Pre-election violence and territorial control: Political dominance and subnational election violence in polarized African electoral systems

Abstract

Cross-national research on African electoral politics has argued that competition increases the prospects for pre-election violence. However, there is a dearth of systematic research on the effect of political competition on pre-election violence at the subnational level. We theorize that in African democracies characterized by competition at the national level but low subnational competitiveness (polarization), violence is often a manifestation of turf war and a tool to maintain and disrupt political territorial control. Consequently, contrary to expectations derived from the cross-national literature, pre-election violence is more likely in uncompetitive than competitive constituencies. Locally dominant as well as locally weak parties have incentives to perpetrate violence in uncompetitive constituencies. For locally dominant parties, violence is a tool to shrink the democratic space in their strongholds and maintain territorial control. For locally weak parties, violence can disturb the dominance of the opponent and protect their presence in hostile territory. We hypothesize that pre-election violence will be particularly common in opposition strongholds. In such locations, ruling parties can leverage their superior repressive resources to defend their ability to campaign, while the opposition can use their local capacity to reinforce the politics of territoriality. We test our hypotheses with original constituency-level election violence data from the 2016 Zambian elections. Data come from expert surveys of domestic election observers and represent a novel way of measuring low-level variations in election violence. Our analysis shows patterns of pre-election violence consistent with our theory on pre-election violence as a territorial tool.

Keywords

Africa conflict election manipulation election violence geography Zambia

A recent estimate of election violence across sub-Saharan Africa shows that a majority of all election campaigns featured some degree of violence (Straus & Taylor, 2012). Empirical research has suggested that competitive elections, where the ultimate outcome is uncertain, are more prone to violence than less competitive contests. When competition is high, political elites are more willing to engage in illicit tactics, including violence, to maintain or gain power over the state machinery and vital resources (Fjelde & Höglund, 2016; Hafner-Burton, Hyde & Jablonski, 2014).

The relationship between competition and election violence has been confirmed in several African countries such as Kenya (Kanyinga, 2009; Boone, 2011), Côte d’Ivoire (Straus, 2011), Nigeria (Bratton, 2008), and Zimbabwe (Bratton & Masunungure, 2007; LeBas, 2006), where political actors have resorted to violence to manage electoral uncertainty. However, apart from high levels of competition, the aforementioned countries have one additional similarity: they all have highly regionalized voting patterns and consequently low levels of subnational competition. Kenya, Côte d’Ivoire, Nigeria, and Zimbabwe all belong to a group of countries we will here refer to as polarized electoral systems. We define these systems as those that combine high levels of national competition with low levels of subnational competition. Such systems make up a great share of Africa’s more competitive democracies, where electoral patterns remain highly regionalized and local bloc voting is prevalent despite increased levels of national competition (Ichino & Nathan, 2013; Wahman, 2017). Polarized electoral systems introduce an important puzzle: if competition breeds violence, how can violence persist in countries where few subnational locations are competitive?

The conventional view of competition and pre-election violence does not offer much of an explanation of persistently high levels of election violence in polarized electoral systems. Like other forms of manipulation, pre-election violence is associated with costs (Robinson & Torvik, 2009). All else being equal, political actors should be more willing to bear those costs when competition is high locally and nationally, particularly if local elections run concurrently with national elections and parties are fighting to win marginal parliamentary seats.¹ This leaves us with three possible explanations for why polarized electoral systems may experience high levels of election violence despite low levels of local competition: (i) subnational levels of competition are inconsequential for aggregate levels of violence; (ii) most violence in polarized electoral systems occurs within the rare islands of competitiveness; or (iii) competition plays a markedly different role in determining election violence at the national and subnational levels. We will make an argument for the last proposition.

In this article, we shift the focus from the national to the subnational level to uncover a more complex relationship between competition and election violence in polarized electoral systems than previously acknowledged. Although most quantitative work on election violence has studied cross-national variation (Daxecker, 2014; Fjelde & Höglund, 2016; Fjelde, 2020; Hafner Burton, Hyde & Jablonski, 2014; Straus & Taylor, 2012; Taylor, Pevehouse & Straus, 2017; von Borzyskowski, 2019), several recent studies have concentrated on subnational dynamics of election-related conflict (e.g. Daxecker, 2020; Dercon & Guitérrez-Romero, 2012; Goldring & Wahman, 2018; Malik, 2018; Reeder & Seeberg, 2018). However, these studies do not theorize that competition plays a markedly different role at the subnational level than it does at the cross-national level. Focusing particularly on pre-election violence,² the most common form of election violence in sub-Saharan Africa (Daxecker, 2014; Taylor, Pevehouse & Straus, 2017), we argue that subnational dominance – where one party is significantly stronger than its closest rival – rather than competition, is conducive to higher levels of pre-election violence. Thus, our argument for the subnational level contrasts with the dominant argument presented in the cross-national literature.

We argue that political geography is crucial to understanding elections in polarized electoral systems. Pre-election violence in polarized electoral systems can be understood as a tool for what Robert Sack (1986: 1) labels ‘territoriality’, that is, ‘a spatial strategy to affect, influence, or control resources and people, by controlling area’. We argue that pre-election violence fills a dual role. First, it can be used to shrink the democratic space, by preventing effective campaigning by rival parties inside a party’s stronghold. Second, it can be used to break the politics of territoriality and defend a party’s presence inside an area dominated by a rival party. It is this combination of high-level national competitiveness and low-level subnational competition that is particularly conducive to violence. On the one hand, parties have incentives to use violence to tip the balance in nationally competitive elections. On the other hand, subnational dominance makes violence a useful tool to preserve and contest the politics of territoriality. If competition is high at the national level, parties cannot afford defection in their strongholds. Without securing super majorities in their strongholds they are likely to lose close national elections. We argue that pre-election violence is more likely in all uncompetitive constituencies, but the highest level of violence will be observed in opposition strongholds. In these constituencies, ruling parties can leverage their superior resources to perpetrate violence outside their own home turf, whereas opposition parties can use their considerable local capacity in an attempt to reinforce their territoriality.

We leverage a subnational analysis of the 2016 Zambian elections to test our theory on pre-election violence and subnational dominance. The analysis uses a dataset based on 464 surveys with domestic election observers and offers an innovative way to measure subnational variations in election violence. Our method for measuring election violence across space assures systematic coverage and local disaggregation to a level not possible with frequently used event-datasets that tend to introduce systematic measurement errors (Eck, 2012; Weidmann, 2015; von Borzyskowski & Wahman, forthcoming).

