are grateful for comments and suggestions from Daniel Arnon, Benjamin Valentino, the MacMillan Political Violence and its Legacies workshop at Yale University, and several anonymous reviewers. An online appendix is available at doi.org/10.7910/DVN/U8LYW1.

1.

Giorgio Agamben, “State of Exception,” Nova srpska politicka misao [New Serbian political thought], Vol. 12, No. 1–4 (2005), pp. 135–145; Tiberiu Dragu and Mattias Polborn, “The Rule of Law in the Fight against Terrorism,” American Journal of Political Science, Vol. 58, No. 2 (April 2014), pp. 511–525, doi.org/10.1111/ajps.12061; and Tiberiu Dragu and Xiaochen Fan, “Self-Enforcing Legal Limits: Bureaucratic Constraints on Repression under Emergency Powers,” Journal of Politics, Vol. 82, No. 2 (April 2020), pp. 687–699, doi.org/10.1086/706981.

2.

Lorraine Mazerolle and Janet Ransley, “Policing Health Regulations in Democratic Societies: A Focus on COVID-19 Challenges and Opportunities in Australia,” International Journal of Comparative and Applied Criminal Justice, Vol. 45, No. 3 (2021), pp. 315–327, doi.org/10.1080/01924036.2021.1907605; and Joan Barceló et al., “Windows of Repression: Using COVID-19 Policies against Political Dissidents?” SocArXiv (Baltimore: University of Maryland, November 12, 2020), https://osf.io/preprints/socarxiv/yuqw2/

3.

Barceló et al., “Windows of Repression.”

4.

Nicolas Berman et al., “Shutdown Policies and Worldwide Conflict,” Economic Science Institute Working Paper 20-16 (Irvine, Calif.: Chapman University, 2020); Jeffrey R. Bloem and Colette Salemi, “COVID-19 and Conflict,” World Development, Vol. 140 (April 2021), doi.org/10.1016/j.worlddev.2020.105294; Marius Mehrl and Paul W. Thurner, “The Effect of the Covid-19 Pandemic on Global Armed Conflict: Early Evidence,” Political Studies Review, Vol. 19, No. 2 (May 2021), pp. 286–293, doi.org/10.1177/1478929920940648; and Kim Yi Dionne and Fulya Felicity Turkmen, “The Politics of Pandemic Othering: Putting COVID-19 in Global and Historical Context,” International Organization, Vol. 74, No. S1 (December 2020), pp. E213–E230, doi.org/10.1017/S0020818320000405.

5.

Benjamin E. Bagozzi, “On Malaria and the Duration of Civil War,” Journal of Conflict Resolution, Vol. 60, No. 5 (August 2016), pp. 813–839, doi.org/10.1177%2F0022002714550202; Ada González-Torres and Elena Esposito, “Epidemics and Conflict: Evidence from the Ebola Outbreak in Western Africa,” SSRN (2016), doi.org/10.2139/ssrn.3544606; and Remi Jedwab, Noel D. Johnson, and Mark Koyama, “Negative Shocks and Mass Persecutions: Evidence from the Black Death,” Journal of Economic Growth, Vol. 24 (2019), pp. 345–395, doi.org/10.1007/s10887-019-09167-1.

6.

Christian Bj⊘rnskov and Stefan Voigt, “When Does Terror Induce a State of Emergency? And What Are the Effects?” Journal of Conflict Resolution, Vol. 64, No. 4 (April 2020), pp. 579–613, doi.org/10.1177/0022002719865994.

7.

Sabine C. Carey, “The Dynamic Relationship between Protest and Repression,” Political Research Quarterly, Vol. 59, No. 1 (March 2006), pp. 1–11, doi.org/10.1177/106591290605900101; Will H. Moore, “The Repression of Dissent: A Substitution Model of Government Coercion,” Journal of Conflict Resolution, Vol. 44, No. 1 (February 2000), pp. 107–127, doi.org/10.1177/0022002700044001006; Mark Irving Lichbach, “Deterrence or Escalation? The Puzzle of Aggregate Studies of Repression and Dissent,” Journal of Conflict Resolution, Vol. 31, No. 2 (June 1987), pp. 266–297, doi.org/10.1177/0022002787031002003; Conway W. Henderson, “Conditions Affecting the Use of Political Repression,” Journal of Conflict Resolution, Vol. 35, No. 1 (March 1991), pp. 120–142, doi.org/10.1177/0022002791035001007; Ted Robert Gurr, Political Rebellion: Causes, Outcomes, and Alternatives (New York: Routledge, 2019); and Ragnhild Nordås and Christian Davenport, “Fight the Youth: Youth Bulges and State Repression,” American Journal of Political Science, Vol. 57, No. 4 (October 2013), pp. 926–940, doi.org/10.1111/ajps.12025.

