Mixing Votes with Violence:Election Violence around the World
Elections are now held in almost every country in the world. Yet the introduction of electoral processes in developing countries has led to a mix of voting and violence rather than the establishment of peace and stability, as violence in recent elections in Afghanistan, Bangladesh, Kenya, and Zimbabwe illustrates. Developing countries display various forms of violence closely linked to elections, such as incumbents intimidating opposition candidates and voters, armed groups using violence to disrupt electoral processes, or rioting between the supporters of opposing political parties. Until recently, a lack of globally available data complicated efforts to properly describe and understand election violence. This paper uses disaggregated event data from the Electoral Contention and Violence (ECAV) dataset to provide a global, post-Cold War assessment of election violence in the developing world. We first present a descriptive assessment, beginning with temporal and regional patterns, which show that most election violence occurs in Asia, the Middle East, and Africa. We then examine timing, election type, electoral systems, and the actors and targets of election violence. Second, we probe the link between armed conflict and election violence. Violent elections in countries such as Afghanistan or Iraq highlight the challenges of holding elections during civil war, yet existing research has not been able to assess to what extent election violence takes place inside or outside of ongoing armed conflict. We conclude our systematic assessment of election violence with implications for scholarship and policy.
Elections are now held in almost every country in the world, yet the introduction of electoral processes in developing countries has led to a mix of voting and violence rather than the consistent establishment of peace and stability. Moreover, highly violent elections in Afghanistan and Iraq indicate that election-related violence is often intertwined with other forms of political violence. Our paper first reviews existing empirical and theoretical work on electoral violence as a distinctive type of political violence, including literature [End Page 53] on the link between armed conflict and elections. Second, we present regional and temporal patterns in election violence, examine the timing of election violence, the incidence of violence across election types and electoral systems, and catalogue actors and targets involved in violence. Departing from findings in the existing literature, we show that the state is not the primary perpetrator of violence, that Asia and the Middle East experience more election violence than Africa, and that parliamentary elections are more violent than other elections in a global sample. Third, we explore the link between armed conflict and election violence, showing that the majority of election violence occurs during armed conflict. Existing research has sometimes assumed that election violence is primarily the result of incumbent or opposition parties trying to influence election outcomes. However, election violence during armed conflict involves armed groups as the primary perpetrators, and these groups' goals do not necessarily align with parties competing in elections. We conclude with implications of this empirical assessment for democracy promotion and conflict resolution.
What We Know About Electoral Violence, and What is Missing
Elections are often regarded as a peaceful means for choosing political representation, but in reality, they are frequently accompanied by violence. Election violence is a subtype of political violence that either aims to influence electoral processes in the run-up to election day or takes place as a violent response to elections because of concerns over electoral conduct.1
Academic and policy work has prioritized some of the most violent and lethal recent elections, such as the 2007 Kenyan general elections, the 2008 Zimbabwean general elections, or the 2010 Cote d'Ivoire presidential elections.2 Focused on African elections, existing work suggests that incumbents are the primary perpetrators of electoral violence, a finding that has become an implicit assumption in other research.3 Regarding the timing of violence, research has shown that pre-election violence is more common.4 The literature also suggests that presidential elections are more violent than parliamentary elections.5 Empirical research on the causes of election violence has shown that more competitive elections, those held under majoritarian electoral rule, and elections assessed as fraudulent by observers are more likely to experience election violence.6 Yet with evidence coming primarily from African elections, it is impossible to know whether these findings hold for other world regions.