Our analyses show results consistent with our argument of pre-election violence as a territorial tool. The quantitative tests reveal that violence in the 2016 Zambian elections was particularly common in uncompetitive constituencies and in constituencies controlled by the opposition. Our findings have important implications for research on election violence in Africa and beyond. They highlight the need to understand not only how election violence affects micro-level political behavior, but also the way in which it fills a role in shaping local electoral environments. They also contribute to the burgeoning literature on subnational authoritarianism and highlight the importance of political geography in the study of election violence.

Pre-election violence and political campaigns in polarized electoral systems

To understand how pre-election violence persists in polarized electoral systems, we first need to understand the role of pre-election violence in electoral competition. Most theorizing on the matter is skeptical that pre-election violence can be used to effectively persuade swing voters or attract the support of new voters (e.g. Collier & Vicente, 2012; Robinson & Torvik, 2009). Some have even questioned its effectiveness as an electoral strategy and argued that voters respond negatively to parties engaging in violence (Gutiérrez-Romero & LeBas, 2020; Rosenzweig, 2017).

If violence is ineffective for attracting swing voters and it can backfire on the perpetrator, why has pre-election violence persisted in African elections? The most plausible answer is provided by scholars who stress the potentially demobilizing effect that violence may have among voters. Through peddling fear, parties can utilize pre-election violence to reduce turnout for rival parties (Collier & Vicente, 2012; Gutiérrez-Romero & LeBas, 2020; Rauchenbach & Paula, 2019).³ Although this theory is intuitively sound, Bekoe & Burchard (2017) found no aggregate negative relationship between levels of pre-election violence and electoral turnout in African countries.

While concentrating mainly on micro-level behavioral consequences of violence (e.g. vote choice or decisions to abstain or participate in voting), earlier work on pre-election violence paid insufficient attention to political geography and the significance of violence in shaping the local electoral environment; however, the effect of violence may be more noticeable on the meso than the micro level. Existing studies have recognized that pre-election violence may shape political participation but have understood participation narrowly by concentrating only on voting. We do not contest the idea that violence may be used to deter turnout, but participation involves a much wider array of activities, such as the ability to campaign, attend political rallies, articulate opinions, or canvas. When violence affects parties’ abilities to engage in such activities, it may affect vote choice. This occurs not directly, through changes in voter preferences, but indirectly by shrinking the democratic space and disabling effective competition.⁴ For example, after experiencing systematic violence from government party cadres in the 2016 Zambian elections, supporters of the opposition were asked by their party not to display their party colors publicly.⁵ These supporters were unlikely to change their vote choice, but the visual absence of opposition party colors most likely affected the impression among voters of the viability of the opposition.

Due to low penetration of national media and low levels of nationalization, African parties have invested heavily in political rallies as a central campaign tool, particularly in polarized electoral systems such as Ghana, Kenya, and Zambia (Beardsworth, 2018a; Brierley & Kramon, 2017; Horowitz, 2010; Paget, 2019; Rauchenbach, 2017). The aim of such campaigning techniques is to show local viability. Recent research has shown that African voters tend to favor locally viable candidates (Conroy-Krutz, 2013; Ichino & Nathan, 2013) and overestimate the national strength of locally popular parties (Horowitz & Long, 2016). Parties active on the ground can shape public opinion (Letsa, 2019).

Parties cannot rely on national communication to win support in an area. A party that is unable to show local presence cannot expect to win a significant share of the vote in that locality. Parties that have the ability to mobilize impressive crowds can demonstrate their local presence to voters. Drawing on earlier work on electioneering in the West (e.g. Chen & Reeves, 2011; Rohrschneider, 2002), studies on the locations of African rallies have argued that rallies tend to be targeted towards swing districts (Horowitz, 2010) or strongholds in an attempt to mobilize the base (Rauchenbach, 2017). But in a study on Zambia, Beardsworth (2018a) shows that parties also target areas where they traditionally had little support. This strategy is aimed at explicitly showing increased viability in the targeted locality.

Political dominance and violence as a territorial tool

In polarized electoral systems where parties strive to maintain and break territorial control and perceptions of electoral dominance, violence fills an important role. In short, parties use violence to perpetuate local dominance in their strongholds and attempt to forge entry into the electoral arena in opposing parties’ strongholds. Violence, according to this understanding, is primarily a tool to shape the local electoral environment.

In the African context, violence has often been used in uncompetitive constituencies by locally dominant parties as a form of ‘anti-campaigning’. Reports from polarized electoral systems like Malawi and Zambia have described the creation of ‘no-go zones’, where local power brokers prevent rival parties campaigning and make voters fearful of attending opposition parties’ rallies or wearing opposition party branded attire. This sends a signal that rival parties are not welcome and that voters should not display affinity towards parties other than the locally dominant one (Authors’ interview, Choma, 22 December 2016).⁶ For example, in the Malawi 2014 election, cadres associated with the Democratic Progressive Party (DPP) perpetrated violence against supporters of the People’s Party (PP) when PP’s presidential candidate, Joyce Banda, held a rally in a DPP stronghold constituency in Thyolo (Dionne & Dulani, 2014). The violence resulted in two fatalities. Nevertheless, it helped reinforce the territoriality of the local hegemon; the DPP won 91% of the vote in Thyolo.

One might argue that this kind of violence, sparked by locally dominant parties, may be counterproductive for the perpetrator, as it could deter turnout in a party’s stronghold. However, such risks are limited if pre-election violence is targeted against minorities, outsiders, and hardcore rival party supporters (Collier & Vicente, 2012). Although the risks are small, the potential electoral benefits are significant: if a locally dominant party is able to deter competitors from actively campaigning or diminish crowds at their rallies, it will be able to perpetuate the perception of local dominance and diminish the losses from elite and voter defections. In polarized electoral systems, parties cannot afford to lose control over their strongholds and allow rival parties to chip away their overwhelming vote margins. Limited defections in strongholds are likely to result in an inability to win national majorities if rival parties are able to maintain their dominance on their own home turf. Violence is unlikely to kill the campaign of a rival party in a competitive area, where the party enjoys the support of a critical mass of supporters, but it can be very effective in places where the rival party is mobilizing mostly local minorities or small pockets of support. Voters that are already prime targets of repression and social exclusion are particularly likely to refrain from participating in the electoral campaign when the campaign turns violent.

Parties not only use violence in their strongholds to protect inroads from challengers, but also use pre-election violence in rival party strongholds to protect their ability to contest and participate in the local electoral environment. If parties can compete and show presence in their rivals’ strongholds, particularly the more populous ones, this undoubtedly imposes high costs on their rivals. When faced with a hostile environment, parties with the ability to show repressive capacity are more likely to contest the dominance of rival parties. In this way, parties can forge entry into enemy turf and protect their place in the environment, thereby widening the space that the opposing party attempts to shut. In Zambia in 2016, the ruling party brought foot soldiers for protection when conducting rallies in the opponent’s strongholds. Such foot soldiers were often responsible for using excessive violence to deter any attempts at ‘de-campaigning’ (Authors’ interview, Choma, 22 December 2016). We therefore formulate Hypothesis 1:

H1: Pre-election violence is more likely in uncompetitive constituencies than in competitive constituencies.