8.

We use the terms preventive repression and preemptive repression interchangeably, as they broadly reflect a similar logic: The state acts first in response to political challengers.

9.

Emily Hencken Ritter and Courtenay R. Conrad, “Preventing and Responding to Dissent: The Observational Challenges of Explaining Strategic Repression,” American Political Science Review, Vol. 110, No. 1 (2016), pp. 85–99, doi.org/10.1017/S0003055415000623.

10.

Tiberiu Dragu and Adam Przeworski, “Preventive Repression: Two Types of Moral Hazard,” American Political Science Review, Vol. 113, No. 1 (2019), pp. 77–87, doi.org/10.1017/S0003055418000552.

11.

Sabine C. Carey, “The Use of Repression as a Response to Domestic Dissent,” Political Studies, Vol. 58, No. 1 (February 2010), pp. 167–186, doi.org/10.1111/j.1467-9248.2008.00771.x.

12.

Christian Davenport, “Licensing Repression: Dissent, Threats, and State Repression in the United States,” Minnesota Journal of International Law, Vol. 16, No. 2 (2007), pp. 311–333, https://scholarship.law.umn.edu/mjil/232; Emilie M. Hafner-Burton and Kiyoteru Tsutsui, “Human Rights in a Globalizing World: The Paradox of Empty Promises,” American Journal of Sociology, Vol. 110, No. 5 (March 2005), pp. 1373–1411, doi.org/10.1086/428442; Emilie M. Hafner-Burton, “Sticks and Stones: Naming and Shaming the Human Rights Enforcement Problem,” International Organization, Vol. 62, No. 4 (October 2008), pp. 689–716, doi.org/10.1017/S0020818308080247; James Raymond Vreeland, “Political Institutions and Human Rights: Why Dictatorships Enter into the United Nations Convention against Torture,” International Organization, Vol. 62, No. 1 (January 2008), pp. 65–101, doi.org/10.1017/S002081830808003X; Oona A. Hathaway, “Why Do Countries Commit to Human Rights Treaties?” Journal of Conflict Resolution, Vol. 51, No. 4 (August 2007), pp. 588–621, doi.org/10.1177%2F0022002707303046; and Nic Cheeseman and Brian Klaas, How to Rig an Election (New Haven, Conn.: Yale University Press, 2018).

13.

González-Torres and Esposito, “Epidemics and Conflict.”

14.

Berman et al., “Shutdown Policies and Worldwide Conflict.”

15.

Tiberiu Dragu and Yonatan Lupu, “Collective Action and Constraints on Repression at the Endgame,” Comparative Political Studies, Vol. 51, No. 8 (2018), pp. 1042–1073, doi.org/10.1177/0010414017730077.

16.

David R. Mayhew, Congress: The Electoral Connection, 2nd ed. (New Haven, Conn.: Yale University Press, 2004); Milan W. Svolik, The Politics of Authoritarian Rule (Cambridge: Cambridge University Press, 2012); Jennifer Gandhi, Political Institutions under Dictatorship (Cambridge: Cambridge University Press, 2008); and Beatriz Magaloni, Voting for Autocracy: Hegemonic Party Survival and Its Demise in Mexico (Cambridge: Cambridge University Press, 2006).

17.

Christian Davenport, “State Repression and Political Order,” Annual Review of Political Science, Vol. 10 (2007), p. 7, doi.org/10.1146/annurev.polisci.10.101405.143216.

18.

Ibid., pp. 1–23.

19.

Carey, “The Use of Repression as a Response to Domestic Dissent.”

20.

Jennifer Earl, Sarah A. Soule, and John D. McCarthy, “Protest under Fire? Explaining the Policing of Protest,” American Sociological Review, Vol. 68, No. 4 (August 2003), pp. 581–606, doi.org/10.2307/1519740.

21.

