Through a wide array of actions that include harassment, intimidation, displacement, and physical attacks, political actors have employed violence to influence voter turnout and vote choice. By exploring patterns of electoral violence in Africa, this chapter evaluates the scope and character of Africa’s electoral violence; key conclusions from emerging datasets on incidences of electoral violence; various political strategies for the use of violence; the temporal nature of electoral violence; and the different efforts international and domestic actors undertake to prevent election violence. Further research is needed to refine the measurement of electoral violence, better predict its occurrence, and fully understand its consequences.
Political violence, commonly considered “organized violence carried out by political actors, including rebel groups, insurgents, or terrorist organizations,” targets both civilians and armed combatants (Valentino 2014, 91). For a long time, the literature focused on political violence during civil wars and the harm committed against civilians. However, since around 2010, scholars and policymakers have come to recognize the significant role violence plays in elections, especially in new democracies with weak or biased institutions. Now, nine years later, studies of electoral violence in third wave democracies and older academic studies, such as Fischer’s (2002) seminal work on electoral conflict and violence, create a robust field of study.
Electoral violence can be distinguished from other forms of political violence by its timing (before an election, on election day, or after an election); objectives (winning, sabotaging, preventing, altering, or challenging election outcomes); and methodology (it may be physical, verbal, or psychological). Electoral violence functions as one of many tools used by political actors to acquire or retain power. While some political actors use electoral violence to strategically winnow the field of competition or to coerce voters, it may also occur for incidental or non-strategic reasons (Burchard 2015). The decision to use violence or the unplanned eruption of violence depends largely on the local political context and the integrity of critical institutions. For example, countries where some communities view critical institutions, such as the judiciary or security forces, as biased may suffer more electoral violence, as those avenues for adjudication appear to be closed off.
Another important distinction when it comes to studies of electoral violence, as compared to analyses of the broader category of political violence, is the array of harmful actions it tends to encompass. While analysts of political violence tend to focus on civilian fatalities, in the majority of electoral violence studies, analysts highlight how citizens suffer from a wide range of actions including harassment, intimidation, detention, and displacement by political actors seeking to influence voter turnout and vote choice. The emerging analysis of gender and electoral violence reveal even broader categories of injuries, to include social-psychological harm and economic hardship (Bardall 2011).
This chapter explores the patterns of election violence in Africa and the conditions under which electoral violence emerges. The following sections will evaluate the scope and quality of Africa’s electoral violence, conclusions from emerging datasets, political strategies for using violence, the temporal nature of electoral violence, and the various efforts international and domestic actors employ to prevent election violence.
Several different sources provide data on electoral violence. Our chapter relies on the African Electoral Violence Database (AEVD), first compiled by Straus and Taylor (2012) for all national-level elections (legislative and executive) held in sub-Saharan Africa from 1990 to 2008. It was the first database to record and categorize levels of electoral violence in Africa. The AEVD measures electoral violence on an ordinal scale from one to four, with one representing a peaceful election and four representing an election in which widespread violence and more than twenty fatalities occurred. For their coding, Straus and Taylor relied upon US State Department Human Rights reports. Using their criteria and coding scheme, we updated the AEVD through 2017, to include a total of 333 elections. 1 Straus and Taylor consider the pre-election period to be the six months prior to election day and the post-election period to include the three months after election day. This coding decision makes longitudinal and cross-national comparison possible, and helps distinguish electoral violence from more general political violence, although it may overlook key events that occur outside this period. Nonetheless, we are confident that this nine-month period captures the vast majority of election violence.
A majority of Africa’s elections—roughly 65 percent—experience some form of electoral violence, ranging from threats and voter intimidation to physical conflict leading to mass fatalities (Figure 18.1). Fortunately, among the 216 violent elections, extreme violence is not the norm; 60 percent of electoral violence is low-intensity, which comprises incidents such as verbally harassing voters, forcefully discouraging aspirants from running for office, tearing down or otherwise destroying election materials, rioting, arresting candidates, and shuttering critical press outlets. In about 22 percent of violent elections, security services repeatedly detain opposition candidates and supporters, often for long periods of time, and political assassinations occur with a limited number of fatalities. In the most severe instances, which accounts for around 18 percent of all violent elections, widespread violence breaks out, with fatalities numbering anywhere from the dozens to the thousands, and a sizable number displaced from their homes. In Kenya in 2007 and Côte d’Ivoire in 2010, two of the worst examples of election violence in recent history, thousands died (approximately 1,100 and 3,000 respectively) and hundreds of thousands were displaced (approximately 700,000 and 1,000,000 respectively) (Freedom House 2012; Lynch 2013). Legislative elections experience less violence than presidential elections: an estimated 46 percent of legislative elections held in Africa experience violence compared to 61 percent of presidential elections and 70 percent of concurrent elections (where presidential and legislative elections take place at the same time).