The importance of capacity to engage in election violence

As emphasized previously in studies on election violence and other forms of electoral manipulation, pre-election violence is not simply a function of demand. Violence and other forms of manipulation require capacity and agents willing to engage in illicit acts on behalf of a reliable patron (e.g. Rundlett & Svolik, 2016; Taylor, Pevehouse & Straus, 2017). Such capacity varies across incumbent and opposition parties.

Local dominance enhances a party’s capacity for violence. Locally dominant parties can offer agents credible access to patronage networks and shield perpetrators from repercussions for engaging in violence. Although locally weak parties have incentives to use violence to disturb their opponent’s territorial control, organizationally weak and cash-strapped African parties often lack the capacity to do so outside their strongholds. Indeed, empirical research has shown that African incumbent parties are responsible for the bulk of pre-election violence on the continent (Birch & Muchilinski, 2017; Hafner-Burton, Hyde & Jablonski, 2014; Straus & Taylor, 2012). In the Zambia 2016 elections, the ruling party had enough resources to recruit predominantly young, local, underemployed men in the city to work in political campaigns. These men were paid small sums of money and were often equipped with military-style attire to project strength and instill fear in rival parties’ supporters. The cadres operated like organized crime groups, extracting small rents from bus drivers, taxi drivers, and market vendors under the protection of the ruling party.⁷ Economic benefits that cadres received owing to their activities on behalf of the ruling party increased the cadres’ stakes in the election. Many of the cadres used for violence by the ruling party were imported from Lusaka and Copperbelt, two major urban areas, and transported to populous constituencies in nearby opposition strongholds in places like Southern and North Western Province. Cadres came in buses and were prepared to act violently, particularly when their president was campaigning (Authors’ interview, Solwezi, 19 December 2016). The incumbent party’s ability to control these cadres provided capacity advantages vis-à-vis the opposition with regard to mobilizing resources for violence (Authors’ interview, Lusaka, 2 January 2017).

The resource advantage enjoyed by African incumbent parties makes them more capable of perpetrating violence in opposition parties’ strongholds. Violence in ‘foreign’ territory can be enabled by control of the formal state machinery. For instance, Hassan (2017) shows that Kenyan President Moi’s government tended to post co-ethnic Kalenjin security officers in opposition strongholds to enhance government coercive capacity. Government parties may also use state resources to transport supporters, hire thugs, and recruit party foot soldiers to perpetrate violence across constituencies outside their home turf. For example, in the Zambia 2016 elections, the opposition stronghold, Southern Province, was particularly affected by pre-election violence. Much of this was connected with ruling party rallies. In a focus group with domestic election observers, one observer explained this violence with territorial politics.

Southern Province is known to be an opposition area, so when there was a [PF] campaign rally we saw all these new faces. And they came prepared. They were going to a zone where they weren’t wanted. People were being imported from one province to another, so that fueled violence. You see, when you’re in another world, you can act violently. (Authors’ interview, Choma, 22 December 2016)

The logic underpinning H1 suggests that pre-election violence is more likely in uncompetitive constituencies. However, the argument above implies that we should witness more violence in opposition-dominated constituencies. As such, constituencies are likely to see spiraling violence between a locally strong opposition with the incentives to maintain territorial control and a ruling party with the incentives and resources to use violence to defend their presence in opposition strongholds. The second hypothesis is therefore:

H2: Pre-election violence is more likely in constituencies that strongly support the opposition than in those that do not.

Case introduction and case selection

Zambia held executive, parliamentary, and local elections in August 2016. Parliamentary elections were decided in

Figure 1.

Zambian 2016 presidential election, constituency margins of victory

single member district (SMD) elections and presidential elections were decided by a national majority election. The election was highly competitive at the national level. The 2016 election was contested just a year after the last presidential election, which was called to fill the vacancy when the then-president, Michael Sata of the Patriotic Front (PF), died. The 2015 election between Edgar Lungu (PF) and opposition front-runner Hakainde Hichilema of the United Party for National Development (UPND) ended with a small margin for PF, as Lungu defeated Hichilema by 25,575 votes, or 1.66%. The 2016 election was widely seen as a rerun of 2015. Lungu and Hichilema were again the main contenders (together they won 98% of the valid vote). However, factors including government party defections, poor economic performance, and an updated voter roll added significant uncertainty. Moreover, a recently passed constitutional amendment introduced important institutional innovation: specifically, a new provision introduced the possibility of a run-off election if no candidate reached at least 50% of the vote. This made it vital for the ruling party not only to beat its closest opponent, but also to win an outright majority (Goldring & Wahman, 2016).

Although highly competitive at the national level, the election was extremely uncompetitive subnationally, placing Zambia squarely in the category of polarized electoral systems. UPND swept the Southern, Western, and North-Western Provinces, while PF secured astonishing margins in Northern, Muchinga, Eastern, and Luapula Provinces. The two major urban regions, Lusaka and Copperbelt, voted predominantly for PF, but were more competitive. Central Province was the only province outside the major urban areas where elections were relatively competitive. Figure 1 shows there was a clear division between a solidly ‘red’ West and an overwhelmingly ‘green’ East. Although ethnicity is not irrelevant in explaining geographic polarization of Zambian voting, it would be a mistake to attribute polarization solely to

Figure 2.

Zambian pre-election violence and national and local level competition

ethnic voting. High levels of ethnic fractionalization mean that Zambian parties need to draw multi-ethnic support to be nationally viable. In 2016, most Zambians did not vote for a coethnic presidential candidate (Beardsworth, 2018b). Certain ethnically homogenous provinces such as Western and North Western had previously been highly competitive, but became uncompetitive in 2016 (Cheeseman & Hinfelaar, 2009; Larmer & Fraser, 2007). As in other African contexts, chiefs, religious leaders, and business leaders play important roles as vote brokers, but do not necessarily mobilize voters for coethnic candidates (Baldwin, 2013; Koter, 2017). Fraser (2017: 468) argues that it is a mistake to think of modern Zambian politics as primarily ethnic. Instead, Zambian parties are made up of networks of political celebrities that jump between political parties and have the ability to mobilize voters. However, patronage networks are contested at the local level and political brokers of national and local standing need to convince electorates that they are advocating for a viable candidate. Political figures representing unviable candidates are likely to lose in political influence and following.