An additional blind spot in some existing work is that it treats election violence as a campaign strategy employed by incumbents or the opposition, ignoring the presence of armed groups as potential perpetrators of violence in countries struggling with armed conflict. It seems intuitive to expect that elections held during or immediately after armed conflict experience more [End Page 54] violence. During armed conflict, rebel groups excluded from participating in elections may use violence to express their opposition to the political system, while spoiler dynamics could contribute to violence in post-conflict elections. Academic work has therefore also recognized important links between armed conflict and elections.7 For example, research highlights that many armed groups engage in both electoral and violent strategies, suggesting that they are complements rather than substitutes.8 It is challenging to empirically distinguish between violence related to elections and violence related to the armed conflict more generally, but such a distinction is crucial if the study of election violence is indeed an analytically distinct area of research.9 Lacking systematic and disaggregated data, however, we do not yet know how much election violence happens during, after, or outside of armed conflict.
To conclude, existing research on election violence relies primarily on data for African elections and highlights incumbents as the primary perpetrators. This work also suggests that presidential elections and those using majoritarian electoral rule experience more violence, but these patterns have not been established globally. Furthermore, while highlighting the importance of armed conflict in creating incentives for election violence, existing research has not provided an empirical assessment of the link between armed conflict and the incidence of election violence. We proceed to describe where election violence is most common, its timing, and who engages in it by leveraging the Electoral Contention and Violence (ECAV), a new global event dataset on election violence.10 We then use ECAV to examine the link between armed conflict and election violence. We assess how much of election violence happens inside and outside of armed conflict, and whether election violence in conflict countries is the result of the involvement of conflict actors.
Election Violence Across the World
Regional and Temporal Patterns
ECAV contains approximately 10,000 events of election violence coded for 2,000 election rounds held in 135 countries from 1990 until 2012.11 While this paper presents election-level trends, the ECAV data are coded at the level of the event-day-location, allowing for temporally and spatially disaggregated analyses. Figure 1 shows the regional incidence of election violence, plotting all events in which location information was available at the first-order administrative unit level, and omitting events where only the country was known. Election violence is most common in Asia (35% of events), followed by Africa [End Page 55] (28%), the Middle East (20%), the Americas (9%), and Eastern and Southeastern Europe (7%). The figure thus highlights the importance of studying election violence in regions other than Africa. In terms of countries, most violent incidents occur in elections held in Iraq (11.2%), India (6.5%), Bangladesh (6%), Afghanistan (5.9%), and Kenya (5.2%).
Figure 2 shows the regional distribution of violent events over time. We divide this period into six-year sub-periods, beginning in 1989 (since elections held in 1990 also include some violence in 1989) and ending in 2013 (since elections held in 2012 could experience post-election violence in 2013). We observe an increase of election violence in the first years of the post-Cold War period. Election violence in Europe peaks in the years between 1995 and 2000 because of violent elections in Albania, Bosnia and Herzegovina, and Russia. Likewise, violent elections in Bangladesh and India led to an increase of violence in Asia in 1995-2000. From 2000 onwards, violence increases again because of violent elections in Haiti and Peru in the Americas, but also in Egypt, Iran, and Zimbabwe. Violence decreases thereafter in most regions except Asia (especially elections in Bangladesh, Sri Lanka, and the Philippines in 2001), and Africa (violence during Kenyan and Zimbabwean elections in 2002). From 2001-2007, violence peaks overall because of highly contentious and violent elections in Afghanistan, Egypt, Iraq, and Ukraine (particularly in the years 2004 and 2005). After 2008, we observe an increase [End Page 56] in Africa, again because of violent elections in Kenya and Zimbabwe. Violent elections in Afghanistan, Egypt, India, and Iraq lead to substantial violence in Africa and the Middle East, but violence decreases toward the end of the period, especially in 2011 and 2012.
Figure 3 shows that most election violence is related to parliamentary elections (53%), followed by presidential elections (22%), concurrent elections (15%), and other types (10%).12 This finding stands in contrast to research on African elections, which has shown that presidential elections attract more violence.13 When examining ECAV patterns for election type and violence by region (figure not shown), we confirm that presidential elections attract more violence in Africa, but the opposite pattern holds in Asia, the Middle East, and Latin America. The predominance of violence during parliamentary elections stems partly from the fact that parliamentary election rounds make up the majority (55%) of the elections included in ECAV. Moreover, several countries with highly violent elections have parliamentary (Bangladesh, India, Iraq) or mixed electoral systems (Afghanistan, Egypt, Zimbabwe). [End Page 57]
Figure 4 examines the incidence of violence across electoral systems. We find that election violence is more common in elections with majoritarian systems, confirming earlier research.14 The figure shows that election violence happens in 42 percent of majoritarian elections compared to 36 percent of elections with proportional representation.