Figure 18.1 Electoral violence in sub-Saharan Africa, 1990–2017
Violence can be used strategically, but its occurrence can also be incidental or opportunistic. Political actors use strategic violence to intimidate aspirants and deter them from running for office. They can also use it in an attempt to influence voters’ ability or decision to turn out, their candidate selection, or to punish those perceived to have voted in the wrong way. Incidental electoral violence has no specific prior motive and requires no advanced planning, but that does not mean it cannot be anticipated or prevented. Campaign rallies or protests can rapidly turn violent when police, the military, or members of the presidential guard try to stop these demonstrations, responding with excessive force. Finally, when strategic or incidental violence occurs, opportunistic violence often follows as individuals take advantage of uncertainty and insecurity to loot, acquire land, or settle old scores.
Strategic, incidental, and opportunistic violence may occur in the same election and, while theoretically distinct, can often be difficult to differentiate. Because any form of election violence is generally considered a criminal act, some perpetrators will disguise their acts to prevent all but the targeted population from discerning the purpose of the violence. Additionally, strategic violence directed at specific targets can provoke incidental violence. It is also sometimes difficult to discern between violent acts that are motivated by electoral concerns and those that are simply rendered possible by a period of insecurity. Despite these difficulties, different types of election violence with different origins require different policy responses.
Around 90 to 95 percent of cases of election violence take place prior to an election (Taylor, Pevehouse, and Straus 2017). Post-election violence happens less frequently, but it tends to be more intense and deadlier (Burchard 2015). Of the 333 national-level elections held in Africa between 1990 and 2017, only thirty-one resulted in post-election violence. But of these thirty-one cases, twenty-two (71 percent) resulted in fatalities. In thirteen cases, the fatalities numbered in the dozens, hundreds, or even thousands. 2
Plurality, majoritarian, or proportional representation systems each create a distinct set of consequential variables for electoral violence. Presidential elections that use simple plurality systems, in which the candidate with the most votes wins, are more likely to experience election violence than majoritarian systems, whereby elections proceed to a second round should no candidate secure over 50 percent of the vote in the first round (74 percent versus 51 percent). For legislative elections, proportional representation systems are less likely to experience violence than elections held using plurality or majoritarian systems. This is because elections held using plurality and majoritarian rules encourage a “winner takes all” mentality and tend to result in the exclusion of larger groups of people from state power (Fjelde and Höglund 2016). Of countries using plurality electoral systems for their legislative contests, 65 percent of elections were violent, compared to 45 percent using proportional representation. Mixed systems, or those that combine aspects of plurality and proportional representation, are less likely to experience violence; roughly 43 percent of these elections resulted in violence. The decrease in violence in systems that incorporate aspects of proportional representation reinforces the argument for more inclusive politics as a conflict management strategy in African states (Sisk and Reynolds 1998).
Incumbent governments are the most common perpetrators of election violence. Taylor, Pevehouse, and Straus (2017) estimate that incumbent governments are responsible for 80 percent of violent elections in Africa. Opposition parties in more competitive systems such as Kenya and Nigeria have, however, also participated in election violence. In both countries, candidates of all persuasions have routinely hired thugs to disrupt rallies and intimidate rival supporters. In Ghana, vigilante groups associated with both main parties have provided candidates with security, intimidated voters, and stolen or destroyed ballot boxes (Attuquayefio and Darkwa 2016).