The 2016 campaign was significantly more violent than previous contests (see Figure 2). Increased levels of violence are noted in both expert data and in public perception. The Afrobarometer survey indicates that the percentage of Zambians who fear political intimidation and violence ‘a lot’ during elections has increased from 17% after the 2006 election to 29% after the 2011 election, and to 35% after the 2016 election.⁸ Although we cannot point towards a causal relationship between polarized voting patterns and higher levels of election violence, Figure 2 shows preliminary descriptive data suggesting that the combination of high national-level competition and low local-level competition has been associated with higher levels of pre-election violence. Whereas the election in 1996 was uncompetitive at the local level,⁹ it was also highly uncompetitive at the national level. In such cases, ruling parties are not overly concerned by the rare islands of opposition support that are unlikely to impact national vote totals and have small incentives to perpetrate violence. From 2001 to 2011, elections were highly competitive at the national level, but also fairly competitive at the subnational level. Several provinces like Northwestern and Western that were uncompetitive in 2016 were highly competitive in 2011. The election that most clearly illustrates a polarized electoral system is 2016.

While most studies on election violence have concentrated on a handful of cases with extreme levels of election violence, such as Côte d’Ivoire, Nigeria, and Kenya, few have examined pre-election violence in countries with levels of violence closer to the continent average. Zambia occupies this middle ground between peaceful cases such as Botswana and extremely violent cases such as Zimbabwe.¹⁰ Zambia also shares many institutional characteristics with other countries that have highly regionalized patterns of party support (Wahman, 2017). Zambia combines a small average district magnitude with a directly elected head of government and has a high level of ethnic fragmentation (Posner, 2004). Moreover, as in many other polarized electoral systems like Ghana, Kenya, Nigeria, and Malawi, SMD parliamentary elections are held concurrently with presidential elections. This particular institutional setup makes for a hard test of our theory. In contrast to our theory, if competition creates incentives for pre-election violence, this should be particularly visible in countries such as Zambia where parliamentary elections are held in single-member districts, and local parliamentary candidates would need to engage in violence to tip the balance in close local races. However, if our theory is correct, we should see increased violence in uncompetitive constituencies, especially those dominated by the opposition. An important caveat, however, is that the argument presented here would not hold if elections are only decided in local constituencies and candidates are not competing for national majorities. For instance, in countries like Botswana, where candidates strive to win constituency majorities, not national majorities, parties do not necessarily win leverage by winning supermajorities in their stronghold constituencies.

Data and empirical strategy

We expect that pre-election violence varies across space depending on local competition and opposition party support. Unfortunately, existing datasets are unsuitable to test these hypotheses. Earlier research has used subnational election violence data from sources including the Social Conflict Analysis Database (SCAD) (Salehyan et al., 2012) and the Armed Conflict Location and Event Data Project (ACLED) (Raleigh et al., 2010). However, these datasets are likely to bias results as they are more likely to pick up violence occurring in urban districts (von Borzyskowski & Wahman, forthcoming). Urban bias is particularly problematic given the hypotheses in this article because constituency-level population density in Africa is highly correlated with variables such as competition and ethnic fragmentation (Wahman & Boone, 2018).¹¹

As an innovative alternative, we use data from our original Zambia Election Monitor Survey (ZEMS). ZEMS is a post-election survey with 464 domestic election monitors in the month after the 2016 Zambian elections.¹² Respondents were domestic election observers from the internationally supported and non-partisan Foundation for Democratic Process (FODEP) and the Southern African Center for Constructive Resolution of Disputes (SACCORD) and represented every Zambian constituency. Respondents were intentionally not a random sample; they were targeted as unbiased experts, similar to other expert surveys on election integrity (Norris, Frank & Martínez i Coma, 2014). We worked with FODEP and SACCORD to secure respondents with relevant monitor training, a clear awareness of the entire electoral cycle in their respective constituency, and no known partisan bias.¹³ Interviews were conducted via phone by seven trained research assistants. For a lengthier discussion on the data and its comparative advantages to other available alternatives, see von Borzyskowski & Wahman (forthcoming).

Dependent variables

When hypothesizing about which constituencies will have more pre-election violence, it is not entirely clear what ‘more’ means. Is ‘more’ a reflection of the number of violent events or also the severity of those events? Are political actors more likely to use more intense violence in some locations and more frequent low-scale violence in others? We are theoretically agnostic on this issue. We argue that the electoral environment would be more affected by many events than by one. Similarly, it will be more affected by one severe event than a less severe event. We therefore measure the frequency and severity

Table I.

Dependent variables

Severity	Comfortable incumbent stronghold	Incumbent stronghold	Incumbent marginal	Opposition marginal	Opposition stronghold	Comfortable opposition stronghold	Total number of constituencies
No violence or property damage	11	33	16	8	9	16	93
Low severity	5	14	6	1	4	9	39
High severity	3	4	1	2	2	12	24
Frequency
No events	11	33	16	8	9	16	93
Single event	4	16	6	2	4	16	48
Multiple events	4	2	1	1	2	5	15
Violence (binary)
No reports of violence	11	33	16	8	9	16	93
Reports of violence	8	18	7	3	6	21	63

Based on the 2016 Zambian presidential election, ‘comfortable stronghold’ means the margin of victory is at least 67%; ‘stronghold’ has a margin of victory from 34 to <67%; and a ‘marginal constituency’ has a margin of victory between >0 and <34%.

of events to ensure our results are not contingent on this conceptual issue. First, we code Severity based on whether there is: no pre-election violence (coded as 0); low-severity pre-election violence (1); and high-severity pre-election violence (2). The ‘low’ category includes events where there was property damage that did not result in total destruction of property, or violence without reported injuries or fatalities. The ‘high’ category includes when people were injured or killed, or property was completely destroyed. Our classification is based on the narratives provided by respondents, not the subjective judgment by individual monitors. To give a sense of differences between categories, the violence in Ndola Central (Copperbelt Province) is coded as low; the opposition party held a rally where supporters of the governing party threw stones, damaging the opposition candidate’s car. Conversely, Mazabuka (Southern Province) is classified as high because, during a visit by the incumbent president, government vehicles were wrecked and several people were injured and hospitalized.

Second, we code Frequency – the number of events. Given the problem with delineating events, we code frequency as an ordinal scale, rather than a count of events. We code Frequency as no events (0); a single event (1); and multiple events (2).

Third, to ensure we are not extrapolating from our data, we also estimate models based on a binary dependent variable of whether there was any violence (see Table IV in the Online appendix). For all three dependent variables, we identify violence occurring if any of the three observers in the constituency recorded violence. We have no reason to suspect that monitors would falsely report violence, and it is possible that one monitor witnessed or received credible reports of violence but the other two monitors did not, particularly in many of the geographically large constituencies in rural Zambia.¹⁴ Table I contains descriptive statistics of how many constituencies experienced these types of violence.