Tavishi Bhasin and Jennifer Gandhi, “Timing and Targeting of State Repression in Authoritarian Elections,” Electoral Studies, Vol. 32, No. 4 (December 2013), pp. 620–631, doi.org/10.1016/j.electstud.2013.07.011; Christopher M. Sullivan, “Undermining Resistance: Mobilization, Repression, and the Enforcement of Political Order,” Journal of Conflict Resolution, Vol. 60, No. 7 (October 2016), pp. 1163–1190, doi.org/10.1177/0022002714567951; and Rory Truex, “Focal Points, Dissident Calendars, and Preemptive Repression,” Journal of Conflict Resolution, Vol. 63, No. 4 (April 2019), pp. 1032–1052, doi.org/10.1177/0022002718770520.

22.

Bhasin and Gandhi, “Timing and Targeting of State Repression.”

23.

Truex, “Focal Points.”

24.

Janina Beiser-McGrath, “Targeting the Motivated? Ethnicity and the Pre-emptive Use of Government Repression,“ Swiss Political Science Review, Vol. 25, No. 3 (2019), pp. 203–225, doi.org/10.1111/spsr.12370.

25.

Sullivan, “Undermining Resistance.”

26.

Truex, “Focal Points.”

27.

Ibid.

28.

Consider, for example, how governments facing a “youth bulge” are more likely to engage in repression. Nordås and Davenport, “Fight the Youth.” Another study finds that autocrats strategically shuffle their security apparatus to resolve principal-agent challenges of repression. Mai Hassan, “The Strategic Shuffle: Ethnic Geography, the Internal Security Apparatus, and Elections in Kenya,” American Journal of Political Science, Vol. 61, No. 2 (April 2017), pp. 382–395, https://www.jstor.org/stable/26384738.

29.

Peter D. Carey II et al., “Oil Discoveries, Civil War, and Preventive State Repression,” Journal of Peace Research (forthcoming).

30.

Davenport, “Licensing Repression.”

31.

Ibid.

32.

We leave for future study the possibility that opportunistic repression may arise from domestic crises. Some scholars have covered this phenomenon, such as: Davenport, “Licensing Repression”; and Jane Esberg, “The Audience of Repression: Killings and Disappearances in Pinochet's Chile,” SSRN (2018), doi.org/10.2139/ssrn.3246120.

33.

Martha Finnemore and Kathryn Sikkink, “Taking Stock: The Constructivist Research Program in International Relations and Comparative Politics,” Annual Review of Political Science, Vol. 4, No. 1 (June 2001), pp. 391–416, doi.org/10.1146/annurev.polisci.4.1.391.

34.

UN General Assembly, International Covenant on Civil and Political Rights, Treaty Series, Vol. 999 (New York: United Nations, December 16, 1966), p. 171, https://www.refworld.org/docid/3ae6b3aa0.html.

35.

John Reynolds, Empire, Emergency, and International Law (Cambridge: Cambridge University Press, 2017).

36.

Wesley W. Widmaier, Mark Blyth, and Leonard Seabrooke, “Exogenous Shocks or Endogenous Constructions? The Meanings of Wars and Crises,” International Studies Quarterly, Vol. 51, No. 4 (December 2007), pp. 747–759, esp. 757, doi.org/10.1111/j.1468-2478.2007.00474.x.

37.

Hathaway, “Why Do Countries Commit to Human Rights Treaties?”; and Emilie M. Hafner-Burton, “Trading Human Rights: How Preferential Trade Agreements Influence Government Repression,” International Organization, Vol. 59, No. 3 (July 2005), pp. 593–629, doi.org/10.1017/S0020818305050216.

38.

Jacqueline H.R. DeMeritt and Courtenay R. Conrad, “Repression Substitution: Shifting Human Rights Violations in Response to UN Naming and Shaming,” Civil Wars, Vol. 21, No. 1 (2019), pp. 128–152, doi.org/10.1080/13698249.2019.1602805.

39.

Mirjam Edel and Maria Josua, “How Authoritarian Rulers Seek to Legitimize Repression: Framing Mass Killings in Egypt and Uzbekistan,” Democratization, Vol. 25, No. 5 (2018), pp. 882–900, doi.org/10.1080/13510347.2018.1439021.

40.

Colton Heffington, “External Threat and Human Rights: How International Conflict Leads to Domestic Repression,” Journal of Human Rights, Vol. 20, No. 1 (2021), pp. 2–19, doi.org/10.1080/14754835.2020.1803052; and Nathan Danneman and Emily Hencken Ritter, “Contagious Rebellion and Preemptive Repression,” Journal of Conflict Resolution, Vol. 58, No. 2 (March 2014), pp. 254–279, doi.org/10.1177/0022002712468720.

41.

Heffington, “External Threat and Human Rights.”

42.