Victims of election violence are often younger and poorer than the average citizen. Based on an analysis of Afrobarometer data from a 2015 to 2016 survey of thirty-six African countries, 3 50 percent of those under twenty-five years old feared election violence compared to 39 percent of those over the age of forty-five (Afrobarometer 2016). Fifty-five percent of those who reported that they “always go without food” also stated that they personally feared election violence, whereas only 43 percent of those who “never go without food” feared election violence. Opposition party supporters fear election violence more than ruling party supporters—52 percent of self-identified opposition supporters claim to fear election violence versus 41 percent of ruling party supporters. This makes sense as incumbent governments may disproportionately engage in electoral violence and opposition supporters are more likely to be targets.
Finally, there is a gender dimension to electoral violence. According to Afrobarometer data, women fear electoral violence slightly more than their male counterparts—48 percent of women versus 44 percent of male respondents reported fearing election violence. While women tend to be physically harmed less frequently than men during elections, women report higher levels of intimidation and psychological abuse (Bardall 2011). Women candidates and politicians are acutely affected by election violence, as some may bristle at the prospect of women’s direct participation in the political sphere, traditionally viewed as a man’s domain (Krook 2017).
Electoral violence is made up of a constellation of illicit behaviors. Although enforcement of electoral laws may be infrequent and violations of electoral infractions rarely punished, measuring illegal activities can still pose a problem. Some election violence is covert, especially in cases resulting in fatalities. Some violence is tailored for a specific audience and outside observers may be unable to detect it. Those who engage in violence may also have differing motivations. Some may be solely motivated by an electoral outcome, whereas some may opportunistically take advantage of an election dispute to loot, or seize property or land. Finally, accusations of electoral violence may be politicized and used to impugn an incumbent government or, alternatively, as a pretense for state repression.
Different quantitative measures present their own strengths and weaknesses. Because of the difficulty posed by collecting comprehensive data on election violence, some researchers opt to measure it using binary or ordinal variables. The AEVD assigns an election a score ranging from one to four. Examining election violence this way is less precise, but also less prone to error. Some sources contain event-specific data on electoral violence. The Armed Conflict Location & Event Data Project (ACLED) uses a variety of media and non-governmental reporting sources to collect data on different acts of political violence, including election-related events, from a select number of countries in Africa, the Middle East, and Asia (Raleigh et al. 2010). The Countries at Risk of Electoral Violence (CREV) dataset uses event count methodology that specifically collects information on electoral violence events (differentiated between verbal and material attacks) (Birch and Muchlinski 2017). Event count datasets provide more nuanced detail on each violent event but their comprehensiveness is a direct function of the quality of the original source data (i.e., news reports). International news sources tend to favor urban areas, report information in French, English, or Portuguese (finding vernacular languages inaccessible), and generally only capture obviously violent events.
Every measure of election violence uses some sort of temporal cutoff point to isolate election-related violence from political violence more generally. Straus and Taylor (2012) and Birch and Muchlinski (2017) use the six-month period prior to an election and the three-month period following an election as their cutoffs, which provides uniformity to the data. While election-related events certainly could occur outside of this nine-month period, the further an event occurs from an election, the harder it is to connect it to an election, especially without in-depth analysis. However, the temporal cutoff is costly; violence occurring outside of the period—such as the 2005 post-election violence in Ethiopia that broke out six months after the May election, or the year-long violence that preceded and then also followed Kenya’s 1992 founding elections—is omitted from these datasets.
Elections are cyclical processes that take place within dynamic political, social, economic, and security contexts. Uncovering root and proximate causes of different forms of electoral violence necessitates understanding the contexts in which the breadth of an electoral contest evolves. Past occurrences of electoral violence may indicate continued occurrences, but in an equal number of cases, that does not hold. For example, whereas every election in Kenya since the adoption of multiparty politics in 1992 has been accompanied by violence, not all of Tanzania’s elections have included violence. Uncovering the differences between Kenya, Tanzania, and other countries requires deep contextual understanding of where elections take place, a country’s conflict resolution mechanisms, and the strength of its institutions.
Electoral violence is frequently a tool that political actors may choose to use or not. As Hafner-Burton, Hyde, and Jablonski (2014, 150) note, the likelihood of electoral violence stems from identifying a political actor’s motivation for violence, as well as the institutional permissiveness that enables its use. In other words, many political actors may harbor grievances regarding roadblocks to their political objectives, but not all resort to violence.