Figure 3 maps the values of the various dependent variables.¹⁵ The patterns of pre-election violence displayed in Figure 3 correspond well with our hypotheses: violence is widespread across the country and not confined to competitive urban and peri-urban areas. For instance, highly PF-dominant constituencies in Muchinga Province show high severity and frequency of pre-election violence. In accordance with H2, there are clusters of severe violence in UPND-dominated Southern and North-Western provinces, particularly in the northern part of Southern province, close to Lusaka, and the eastern part of North-Western province, close to Kitwe. In Southern Province, most of the violence happened in constituencies such as Monze, Choma, and Mazabuka, all towns connected to the T1 highway running between Lusaka and Livingstone. In focus groups, election observers noted how opposition constituencies with strong connectivity to Lusaka and the Copperbelt were primary targets of election violence as the government party could easily transport their supporters to these areas during campaigns.

Figure 3.

Zambia 2016 presidential election, severity and frequency of pre-election violence

Clashes often happened between government and opposition supporters during presidential rallies.

Independent variables

We first hypothesized that pre-election violence is more likely in uncompetitive constituencies. Competition is measured by the runner-up’s number of votes divided by the winner’s votes; higher values indicate a constituency is expected to be more competitive (Cox, 1997).¹⁶

Our second hypothesis expects that pre-election violence is more likely in constituencies that strongly support the opposition, and our third hypothesis suggests this is especially likely in constituencies with high connectivity to urban government strongholds. We measure the strength of a constituency’s support for the opposition by Opposition votes. This is the number of thousands of votes that the opposition presidential candidate received in the previous (2015) election.¹⁷

We also include a number of control variables. We account for Population density, which measures the number of people per km² (Central Statistical Office, 2012); General violence, approximated by ACLED;¹⁸ Wealth, approximated by the use of night light density data;¹⁹ Literacy, measured as the percentage of literate citizens over the age of 16 (Central Statistical Office, 2012); and Ethnic fractionalization, calculated with a Herfindahl Index (Central Statistical Office, 2012).²⁰

The ordinal nature of our dependent variables – Severity and Frequency – where the distances between the categories are unknown, violates an assumption of the linear regression model. We therefore estimate ordered

Table II.

Ordered probit models of the Severity and Frequency of pre-election violence

	Hypothesis 1		Hypothesis 2
	Model 1	Model 2	Model 3	Model 4
Dependent variable	Severity	Frequency	Severity	Frequency
Opposition votes			.06* (.02)	.06* (.02)
Competition	–1.52** (.50)	–1.32** (.43)	–1.15* (.50)	–.97* (.42)
Population density	.00 (.00)	–.00* (.00)	–.00 (.00)	–.00* (.00)
General violence	.04 (.62)	–.04 (.64)	–.11 (.64)	–.18 (.66)
Wealth	–.01 (.05)	.08 (.05)	.03 (.07)	.12 (.07)
Literacy	–.00 (.01)	–.01 (.01)	–.01 (.01)	–.01 (.02)
Ethnic fractionalization	.96 (.71)	.65 (.66)	.73 (.68)	.43 (.61)
τ₁	.08 (.85)	–.09 (.89)	–.10 (.79)	–.27 (.85)
τ₂	.90 (.81)	1.03 (.89)	.75 (.76)	.89 (.84)
Pseudo R ²	.03	.03	.06	.06
N	156

Data come from ZEMS 2016. Standard errors clustered by province.

***p < 0.001, **p < 0.01, *p < 0.05.

probit models for the dependent variables based on the severity and frequency of election violence, and probit models for the binary indicator.²¹ We cluster standard errors by province to account for the lack of independence between constituencies within the same province.²²

Analysis

We find strong support for both hypotheses irrespective of whether we use Severity or Frequency of pre-election violence as the dependent variable. The full results are in Table II.

H1 expects that pre-election violence is more likely in uncompetitive constituencies. We find strong support for our expectation that, contrary to previous findings, uncompetitive constituencies were more likely to experience pre-election violence (Models 1 and 2 in Table II). As Figure 4 shows, non-competitive constituencies experienced significantly more minor and severe pre-election violence.²³ There is a less than 20% likelihood that the most competitive constituencies experience any form of pre-election violence.²⁴

H2 expects that pre-election violence is more likely in constituencies that strongly support the opposition; Models 3 and 4 in Table II suggest there is also support for this hypothesis. To demonstrate the substantive effect, we depict the probability of the different categories of Severity based on varying levels of Opposition votes in Figure 5, holding all other variables constant.²⁵ As the opposition’s support in a constituency increases, the likelihood of pre-election violence, both relatively minor and severe, notably increases. In particular, the likelihood of severe pre-election violence goes from less than .1 to around .5 at the highest value of Opposition votes.²⁶

A potential critique of our research design is that the observable implications for H1 and H2 are similar. Recall that H1 expects that pre-election violence is more likely in constituencies where competition in general is low, while H2 expects that pre-election violence is more likely in constituencies where the opposition is increasingly dominant. It is possible that the entire variation in election violence can be explained by the level of support for the opposition. However, while much of the most severe election violence occurred in opposition strongholds, our data include cases of violence organized by locally strong parties in incumbent party strongholds. However, to rule out the possibility that violence is strictly a consequence of opposition strength, we

Figure 4.

Probability of Severity of pre-election violence based on Competition

Figure 5.

Probability of Severity of pre-election violence based on Opposition votes

Table III.

Probit and ordered probit models to probe the mechanisms of pre-election violence

	Model 5	Model 6	Model 7
Dependent variable	Violence (binary)	Severity	Frequency
PF share	–.04* (.02)	–.05** (.02)	–.04* (.02)
PF share²	.00* (.00)	.00* (.00)	.00* (.00)
Population density	–.00 (.00)	.00 (.00)	–.00** (.00)
General violence	–.37 (.66)	–.01 (.63)	–.08 (.64)
Wealth	.09 (.10)	.04 (.07)	.12 (.07)
Literacy	–.01 (.02)	–.01 (.02)	–.01 (.02)
Ethnic fractionalization	.76 (.76)	.79 (.77)	.51 (.69)
Constant	.66 (1.17)
τ₁		–.66 (1.05)	–.71 (1.11)
τ₂		.16 (1.01)	.42 (1.11)
Pseudo R ²	.04	.04	.04
N	156

Data come from ZEMS 2016. Standard errors clustered by province.

***p < 0.001, **p < 0.01, *p < 0.05, ^†p < 0.10.

compare government strongholds, competitive constituencies, and opposition strongholds. We therefore estimate ordered probit and probit models including a quadratic function of the share of the incumbent vote in 2015.²⁷ This serves as an indicator for whether the incumbent or opposition was expected to win in 2016, and thus whether violence occurred in constituencies where the lack of competition was driven by either incumbent or opposition dominance. The results are displayed in Table III.