[End Page 58]
Figure 5 presents information on the timing of election violence. Confirming previous research on African elections, Figure 5 shows that most election violence occurs before elections (60%), followed by post-election violence (29%), and violence on the election day (11%). To be fair, establishing a greater incidence of pre-election violence is, in some ways, built into the ECAV dataset because it codes violent events six months before elections, but only three months thereafter. While we think that most post- election violence takes place within this time frame, occasionally a longer time frame would be preferable. For example, post-election violence in Cote d'Ivoire did not escalate fully until more than four months after the 2010 elections.
We now present information on perpetrators and targets of election violence. Table 1 distinguishes between events in which the initiator of violence could be established, and events where it was not possible. The initiator of a violent event could be established in 80 percent of ECAV violent events, leaving 20 percent in which the initiator could not be determined. Interestingly, we see that unknown actors and armed groups are the most common perpetrators of election violence, which contradicts other evidence implicating the state as the primary perpetrator.15 In fact, for events with an established initiator, the state is [End Page 59] responsible only for 16 percent of violent events. In terms of targets, we observe that state actors are the most common targets, followed by citizens and party actors. In events where we could not establish the perpetrator or target of an event, citizens are the most common participants, followed by state and party actors. Overall, our evidence contradicts the common assumption of state actors as primary perpetrators, although this finding could be a result of the high incidence of election violence during armed conflict, which we turn to next.
Armed Conflict and Election Violence
Out of the 135 countries included in ECAV, 67 countries experienced armed conflict between 1985 and 2012, and 54 of these countries held elections while conflict was ongoing (leaving only 13 conflict countries without elections). As Figure 6 displays, 59 percent of all election violence took place during ongoing armed conflict, suggesting that election violence is often intrinsically linked to other political conflict. We note that ECAV codes only violence related to elections rather than all violence, meaning that violence occurring during armed conflict has to be temporally and substantively linked to elections.16 The high incidence of election violence during armed conflict is thus not simply an artifact of more violence taking place in conflict countries.
For the 41 percent of events of election violence occurring outside of conflict, we distinguish between those taking place in countries that never experienced armed conflict and events that happen more than five years after conflict ends (what we call peacetime elections) and those that take place within five years of an armed conflict ending (what we call post-conflict elections).17 Of the 41 percent of events outside of ongoing armed conflict, 79 percent take place in peacetime, while the remaining 21 percent occurs in the five years after a conflict ended. [End Page 60]
Regionally, almost half of election violence during armed conflict occurs in Asia (49%), followed by Middle East (25%), and Africa (17%). A look at particular elections shows that a small number of conflict countries are responsible for a majority of violence. This pattern is not unusual in the study of political violence, where individual conflicts are often behind temporal trends.18 The 2005 Iraqi constitutional elections alone are responsible for almost 13% of events, followed by the 2005 parliamentary elections (6%), the 1996 Bangladeshi parliamentary elections (5%), and the 2009 Afghan presidential elections and the 2001 Bangladeshi parliamentary elections (4% each).