When might violence occur? An important factor in answering this question requires understanding the conditions in which the electoral cycle takes place. Depending on where the electoral planning process rests and how it is progressing, political actors and their supporters have various perceptions of the credibility and integrity of the upcoming (or recently past) election and therefore different reasons for employing violence (Hafner-Burton, Hyde, and Jablonski 2014). Not all parts of the electoral cycle are the same; it is costlier—in terms of material resources and lives lost—to employ violence after an election than before it. This arguably supports the finding by Straus and Taylor (2012) that nearly all electoral violence in Africa takes place before the election. Equally important, the probability of incidental violence may increase with poor security sector training and a lack of protection of civil liberties.
Political actors respond to different threats at different points in the electoral cycle. As indicated previously, most perpetrators of electoral violence are incumbent political actors (Taylor, Pevehouse, and Straus 2017). Concerned they may lose an election, incumbents may use violence in the pre-election period to prevent an unfavorable outcome, such as reducing electoral competition or influencing voter turnout in their favor (Hafner-Burton, Hyde, and Jablonski 2014, 150). The stylized electoral cycle in Figure 18.2 identifies the many activities a country may undertake during an electoral period.
Figure 18.2 The electoral cycle
The pre-election phase can create expectations for an election and provide valuable information about the potential for political actors to attain their objectives. It can set the stage for either a peaceful transition or violent clashes. Processes such as registering voters and political parties, the availability of information about the election, training programs for electoral officials, and the allocation of adequate funding can determine if political actors will perceive the election as credible. For example, planning and budgeting delays in the Democratic Republic of Congo forced the postponement of the 2016 general elections, leading many to believe that President Joseph Kabila was deliberately hindering the election to remain in power. The delay resulted in several protests by opposition groups, many of which turned into violent clashes with security services; nearly 120 Congolese died protesting the postponement of the elections (HRW 2018).
Governments can also use violence to prevent citizens from voting. As noted before, while Kenya prepared for foundational elections in 1992, the Daniel arap Moi government engineered ethnic clashes to, among other things, displace potential opposition voters. Moi’s Kalenjin ethnic group fought (mainly) the Kikuyu over long-held beliefs that they had unfairly appropriated land and jobs from the Kalenjin. From October 1991 until the end of 1993, the ethnic clashes resulted in approximately 1,100 deaths and 300,000 displaced. No punishments ensued, although the violence was documented and blamed on the government (Klopp 2006, 66). Official institutions opted not to prosecute (Mueller 2011, 116).
While many countries, like Kenya, experience ethnic tensions, violence is not a foregone conclusion. What triggers violence? Hafner-Burton, Hyde, and Jablonski (2014, 154–6) posit that the likelihood of violence increases when governments possess little information regarding their popularity, fear they will lose, or lack institutional constraints. Kenya’s repressive government faced steep internal and external pressure to democratize (Mueller 2011; Klopp 2006), thereby indicating Moi’s fear of losing. Furthermore, Kenya’s simple plurality electoral system meant that the highest vote-getter won office, a relatively low threshold for victory. These two conditions likely reinforced Moi’s decision to foment violence by exploiting long-held local grievances to secure his win. Moi was reelected with less than 37 percent of the vote (EISA 2010). With Kenya’s 2010 constitutional reforms requiring a qualified majority electoral system for the presidential contest, winners must receive 50 percent plus one of the total vote and at least 25 percent of the vote in more than half of the country’s forty-seven counties (Constitution of Kenya, 2010, Article 138). This forces candidates to reach out to a wider segment of the electorate, hopefully lessening ethnic appeals and reducing incentives to resort to violence.