The results show it is not just opposition dominance that influences pre-election violence. The statistically significant coefficient for PF share ² buttresses our claims from Models 1 and 2 in Table II that a lack of subnational competition, regardless of whether it is PF or UPND that is dominant, makes pre-election violence more likely than in competitive constituencies. The predicted probability of Violence (binary) based on this function is displayed in Figure 6. This finding is also important to address an important alternative explanation to our findings: that election violence was mostly a form of spontaneous protest by opposition supporters towards the general quality and bias of the election. Although it is true that violence was more common in opposition constituencies, we also see higher levels of violence in ruling party strongholds compared to more competitive constituencies. Similar to the opposition, ruling parties also have incentives to maintain control over their strongholds and prevent the opposition from making inroads. For instance, when Hichilema tried to campaign in Shingwang’ndu (Luapula), a PF stronghold, PF cadres attacked his helicopter and managed to abort the opposition leader’s rally.

The reverse J-shaped relationship between PF support and pre-election violence is precisely what we expect given H1 and H2: violence is more common in uncompetitive constituencies but particularly those dominated by the opposition. The highest propensity for pre-election violence occurs when PF vote share is at the lowest level. The lowest propensity for pre-election violence occurs when PF vote share approaches 50%. The propensity for pre-election violence increases again as the PF vote share rises above 50%.

Robustness tests

We conduct several robustness tests to confirm our findings. As discussed, we estimate probit models using binary indicators of pre-election violence. We also estimate models using all three dependent variables with the squared term of PF share to show the observable implications of the two hypotheses are distinct and supported. Long & Freese (2014: 310) also caution against analyzing a model as ordinal when the dependent variable does not warrant this. We therefore estimate multinomial probit models, as well as re-estimating the original models using district-clustered standard errors. We then re-estimate our models using Posner’s (2004) PREG index to account for fractionalization among only politically relevant ethnic groups. We also show our findings are robust to an alternate measure of competition. Finally, we re-estimate our models using a dependent variable of the number of monitors who identified violence in each constituency. We then provide the figures showing substantive effects using Frequency of pre-election violence as the dependent

Figure 6.

Probability of pre-election violence based on low competition due to incumbent or opposition dominance

variable to demonstrate that the substantive effects are similar to Severity. Full details of these tests are in the Online appendix. The results are robust to these alternate specifications.

Conclusion

Scholars of African politics have argued that a lack of competition is one of the fundamental problems in contemporary African democracy (Lynch & Crawford, 2011; Rakner & van de Walle, 2009). With power highly concentrated in the executive and economic resources overwhelmingly centered with the state machinery (Bratton & van de Walle, 1997), government parties have generally controlled the national electoral arena. Consequently, entire groups have been effectively sidelined in the democratic project and poorly represented at the national level. However, lack of competition is not only a national phenomenon. In fact, with high degrees of political polarization, competition remains significantly lower at the local than at the national level (Koter, 2017). A number of African countries, such as Kenya, Malawi, Nigeria, and Zambia, have recently experienced electoral turnovers. Nevertheless, most African voters, even in these nationally competitive systems, belong to local political contexts where competition is virtually non-existent.

Given earlier research suggested that national-level competition is conducive to pre-election violence (Fjelde & Höglund, 2016; Hafner-Burton, Hyde & Jablonski, 2014), it might seem puzzling that problems of pre-election violence persist despite low levels of subnational competition. However, our results show that low subnational competition is, in fact, part of the problem. We argue that uncompetitive constituencies are more prone to violence due to a combination of incentives by locally dominant parties and locally weak parties to contest and maintain territorial control. Whereas locally dominant parties have incentives to preserve their hegemony, shrink the democratic space, and establish a no-go zone for political competition, locally weak parties have incentives to use violence to disturb electoral dynamics and forge their ways into foreign territory.

The empirical scope of this article has been Africa and we have particularly focused on polarized electoral systems. Political parties in Africa have remained less nationalized than parties in other comparable regions with shorter contemporary histories of multiparty democracy, such as Latin America, Eastern Europe, and Asia (Golosov, 2016; Wahman, 2017). Regionalism and lack of local competition have been particularly noted in much of the scholarship on African democracies (e.g. Kaspin, 1995; Ferree & Horowitz, 2010), but the story presented here is not uniquely Africa. In fact, regionalism is a striking feature for much of the developing world where low spatial integration, lack of infrastructure, spatial inequality, local grievances, and ethnic segregation have created distinct local market places. Regional variations in voting remain strong in countries as different as Mexico (Harbers, 2017), India (Chhibber & Verma, 2018), and the Ukraine (Barrington & Herron, 2004).

We would particularly like to highlight three important further implications from this study on the broader study of election violence, beyond the African context. First, we show that local political geography has consequences for the probability of election violence. Cross-national research on election violence should further consider interactions between subnational and national-level dynamics and how such dynamics may create environments with varying propensities for election violence. Although much of the conflict literature has moved in the direction of bridging micro-, meso-, and macro-level approaches to violence (Balcells & Justino, 2014), more election violence research in this vein is required. Our findings suggest that it is insufficient to account for national-level competition to explain varying levels of election violence. One must also consider the distribution of electoral support across space. It is conceivable that the highest levels of election violence will be observed in countries that combine national-level competition and subnational dominance.

Second, in this article we have stressed the theoretical importance of not narrowly focusing on the micro-level consequences of election violence. As the field has moved more towards identifying the causal effects of election violence on micro-level electoral behavior, we risk underestimating the way in which election violence shapes the electoral environment in more fundamental ways. This is particularly true in contexts where the electoral marketplace is essentially local and political nationalization is low. Survey experiments may identify that voters are less likely to vote for perpetrators of violence (Gutiérrez-Romero & LeBas, 2020; Mares & Young, 2016; Rosenzweig, 2017), but the external validity of such experiments may be questionable if violence is used as a tool to effectively distort local political competition and shape the electoral environment in which voters make decisions on political action.

Lastly, these findings add to an emerging literature on subnational authoritarianism around the globe from regions such as Latin America (Gibson, 2013; Giraudy, 2015) and Eastern Europe and Central Asia (McMann, 2006), a literature that has been primarily focused on political economy but been less attentive to repression as a more immediate tool to maintain subnational dominance. Further research is required to understand the local electoral market and the structures underpinning dominance, particularly with reference to violence.

Footnotes

Replication data

The dataset and do-files for the empirical analysis in this article, along with the Online appendix, can be found at .