Danneman and Hencken Ritter, “Contagious Rebellion and Preemptive Repression.”

43.

Steven Heydemann and Reinoud Leenders, “Authoritarian Learning and Authoritarian Resilience: Regime Responses to the ‘Arab Awakening,‘” Globalizations, Vol. 8, No. 5 (2011), pp. 647–653, doi.org/10.1080/14747731.2011.621274.

44.

Stephen G.F. Hall and Thomas Ambrosio, “Authoritarian Learning: A Conceptual Overview,” East European Politics, Vol. 33, No. 2 (2017), pp. 143–161, doi.org/10.1080/21599165.2017.1307826.

45.

Christian Davenport, “State Repression and the Tyrannical Peace,” Journal of Peace Research, Vol. 44, No. 4 (July 2007), pp. 485–504, doi.org/10.1177/0022343307078940. Furthermore, there is evidence of a complex relationship between regime type and the influence of international factors. See Daniel W. Hill Jr. and Zachary M. Jones, “An Empirical Evaluation of Explanations for State Repression,” American Political Science Review, Vol. 108, No. 3 (August 2014), pp. 661–687, doi.org/10.1017/S0003055414000306.

46.

Edel and Josua, “How Authoritarian Rulers Seek to Legitimize Repression,” p. 893.

47.

Dragu and Fan, “Self-Enforcing Legal Limits,” p. 688.

48.

Davenport, “Licensing Repression,” p. 311; and Esberg, “The Audience of Repression.”

49.

Barceló et al., “Windows of Repression.”

50.

Sheena Chestnut Greitens, Myunghee Lee, and Emir Yazici, “Counterterrorism and Preventive Repression: China's Changing Strategy in Xinjiang,” International Security, Vol. 44, No. 3 (Winter 2019/20), pp. 9–47, doi.org/10.1162/isec_a_00368.

51.

Max Bergmann and Alexandra Schmitt, “A Plan to Reform U.S. Security Assistance,” Center for American Progress, March 9, 2021, https://www.americanprogress.org/issues/security/reports/2021/03/09/496788/plan-reform-u-s-security-assistance/.

52.

Ibid.

53.

Edel and Josua, “How Authoritarian Rulers Seek to Legitimize Repression,” p. 893.

54.

Davenport, “Licensing Repression,” p. 311.

55.

Melissa Pavlik, “A Great and Sudden Change: The Global Political Violence Landscape before and after the COVID-19 Pandemic” (Grafton, Wis.: ACLED, August 4, 2020), https://acleddata.com/2020/08/04/a-great-and-sudden-change-the-global-political-violence-landscape-before-and-after-the-covid-19-pandemic/.

56.

Roudabeh Kishi et al., “ACLED 2020: The Year in Review” (Grafton, Wis.: ACLED, March 2021), https://acleddata.com/acleddatanew/wp-content/uploads/2021/03/ACLED_AnnualReport2020_WebMar2021_PubUpd.pdf.

57.

Barceló et al., “Windows of Repression.”

58.

We thank the anonymous reviewers for this point.

59.

“Covid-19 Triggers Wave of Free Speech Abuse,” Human Rights Watch, February 11, 2021, https://www.hrw.org/news/2021/02/11/covid-19-triggers-wave-free-speech-abuse#.

60.

“Emergency Measures and COVID-19: Guidance” (Geneva: United Nations Human Rights Office of the High Commissioner, April 27, 2020), https://www.ohchr.org/Documents/Events/EmergencyMeasures_Covid19.pdf. There were also calls for proportionality in the response to violations of these restrictions.

61.

Michelle Bachelet, “COVID-19: Exceptional Measures Should Not Be Cover for Human Rights Abuses and Violations“ (Geneva: United Nations Human Rights Office of the High Commissioner, April 27, 2020), https://www.ohchr.org/EN/NewsEvents/Pages/DisplayNews.aspx?NewsID=25828.

62.

“COVID-19 Crackdowns: Police Abuse and the Global Pandemic” (London: Amnesty International, December 17, 2020), https://www.amnesty.org/en/documents/act30/3443/2020/en/.

63.

Mehdi Khawaja, “Modi's Government Is Exploiting the Pandemic to Ramp up Repression in Kashmir,” Jacobin, May 3, 2020, https://www.jacobinmag.com/2020/05/india-modi-coronavirus-police-repression-kashmir.

64.

Dionne and Turkmen, “The Politics of Pandemic Othering.”

65.