Just as governments try to influence who votes, they also may try to influence who competes. Many African countries have criteria that each political candidate must meet in order to stand for election. Incumbents use a wide range of criteria and tactics to limit the nomination of opposition candidates. In Ghana, for example, the Electoral Commission (EC) determined that a number of candidates did not meet the criteria in 2016, reducing the challengers to President John Mahama from seventeen to four (Electoral Commission of Ghana 2016b). Not surprisingly, many accused the EC of attempting to restrict competition (Clottey 2016a). After several candidates petitioned the Supreme Court, they were allowed to resubmit their nominations; in all, seven candidates competed for the presidency (Clottey 2016b). In contrast, Kizza Besigye, who has run against Uganda’s President Yoweri Museveni four times, has suffered frequent arrests, harassment, intimidation, and physical assaults. Among other crimes, the Ugandan government has charged Besigye with treason, a capital offence (Mwenda 2007). During and after the 2016 election, police repeatedly detained Besigye and state security forces routinely attacked his supporters with tear gas and engaged in other violent actions (The Guardian 2016).
Election day and the days immediately following elections remain relatively free of violence, representing only 6 percent of all electoral violence in Africa (Straus and Taylor 2012). Violence during this time may occur if voters encounter difficulties at the polls or getting to the polls, political parties hire thugs to intimidate voters, or events disrupt the vote. Biometric Voter Registration machines (BVRs), ostensibly adopted to reduce fraud by verifying the voter’s identity or to imbue transparency, can still result in delays and suspicion in elections. During Côte d’Ivoire’s 2010 election, biometric registration was introduced to update the voter rolls and clarify citizenship—a central factor in the civil war. However, the biometric process was unable to verify the status of nearly one million people and they were excluded. The government accused the electoral commissioner of attempting to rig the register; a compromise resulted in the inclusion of 400,000 names. The compromise allowed the electoral process to continue, but the Constitutional Court, an ally of incumbent President Laurent Gbagbo, still rejected nearly 500,000 votes in opposition areas and reversed the results that gave victory to his opponent, Alassane Ouattara (Piccolino 2016; US Department of State 2011).
After an election, political actors face an array of choices—each carrying different probabilities for violence. The losers may decide to protest the announced result through legal or extra-legal means; incumbents may create politically diverse cabinets; or incumbents may consider different ways to address electoral fraud that casts doubt on their victory, such as acknowledge it and launch an investigation, or confront citizens’ protests. The integrity of institutions often determines whether incumbents will use violence to retain power or whether the opposition protests violently (Hafner-Burton, Hyde, and Jablonski 2014, 157–8).
Ghana and Kenya vividly illustrate the constraining effect of judicial institutions on political actors’ decisions to use violence. In 2007, after the Electoral Commission of Kenya hurriedly declared Mwai Kibaki the winner of the presidential election, his challenger, Raila Odinga, charged that the electoral commission had stolen the election from him. Rather than seek redress through the courts, which he claimed would be biased in favor of Kibaki, Odinga urged his supporters to the streets (The Guardian 2013). Over the course of two months, approximately 1,100 Kenyans lost their lives at the hands of police officers, political party militants, and hired thugs; approximately 700,000 people were displaced (Haeneit-Sievers and Peters 2008; Lynch 2009, 2013). Ghana experienced nearly the opposite. In 2012, after the EC declared that Nana Addo Dankwa Akufo-Addo lost the presidential election to incumbent John Mahama, Akufo-Addo sought redress through the Supreme Court. Underscoring his faith in the judicial process, Akufo-Addo pledged one month before the court’s verdict that he would abide by the justices’ decision, whatever their conclusion (Modern Ghana 2013).
The actions by Odinga and Akufo-Addo reflect different levels of high court independence in Kenya and Ghana. According to the Varieties of Democracy project, Kenya’s high court was rated a 1.3 in 2006, indicating that it makes decisions in accordance with the wishes of the government. In comparison, in 2011, Ghana’s high court was rated a 3.27, indicating that it seldom makes decisions based on the preferences of the government (Coppedge et al. 2018). Kenya’s perceived lack of judicial independence appears to have influenced the decision to take the grievances to the streets. In contrast, Ghana’s judicial independence led to using the court to arbitrate the election and a peaceful transition; although the court ruled against Akufo-Addo, he had a lower incentive to use violence, because the judicial institutions provided a credible outlet.
Thus, the political and institutional context in which an election or a grievance develops is important. It can determine how political actors assess benefits or costs for violent contestation. The manner in which officials manage different aspects of the electoral cycle can also lead incumbents and challengers to assess their chances of winning. It underscores the need to examine the entirety of the electoral process, rather than just the months or weeks leading up to polling day or the days immediately following.