Acknowledgments

We are grateful to Chimfwembe Mweenge at the Foundation for Democratic Process (FODEP), Boniface Cheembe at the Southern African Center for the Constructive Resolution of Disputes (SACCORD), and Marja Hinfelaar at the Southern African Institute for Policy and Research (SAIPAR). Josephine Chanda, Hope Chipalo, Andrew Kasiya, Lwiindi Mweene, Beauty Nalwendo, Beverly Sichilenge, and Prisca Tembo provided outstanding research assistance. We also thank Bryce Reeder and Kostadis Papaioannou for help with processing geographical data, and Laia Balcells, Catherine Boone, Jeffrey Conroy-Krutz, Scott Straus, and Nicolas van de Walle for invaluable comments. Lastly, we thank the editors of the special issue: Sarah Birch, Ursula Daxecker, and Kristine Höglund, as well as the four anonymous reviewers. This research was approved by University of Missouri IRB Office, IRB#2004569.

Funding

The study was supported by grants from the Economic and Social Research Council (ES/R005753/1), the Magnus Bergvall Foundation (2015-00698) and the University of Missouri Research Council (URC-16-023).

ORCID iD

Michael Wahman

Notes

References

Asunka

Joseph

Brierley

Sarah

Golden

Miriam

Kramon

Eric

Ofosu

George

(forthcoming) Electoral fraud or violence: The effect of observers on party manipulation strategies. British Journal of Political Science. https://doi.org/10.1017/S0007123416000491.

Balcells

Laia

Justino

Patricia

(2014) Bridging micro and macro approaches on civil wars and political violence. Journal of Conflict Resolution 58(8): 1343–1359.

Baldwin

Kate

(2013) Why vote with the chief? Political connections and public goods provision in Zambia. American Journal of Political Science 57(4): 794–809.

Barrington

Lowell W

Herron

Erik

(2004) One Ukraine or many? Regionalism in Ukraine and its political consequences. Nationalities Papers 32(1): 53–86.

Beardsworth, Nicole (2018a) Persuading partisans or courting co-ethnics: UPND campaigns in Zambia. Unpublished manuscript.

Beardsworth

Nicole

(2018b) Electoral coalition-building among opposition parties in Zimbabwe, Zambia and Uganda from 2000 to 2017 . Unpublished PhD dissertation, University of Warwick.

Bekoe

Dorina A

Burchard

Stephanie M

(2017) The contradictions of pre-electoral violence: The effects of violence on voter turnout in sub-Saharan Africa. African Studies Review 60(2): 73–92.

Birch

Sarah

(2007) Electoral systems and electoral misconduct. Comparative Political Studies 40(12): 1533–1556.

Birch

Sarah

(2011) Electoral Malpractice. Oxford: Oxford University Press.

10.

Birch

Sarah

Muchilinski

David

(2017) Dataset of countries at risk of electoral violence. Terrorism and Political Violence 9(28): 1–20.

11.

Bob-Milliar

George M

(2014) Party youth activists in low-intensity violence in Ghana. African Studies Quarterly 15(1): 125–152.

12.

Boone

Catherine

(2011) Politically allocated land rights and the geography of election violence: The case of Kenya in the 1990s. Comparative Political Studies 48(2): 173–202.

13.

Boone

Catherine

(2016) Spatial Inequalities in African Political Economy Dataset. LSE RIIF 2016 (1-BRD-B076).

14.

Bratton

Michael

(2008) Vote buying and violence in Nigerian elections. Electoral Studies 27(4): 621–631.

15.

Bratton

Michael

Masunungure

Eldred

(2007) Popular reactions to state repression: Operation Murambatsvina in Zimbabwe. African Affairs 106(422): 21–45.

16.

Bratton

Michael

Walle

Nicolas van de

(1997) Democratic Experiments in Africa: Regime Transitions in Comparative Perspective. Cambridge: Cambridge University Press.

17.

Brierley

Sarah

Kramon

Eric

(2017) Party campaign strategies. Working paper (https://www.sarahbrierley.com/uploads/2/7/7/1/27711295/bk_campaigns_13dec2017.pdf).

18.

Brosché

Johan

Fjelde

Hanne

Höglund

Kristine

(2020) Electoral violence and the legacy of authoritarian rule in Kenya and Zambia. Journal of Peace Research 57(1): 111–125.

19.

Central Statistical Office (2012) Zambia: 2010 Census of Population and Housing. Lusaka: Central Statistical Office.

20.

Cheeseman

Nicholas

Hinfelaar

Marja

(2009) Parties, platforms, and political mobilization: The Zambian presidential election of 2008. African Affairs 109(434): 51–76.

21.

Chen

Lanhee J

Reeves

Andrew

(2011) Turning out the base or appealing to the periphery? American Politics Research 39(3): 534–556.

22.

Chhibber

Pradeep K

Verma

Rahul

(2018) Ideology and Identity: The Changing Party System of India. New York: Oxford University Press.

23.

Collier

Paul

Vicente

Pedro C

(2012) Violence, bribery, and fraud: The political economy of elections in sub-Saharan Africa. Public Choice 153(1): 117–147.

24.

Conroy-Krutz

Jeffrey

(2013) Information and ethnic politics in Uganda. British Journal of Political Science 43(2): 345–373.

25.

Cox

Gary

(1997) Making Votes Count. Cambridge: Cambridge University Press.

26.

Daxecker

Ursula

(2014) All quiet on election day? International election observation and incentives for pre-election violence in African elections. Electoral Studies 34(June): 232–243.

27.

Daxecker

Ursula

(2020) Unequal votes, unequal violence: Malapportionment and election violence in India. Journal of Peace Research 57(1): 156–170.

28.

Dercon

Stefan

Gutiérrez-Romero

Roxana

(2012) Triggers and characteristics of the 2007 Kenyan election violence. World Development 40(4): 734–744.

29.

Dionne

Kim Yi

Dulani

Boniface

(2014) On the eve of Malawi’s election. Washington Post 9 May.

30.

Eck

Kristine

(2012) In data we trust? A comparison of UCDP, GED and ACLED conflict events datasets. Cooperation and Conflict 47(1): 124–141.

31.

Endersby

James W

Galatas

Steven E

Rackaway

Chapman B

(2002) Closeness counts in Canada: Voter participation in the 1993 and 1997 federal elections. Journal of Politics 64(2): 610–631.

32.

Ferree

Karen

Horowitz

Jeremy

(2010) Ties that bind? The rise and decline of ethno-regional partisanship in Malawi, 1994–2009. Democratization 17(3): 534–563.

33.

Fjelde

Hanne

(2020) Political party strength and electoral violence. Journal of Peace Research 57(1): 140–155.

34.

Fjelde

Hanna

Höglund

Kristine

(2016) Electoral institutions and electoral violence in sub-Saharan Africa. British Journal of Political Science 46(2): 297–320.

35.