“‘Toxic Lockdown Culture’ of Repressive Coronavirus Measures Hits Most Vulnerable,” UN News, April 27, 2020, https://news.un.org/en/story/2020/04/1062632.

66.

“Emergency Measures and COVID-19: Guidance,” April 27, 2020.

67.

“Covid-19 Triggers Wave of Free Speech Abuse,” February 11, 2021.

68.

Ibid.

69.

Robert A. Blair, Benjamin S. Morse, and Lily L. Tsai, “Public Health and Public Trust: Survey Evidence from the Ebola Virus Disease Epidemic in Liberia,” Social Science & Medicine, Vol. 172 (January 2017), pp. 89–97, doi.org/10.1016/j.socscimed.2016.11.016; and “In the Face of Coronavirus, African Countries Apply Lessons from Ebola Response,” World Bank, April 3, 2020, https://www.worldbank.org/en/news/feature/2020/04/03/in-the-face-of-coronavirus-african-countries-apply-lessons-from-ebola-response.

70.

“In the Face of Coronavirus,” April 3, 2020. See also Ciara Staunton, Carmen Swanepoel, and Melodie Labuschaigne, “Between a Rock and a Hard Place: COVID-19 and South Africa's Response,” Journal of Law and the Biosciences, Vol. 7, No. 1 (January–June 2020), doi.org/10.1093/jlb/lsaa052.

71.

Clionadh Raleigh et al., “Introducing ACLED: An Armed Conflict Location and Event Dataset: Special Data Feature,” Journal of Peace Research, Vol. 47, No. 5 (September 2010), pp. 651–660, doi.org/10.1177/0022343310378914.

72.

Additionally, two of the authors of this paper, Hilary Matfess and Melissa Pavlik, have previously worked as analysts for ACLED.

73.

Clionadh Raleigh and Roudabeh Kishi, “Comparing Conflict Data: Similarities and Differences Across Conflict Datasets,” (Grafton, Wis.: ACLED, August 2019), https://www.acleddata.com/wp-content/uploads/2019/09/ACLED-Comparison_8.2019.pdf.

74.

Ibid.

75.

For example, using months in 2005 may not capture a valid counterfactual for what repression would have looked like in a given country in 2020 in the absence of COVID-19 restrictions.

76.

For a more applied discussion of this source, see Berman et al., “Shutdown Policies and Worldwide Conflict”; and Thomas Hale et al., “A Global Panel Database of Pandemic Policies (Oxford COVID-19 Government Response Tracker),” Nature Human Behaviour, Vol. 5 (2021), pp. 529–538, doi.org/10.1038/s41562-021-01079-8.

77.

Hale et al., “A Global Panel Database of Pandemic Policies.”

78.

We thank the members of the MacMillan Political Violence and its Legacies workshop at Yale University for raising this point.

79.

Travis B. Curtice, “How Repression Affects Public Perceptions of Police: Evidence from a Natural Experiment in Uganda,” Journal of Conflict Resolution (2021), doi.org/10.1177/00220027211013097.

80.

Travis B. Curtice and Brandon Behlendorf, “Street-Level Repression: Protest, Policing, and Dissent in Uganda,” Journal of Conflict Resolution, Vol. 65, No. 1 (January 2021), pp. 166–194, doi.org/10.1177/0022002720939304.

81.

Max Bearak, “Ugandan Opposition Reels from Widespread Repression ahead of Thursday's Election,” Washington Post, January 12, 2021, https://www.washingtonpost.com/world/africa/uganda-election-opposition/2021/01/12/187af55c-54c5-11eb-acc5-92d2819a1ccb_story.html.

82.

For a discussion of counterterrorism policies in East Africa, see Beth Elise Whitaker, “Compliance among Weak States: Africa and the Counter-Terrorism Regime,” Review of International Studies, Vol. 36, No. 3 (July 2010), pp. 639–662, doi.org/10.1017/S0260210510000641.

83.

Elizabeth Katana et al., “Violence and Discrimination among Ugandan Residents during the COVID-19 Lockdown,” BMC Public Health, Vol. 21 (2021), doi.org/10.1186/s12889-021-10532-2. It is important to note that this survey is skewed toward urban respondents.

84.

“Uganda: Respect Rights in COVID-19 Response,” Human Rights Watch, April 2, 2020, https://www.hrw.org/news/2020/04/02/uganda-respect-rights-covid-19-response.

85.

Ibid.

86.