Systematic research on the consequences of electoral violence has lagged behind work on its character and causes. Until recently, scholars and practitioners assumed that electoral violence resulted in depressed voter turnout (Bratton 2008). However, in our previous analysis of 287 elections held in Africa from 1990 to 2014, we found no significant aggregate impact of election violence on voter turnout. This is because violence has multiple purposes: it can mobilize support, demobilize opposition, or reengineer constituencies through forced displacement. Our analysis of Kenyan elections from 1992 to 2013 reveals how the same election can achieve all three purposes (Bekoe and Burchard 2017). Our emerging research suggests that electoral violence lowers satisfaction with democracy, and may correlate with corruption and poor governance. Moreover, violence in one election may increase the likelihood of its occurrence in subsequent elections, as it creates revenge cycles among perpetrators. Furthermore, with few punitive consequences, violence can seem like a winning strategy—especially if the perpetrators are not punished and achieve their political objectives. Finally, violence may have contradictory outcomes: it appears to increase the probability that an incumbent wins, but it also raises the prospect of mass protests in the post-election period. These protests, if severe enough, can result in concessions to the opposition, contrary to an incumbent’s objectives (Hafner-Burton, Hyde, and Jablonski 2018, 460).
From an attitudinal perspective, electoral violence is associated with lower levels of satisfaction with democracy and modest decreases in support for democratic systems (Burchard 2015). These factors may spell trouble for the long-term health of a country by decreasing regime stability. Individuals who report less satisfaction with the direction of their country may, in future elections, vote against the incumbent. On the far end of the spectrum, they could be prone to rebel against the state or be convinced to support alternative, non-democratic forms of government. Election violence could also harden social divisions —such as between ethnic groups, religious groups, or linguistic groups—inducing further social conflict, when violence maps onto existing cleavages. A recent study shows that religious affinity has increased in communities along the border between Côte d’Ivoire and Burkina Faso, as a result of its politicization in Côte d’Ivoire’s civil war, which culminated in post-election violence in 2010 (McCauley and Posner 2019). Finally, the prospect of electoral violence can also be used by the state to restrict civil liberties or increase the presence of security, as Zambia did when the government declared a state of emergency following the election of Edgar Lungu (Chanda 2017).
Countries that routinely experience electoral violence may suffer economically. It is estimated that Kenya’s 2008 post-election violence negatively affected household income—even for those not directly affected by the violence—and reduced per capita gross domestic product (GDP) by roughly 5 percent (Dupas and Robinson 2012; Guibert and Perez-Quiros 2012). Fear and violence surrounding the disputed 2017 Kenyan election also resulted in significant decreases in economic growth; tourists and wary investors stayed away and economic productivity was sharply down from the previous year (Aglionby 2017).
Additional socioeconomic consequences, such as lower levels of public service delivery, resource distribution, and increased public corruption, can also result from electoral violence. The large number of displaced Kenyans affected the 2008–09 school year, as teacher shortages were recorded in areas hardest hit by the violence (Kirimi and Njubuna 2014). Violence may also result in changes in land distribution and changes to the ethnic composition of constituencies, due to displacement (Boone 2012; Lang and Sakdapolrak 2015). Corruption may rise, as governments that willingly resort to violence in order to remain in power might do so because of access to state resources and selectively distributed patronage. Political actors who face no consequences for the use of violence during elections could feel emboldened, while citizens might find it difficult to hold such politicians accountable. Future research should focus on the short-term and long-term consequences of election violence at the individual and institutional levels.
There is a growing literature on the impact of electoral violence prevention efforts. A useful categorization of prevention efforts divides interventions into three subsets: changing the attitudes of those most likely to commit violent acts; reinforcing the capacity of electoral institutions; and election monitoring (Birch and Muchlinski 2018).