To understand how armed conflict is linked with election violence, we can use our disaggregated data to identify the actors engaged in violence. In Table 2, we show the distribution of perpetrators and targets during armed conflict. In this table, we only present events where initiators could be established. Armed groups (39%), unknown actors (29%), and states (12%) are the most common perpetrators of election violence during armed conflict. While we can only speculate about unknown perpetrators, such events are often bombings or killings, and armed groups may very well be behind them. For targets, most commonly targeted are the state (38%), citizens (26%), and parties (16%). [End Page 61]
Our paper demonstrates the potential of disaggregated data for addressing important yet unresolved questions in the study of election violence. Our analysis shows that findings from existing research, which has relied primarily on empirical evidence from African elections, are only partly confirmed in a global sample. With regard to geography, election violence is most common in Asia, followed by the Middle East and Africa, which suggests that scholars and practitioners need to look beyond African elections. Regarding time, violence peaks in 2004 and 2005, driven by highly violent elections in Afghanistan, Egypt, Iraq, and Ukraine. We find that violence is more common before elections than on election-day or after voting, which should encourage more research on the conditions leading to violent pre-election interventions. We also confirm that majoritarian elections experience more violence than other electoral systems. For election type, our findings depart from existing research, showing that parliamentary elections experience the most violence. This finding seemingly stems from regional differences; in African elections, presidential contests are indeed more violent, but this trend does not hold in other regions. Examining the actors and targets of violence, we cannot confirm that the state is the primary perpetrator of violence, as is suggested in previous research. Instead, we find that armed groups are the most frequent initiators of violence. Further examination shows that the majority of election violence happens during armed conflict, indicating that election violence is often intertwined with other political violence. This finding has important implications for future scholarship and policy. First, the dynamics of violent interactions between armed groups, the state, and civilians during elections are likely quite different from government-opposition contention during peacetime elections. Violence by armed groups may be aimed to oppose the holding of elections and disrupt the electoral process in its entirety rather than influencing the outcome in favor of the government or opposition.19 These dynamics require more academic [End Page 62] attention. Future analyses should, for instance, examine when and why armed groups target the electoral process as opposed to using other political violence. While our assessment presented evidence at the level of the election, the ECAV are fully geocoded and available at the level of the violent event, making subnational analyses of election violence possible. Second, the finding that most election violence happens during armed conflict suggests that policymakers in democracy promotion should pay more attention to electoral contests in the midst of armed conflict, rather than focus on post-conflict or peacetime elections. Almost all countries experiencing armed conflict in our data are holding elections while conflict is ongoing, and many of these elections generate substantial violence.
Ursula Daxecker is an associate professor in the Department of Political Science at the University of Amsterdam. She is a member of the research group Political Economy and Transnational Governance at the University of Amsterdam. Professor Daxecker emphasizes political violence in her research.
Alexander Jung is a researcher at the University of Amsterdam and also contributes to the Political Economy and Transnational Governance research group.
1. Kristine Höglund, "Electoral Violence in Conflict-Ridden Societies: Concepts, Causes, and Consequences." Terrorism and Political Violence 21 no. 3 (2009): 412-427; Scott Straus and Charlie Taylor. "Democratization and Electoral Violence in Sub-Saharan Africa, 1990-2007." in Voting in Fear: Electoral Violence in Sub-Saharan Africa, ed. Dorina A. Bekoe, (Washington, DC: United States Institute of Peace Press, 2012): 15-38; Paul Staniland, "Violence and Democracy," Comparative Politics 47 no. 1 (2014): 99-118.
2. See, for example, Catherine Boone and Norma Kriger, "Land patronage and elections: winners and losers in Zimbabwe and Côte d'Ivoire," in Voting in Fear: Electoral Violence in Sub-Saharan Africa, ed. Dorina A. Bekoe, (Washington, DC: United States Institute of Peace Press, 2012), 75-117 and Kathleen Klaus and Matthew Mitchell, "Land grievances and the mobilization of electoral violence Evidence from Côte d'Ivoire and Kenya," Journal of Peace Research 52 no. 5 (2015): 622-635.
3. Incumbents are mentioned as main perpetrators in Straus and Taylor (see note 1 above) and Emilie M. Hafner-Burton, Susan D. Hyde, and Ryan S. Jablonski, "When Do Governments Resort to Election Violence?" British Journal of Political Science 44 no. 1 (2014): 149-179.