Halima Athumani, “Ugandan Police Tear Gas Opposition People Power Supporters,” Voice of America, July 17, 2020, https://www.voanews.com/africa/ugandan-police-tear-gas-opposition-people-power-supporters.

87.

“Uganda Arrests Stella Nyanzi at Protest over Coronavirus Response,” Al Jazeera, May 19, 2020, https://www.aljazeera.com/news/2020/5/19/uganda-arrests-stella-nyanzi-at-protest-over-coronavirus-response.

88.

Martina Schwikowski, “Uganda's Unequal Political Campaigns during COVID-19 Times,” Deutsche Welle, August 8, 2020, https://www.dw.com/en/ugandas-unequal-political-campaigns-during-covid-19-times/a-54488099.

89.

Ibid.

90.

Innocent Anguyo, “The Politics of Food Relief in Uganda's COVID-19 Era,” Africa at LSE blog, July 1, 2020, https://blogs.lse.ac.uk/africaatlse/2020/07/01/politics-food-relief-aid-uganda-c19/.

91.

Ibid.

92.

Ibid.

93.

Elias Biryabarema, “Uganda's Tough Approach Curbs COVID, Even as Africa Nears 1 Million Cases,” Reuters, August 5, 2020, https://www.reuters.com/article/us-health-coronavirus-uganda/ugandas-tough-approach-curbs-c-even-as-africa-nears-1-million-cases-idUSKCN251159.

94.

Evelyn Lirri, “How Uganda's Tough Approach to C-19 Is Hurting Its Citizens,” Telegraph, August 26, 2020, https://www.telegraph.co.uk/global-health/science-and-disease/ugandas-tough-approach-covid-19-hurting-citizens/.

95.

Elias Biryabarema, “‘Madness’ to Hold Uganda Vote If Virus Persists: Museveni,” Reuters, May 12, 2020, https://www.reuters.com/article/us-health-coronavirus-uganda-politics/madness-to-hold-uganda-vote-if-virus-persists-museveni-idUSKBN22O10R.

96.

In figure B1 in the online appendix, doi.org/10.7910/DVN/U8LYW1, we map the increase in repression before and after shutdown policies in March 2020 against Museveni's 2016 vote share.

97.

Tilottama Ghosh et al., “Shedding Light on the Global Distribution of Economic Activity,” Open Geography Journal, Vol. 3 (2010), pp. 148–160, doi.org/10.2174/1874923201003010147.

98.

Carl Müller-Crepon and Philipp Hunziker, “New Spatial Data on Ethnicity: Introducing SIDE,” Journal of Peace Research, Vol. 55, No. 5 (September 2018), pp. 687–698, doi.org/10.1177/0022343318764254.

99.

One study, for example, shows that public compliance with COVID-19 requirements in Uganda is positively associated with trust in the government, including the police. See Robert A. Blair et al., “Public Trust, Policing, and the COVID-19 Pandemic: Evidence from an Electoral Authoritarian Regime,” Center for Effective Global Action (CEGA) Working Paper Series No. 164 (Berkeley: University of California Berkeley CEGA, 2021), doi.org/10.26085/C3D01C.

100.

Anguyo, “The Politics of Food Relief in Uganda's COVID-19 Era.”

101.

Ahmed M. Sarki, Alex Ezeh, and Saverio Stranges, “Uganda as a Role Model for Pandemic Containment in Africa,” American Journal of Public Health, Vol. 110, No. 12 (December 2020), pp. 1800–1802, doi.org/10.2105/AJPH.2020.305948; and Lirri, “How Uganda's Tough Approach to Covid-19 Is Hurting Its Citizens.” In contrast, consider the spread of COVID-19 in Tanzania, where the government under President John Magufuli resisted adopting lockdown measures. Nicholas Bariyo, “Tanzania Shunned Lockdowns. Now It's Rejecting Covid-19 Vaccines,” Wall Street Journal, February 3, 2021, https://www.wsj.com/articles/tanzania-shunned-lockdowns-now-its-rejecting-covid-19-vaccines-11612364606.

102.

“Uganda: Authorities Weaponize Covid-19 for Repression,” Human Rights Watch, November 20, 2020, https://www.hrw.org/news/2020/11/20/uganda-authorities-weaponize-covid-19-repression.

103.

Arturas Rozenas, Sebastian Schutte, and Yuri Zhukov, “The Political Legacy of Violence: The Long-Term Impact of Stalin's Repression in Ukraine,” Journal of Politics, Vol. 79, No. 4 (2017), pp. 1147–1161, doi.org/10.1086/692964.