The ideal way to guarantee a reduction of violence—or prevent it all together—is for politicians, political supporters, and others inclined to commit violent acts to refrain from using it as a strategy. These goals underpin the organization of peace marches, caravans, and messaging campaigns that implore politicians and their supporters not to use violence. They also affirm, as well, international actors’ use of preventive diplomacy and peace messaging to convince political opponents to abide by the rule of law, and resolve their differences peacefully. However, attempts to change hearts and minds have had mixed results. In one experiment conducted before Nigeria’s 2011 elections, peace messaging worked to prevent violence (Collier and Vicente 2014). In a cross-national analysis, Birch and Muchlinski (2018, 395) find that preventive diplomacy and mediation can work to reduce violence committed by the state. However, a comparative study by the United States Institute of Peace (USIP) concludes that interventions like peace messaging and preventive diplomacy do not work (Claes 2016, 206); preventive diplomacy, in practice, usually starts too late in the electoral process to make an impact (Claes 2016, 208).
The second category of prevention strategies comprises those focused on improving the capacity of election-related institutions such as the security sector, election management bodies, and training programs. Of these, Claes (2016) shows that security sector capacity-building is more effective than other efforts. At the same time, anecdotes caution that many citizens fall victim to violence at the hands of security services. For example, following Côte d’Ivoire’s 2010 elections, military personnel killed seven women participating in a demonstration calling for Laurent Gbagbo to accept defeat and step down (Bax and Smith 2011). Kenya’s Commission of Inquiry on Post-Election Violence (CIPEV), set up in 2008, estimated that police killed nearly 30 percent of the approximately 1,100 people that died following that country’s election (CIPEV 2008, 342). The security services’ ready access to weapons and the incumbents’ control over the forces facilitates their involvement.
Electoral management bodies (EMBs), which are ultimately responsible for the entire electoral process, have also emerged as key determinants. Charges of bias or appearance of incompetence can adversely affect the integrity of an election, fueling an electorate’s discontent or decision to protest an electoral process (see Gazibo, this volume) Thus, organizations such as the United Nations’ election assistance division and the US-based International Foundation for Electoral Systems (IFES) focus on building the capacity of EMBs to discharge their tasks impartially and professionally. African states have also focused on reforms to EMBs to improve the administration of elections, which can prevent violence. Following the contested 2012 elections in Ghana, the EC ultimately adopted twenty-seven of twenty-nine court-mandated reforms (Electoral Commission of Ghana 2016a). Among the reforms, increasing the accessibility to polling data allowed many more entities to confirm the validity of results in 2016 and reduce the spread of rumors. Other reforms included replacing the restricted Strong Room, which guarded results, with a National Collation Center, where journalists and TV cameras would be permitted (Allotey 2016; Ofori-Boateng 2016).
Among the most overt mechanisms for preventing electoral violence are domestic and international election observation missions. Starting with Zimbabwe’s 1980 election, international observation missions have been deployed across Africa to assess the integrity of national elections (Chan 2017). Accepting international observers and allowing domestic monitors signals a government’s intent to organize free and fair elections (Hyde 2012). The rationale behind election observation is that political actors will feel less inclined to engage in or support electoral malfeasance because of the high cost of international condemnation. As a means to prevent electoral violence, international and domestic observers aim to expose cheating or discourage an assessment that the polls were marred by violence—assuming that such a conclusion can hurt the legitimacy of the elected official. Monitoring organizations assume that political actors have a vested interest in behaving well in their presence.
The empirical evidence is mixed as to whether election monitoring reduces violence. Claes (2016) concludes that monitoring and observation reduce electoral violence, with the most successful cases combining international observation with local parallel vote tabulation efforts (Claes 2016). Yet, Daxecker (2014) suggests that monitoring could contribute to violence. She argues that exposing electoral malfeasance may result in post-election protests and, potentially, violence, should security forces clash with demonstrators. In fact, evidence from Eastern Europe demonstrates that election monitors’ exposure of malfeasance resulted in protests that ultimately toppled authoritarian regimes (Tucker 2007).
Luo and Rozenas (2017) further question the impact of observers. They argue that observers can reduce violence by ensuring the opposition does not protest if they report less fraud, which is most likely to occur when observers convince the incumbent that a fraudulent election will greatly increase serious violence. However, fearful of violence, some observers may choose not to comment on fraud in a tense environment. As a result, election observation that is meant to prevent violence, as well as fraud, will likely do neither. For example, many have argued that monitors and the donor community in Kenya censored their reporting on the elections in the 1990s and 2013 for fear of provoking a violent response to fraud (Cheeseman, Lynch, and Willis 2016; Brown and Raddatz 2014). The Kenyan Supreme Court’s annulment of the internationally endorsed 2017 elections raised similar concerns of censorship (Kahura 2017).