4. Straus and Taylor (see note 1 above).
5. Stephanie M. Burchard, Electoral Violence in Sub-Saharan Africa: Causes and Consequences, (Colorado: Lynne Rienner Publisher, 2015).
6. Burchard (see note 4 above); Ursula Daxecker, "The Cost of Exposing Cheating: International Election Monitoring, Fraud, and Post-Election Violence in Africa," Journal of Peace Research 49 no. 4 (2012): 503-16; Hanne Fjelde and Kristine Höglund, "Electoral Institutions and Electoral Violence in Sub-Saharan Africa." British Journal of Political Science 46 no. 2 (2016): 297-320; Roxana Gutiérrez-Romero, "An Inquiry into the Use of Illegal Electoral Practices and Effects of Political Violence and Vote-Buying." Journal of Conflict Resolution 58, no. 8, (2014): 1500-1527; Hafner-Burton et al. (see note 2 above); Idean Salehyan and Christopher Linebarger, "Elections and Social Conflict in Africa, 1990-2009," Studies in Comparative International Development 50 no. 1 (2015): 23-49; Steven I. Wilkinson, Votes and Violence: Electoral Competition and Ethnic Riots in India (Cambridge University Press, 2004).
7. Sarah Zukerman Daly, Organized Violence After Civil War, (Cambridge University Press, 2016); Thomas Edward Flores and Irfan Nooruddin, "The Effect of Elections on Post- Conflict Peace and Reconstruction." Journal of Politics 73 (2012): 558-570; Carrie Manning, "Armed opposition groups into political parties: Comparing Bosnia, Kosovo, and Mozambique," Studies in Comparative International Development 39 no. 1 (2004): 54-76; Aila M. Matanock, "Using Violence, Seeking Votes: Introducing the Militant Group Electoral Participation (MGEP) Dataset," Journal of Peace Research 53, no. 6 (2016): 845-853; Aila M. Matanock, "Bullets for Ballots: Electoral Participation Provisions and Enduring Peace after Civil Conflict." International Security 41 no. 4 (2017): 93-132; Barbara F. Walter, "Designing Transitions from Civil War: Demobilization, Democratization, and Commitments to Peace," International Security 24 no. 1 (1999): 127-55.
8. Matanock (2016, see note 8 above).
9. Hanne Fjelde and Kristine Höglund, "Electoral Violence: The Emergence of a Research Field." APSA Comparative Democratization Newsletter 14 no. 2 (2016): 8-11.
10. Ursula Daxecker and Elio Amicarelli, "Electoral Contention and Violence (ECAV): A New Dataset," 2018, http://ursuladaxecker.weebly.com/research-projects.html.
11. Countries that were members of the OECD in 1990, the first of year of coding are excluded from the analysis. This means that Western Europe, the US, Canada, Australia, and Japan are not included. For detailed information, please see the data paper and codebook available at http://ursuladaxecker.weebly.com/research-projects.html.
12. Concurrent elections encompass simultaneous presidential and parliamentary elections. Other types include constitutional referenda.
13. Burchard (see note 4 above).
14. Fjelde and Höglund (see note 6 above).
15. See Straus and Taylor (see note 1 above) and Hafner-Burton et al. (see note 3 above).
16. See Daxecker and Amicarelli (see note 11 above) for a detailed discussion.
17. A threshold of five years is of course arbitrary, but it also seems problematic to consider all elections after armed conflict as post-conflict. Since we code elections from 1990 onwards, we consider whether countries experienced armed conflict between 1985-1990 to calculate the descriptive statistics.
18. Marie Allansson, Erik Melander, and Lotta Themnér, "Organized violence, 1989-2016," Journal of Peace Research 54 no. 4 (2016): 574-87.
19. Paul Staniland, "Armed Groups and Militarized Elections," International Studies Quarterly 59 no. 4 (2015): 694-705.