Although less studied, domestic observation missions have produced important outcomes that warrant further examination. Notably, domestic observation missions have introduced innovations that strive not just to assess results, but also to resolve tension and prevent violence. For example, prior to Senegal’s 2012 election, civil society established an election situation room. It served not only to monitor the election, but also to deter fraud. Meeting regularly for more than one year, a network of civil society organizations established clear punitive contingencies in the event of fraud, which they shared with government stakeholders (Doudou Dia, personal interview with the author, March 11, 2012). In another example, in 2008, Ghana’s Coalition of Domestic Election Observers (CODEO) trained a subset of election violence monitors to provide periodic reports of tension and violence to selected stakeholders, in order to determine which actions to take to resolve tension or prevent violence (Oduro 2012).
Timing and institutional strengthening are critical for all the programs or interventions adopted. Given that nearly all violence takes place before an election, interventions to stave off violence must begin well in advance of the polls. Increasingly, research shows that the likelihood of violence decreases when interventions improve the capacity and integrity of electoral management bodies, security sector services, and monitoring organizations. Yet, while focusing on the administrative aspects of elections lends itself to measurable results, they cannot guard against all violence. Nigeria’s well-managed 2011 election, where approximately 800 died in post-election violence, provides a cautionary tale: even credible electoral bodies do not guarantee a violence-free election (Bekoe 2011).
The last decade has brought rapid expansion in practitioners’ and academics’ understanding of electoral violence as a subset of political violence. Three of the most important findings are: (1) the dominance of pre-election violence—underscoring the need to intervene early; (2) the propensity of incumbent governments over challengers to use violence as a tool—providing opportunities for diplomatic and other state-to-state engagements to prevent violence; and (3) the persistence of electoral violence, despite increased acceptance of democratic norms—highlighting the need to consider underlying political and conflict dynamics. Thus, while the field has advanced its understanding of the components of electoral violence, prevention remains a challenge.
At the same time, there remain many opportunities for further research. There is still much we do not know about how to accurately measure electoral violence, the probability of its occurrence, and the consequences of different kinds of electoral violence. Although many international entities rationalize their focus on electoral violence by noting its impact on voter turnout (USAID 2010), further empirical analysis does necessarily corroborate this assumption. Emerging research suggests that violence may affect voter choice, however, this conclusion requires additional work. In line with more hidden consequences, the field has not adequately probed the gender dimensions of electoral violence, especially as they pertain to the home environment. Socioeconomic consequences of election violence are also less well understood. Equally important, while electoral violence is increasingly acknowledged to serve as a tool to obtain political ends, the field has not clearly determined the factors that place particular countries or communities at risk. Moreover, scholars and practitioners still debate motivations for choosing to use violence and how to reverse those incentives. Future researchers focusing on these issues may be able to finally help reduce the level of electoral violence in Africa.
To update the AEVD, we found that we could no longer rely solely on the US State Department’s Human Rights Reports, as revisions in the level of detail rendered them less precise. Instead, we used the same set of criteria as Straus and Taylor (2012) but analyzed newspaper articles and reports by international organizations, in addition to the State Department’s reports.
The thirteen cases of extreme post-election violence are: Angola 1992, Burundi 2015, Côte d’Ivoire 2010, Kenya 2007, Kenya 2017, Lesotho 1998, Madagascar 2001, Nigeria 1993, Nigeria 2007, Nigeria 2011, Togo 2005, Uganda 2016, and Zimbabwe 2008.
Countries included in Round 6 of the Afrobarometer survey that asked the question “During election campaigns in this country, how much do you personally fear becoming a victim of political intimidation or violence?” were: Algeria, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape Verde, Côte d’Ivoire, Egypt, Gabon, Ghana, Guinea, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritius, Morocco, Mozambique, Namibia, Niger, Nigeria, São Tomé and Príncipe, Senegal, Sierra Leone, South Africa, Sudan, Swaziland, Tanzania, Togo, Tunisia, Uganda, Zambia, and Zimbabwe.