Fraser

Alastair

(2017) Post-populism in Zambia: Michael Sata’s rise, demise and legacy. International Political Science Review 38(4): 456–472.

36.

Gibson

Edward

(2013) Boundary Control: Subnational Authoritarianisms in Federal Democracies. New York: Cambridge University Press.

37.

Giraudy

Agustina

(2015) Democrats and Autocrats: Pathways of Subnational Undemocratic Regime Continuity within Democratic Countries. Oxford: Oxford University Press.

38.

Goldring

Edward

Wahman

Michael

(2016) Democracy in reverse: The 2016 general election in Zambia. Africa Spectrum 51(3): 107–121.

39.

Goldring

Edward

Wahman

Michael

(2018) Fighting for a name on the ballot: Constituency-level analysis of nomination violence in Zambia. Democratization 25(8): 996–1015.

40.

Golosov

Grigorii V

(2016) Party system nationalisation in sub-Saharan Africa: Empirical evidence and an explanatory model. International Area Studies Review 19(3): 231–248.

41.

Gutiérrez-Romero

Roxana

LeBas

Adrienne

(2020) Does electoral violence affect willingness to vote and voting choice? Conjoint analysis of a vignette experiment. Journal of Peace Research 57(1): 77–92.

42.

Hafner-Burton

Emilie

Hyde

Susan D

Jablonski

Ryan

(2014) When do governments resort to violence? British Journal of Political Science 44(1): 149–179.

43.

Harbers

Imke

(2017) Spatial effects and party nationalization: The geography of partisan support in Mexico. Electoral Studies 47(June): 55–66.

44.

Hassan

Mai

(2017) The strategic shuffle: Ethnic geography, the internal security apparatus, and elections in Kenya. American Journal of Political Science 61(2): 382–395.

45.

Höglund

Kristine

(2009) Electoral violence in conflict-ridden societies: Concepts, causes, and consequences. Terrorism and Political Violence 21(3): 412–427.

46.

Horowitz

Jeremy

(2010) The ethnic logic of campaign strategy in diverse societies. Comparative Political Studies 49(3): 324–356.

47.

Horowitz

Jeremy

Long

James

(2016) Strategic voting, information, and ethnicity in emerging democracies. Electoral Studies 44(December): 351–361.

48.

Ichino

Nahomi

Nathan

Noah

(2013) Crossing the line: Local electoral geography in Ghana. American Political Science Review 107(2): 344–361.

49.

Kanyinga

Karuti

(2009) The legacies of the white highlands. Journal of Contemporary African Studies 27(3): 325–344.

50.

Kaspin

Deborah

(1995) The politics of ethnicity in Malawi’s democratic transition. Journal of Modern African Studies 33(4): 595–620.

51.

Koter

Dominika

(2017) Beyond Ethnic Politics in Kenya. Cambridge: Cambridge University Press.

52.

Larmer

Miles

Fraser

Alastair

(2007) Of cabbages and King Cobra: Populist politics and Zambia’s 2006 election. African Affairs 106(425): 611–637.

53.

LeBas

Adrienne

(2006) Polarization as craft: Party formation and state violence in Zimbabwe. Comparative Politics 38(4): 419–438.

54.

Letsa

Natalie Wenzell

(2019) The political geography of electoral autocracies: The influence of party strongholds on political beliefs in Africa. Electoral Studies 60(August): 1–12.

55.

Long

J Scott

Freese

Jeremy

(2014) Regression Models for Categorical Dependent Variables Using Stata. College Station: Stata.

56.

Lynch

Gabrielle

Crawford

Gordon

(2011) Democratization in Africa 1990–2010. Democratization 18(2): 275–310.

57.

Malik

Aditi

(2018) Constitutional reform and new patterns of electoral violence. Commonwealth and Comparative Politics 56(3): 340–359.

58.

Mares

Isabela

Young

Lauren

(2016) Do voters punish electoral malfeasance? Working paper (https://www.academia.edu/31970095/Do_voters_punish_electoral_malfeasance).

59.

McMann

Kelly M

(2006) Economic Autonomy and Democracy: Hybrid Regimes in Russia and Kyrgyzstan. New York: Cambridge University Press.

60.

Mueller

Susanne D

(2008) The political economy of Kenya’s crisis. Journal of Eastern African Studies 2(2): 185–210.

61.

Norris

Pippa

Frank

Richard W

Coma

Ferran Martínez i

(2014) Measuring electoral integrity around the world. PS: Political Science and Politics 47(4): 789–798.

62.

Paget

Dan

(2019) The rally intensive campaign: A distinct form of electioneering in Africa and beyond. International Journal of Press/Politics 24(4): 444–464.

63.

Posner

Daniel

(2004) Measuring ethnic fractionalization in Africa. American Journal of Political Science 48(4): 849–863.

64.

Rakner

Lise

Walle

Nicolas van de

(2009) Opposition weakness in Africa. Journal of Democracy 20(3): 108–121.

65.

Raleigh

Clionadh

Linke

Andrew

Hegre

Håvard

Karlsen

Joakim

(2010) Introducing ACLED: Armed Conflict Location and Event Data. Journal of Peace Research 47(5): 1–10.

66.

Rauchenbach

Mascha

(2017) The allocation of campaign rallies in Ghana’s 2012 elections. Working paper.

67.

Rauchenbach

Mascha

Paula

Katrin

(2019) Intimidating voters with violence and mobilizing them with clientelism. Journal of Peace Research 56(5): 682–696.

68.

Reeder

Bryce

Seeberg

Merete Bech

(2018) Fighting your friends? A study of intra-party violence in sub-Saharan Africa. Democratization 25(6): 1033–1051.

69.

Robinson

James A

Torvik

Ragnar

(2009) The real swing voter’s curse. American Economic Review 99(2): 310–315.

70.

Rohrschneider

Robert

(2002) Mobilizing versus chasing: How do parties target voters in election campaigns? Electoral Studies 21(3): 367–382.

71.

Rosenzweig

Steven C

(2017) Dangerous disconnect: Voter backlash, elite misperception, and the costs of violence as an electoral tactic. Working paper (http://cega.berkeley.edu/assets/miscellaneous_files/117_-_Rosenzweig_Dangerous_Disconnect_-_May_14_2015_-_ABCA.pdf)

72.

Rundlett

Ashlea

Svolik

Milan

(2016) Deliver the vote! Micromotives and macrobehavior in electoral fraud. American Political Science Review 110(1): 180–197.

73.

Sack

Robert

(1986) Human Territoriality: Its Theory and History. Cambridge: Cambridge University Press.

74.

Salehyan

Idean

Hendrix

Cullen

Hamner

Jesse

Case

Christina

Linebarger

Christopher

Stull

Emily

Williams

Jennifer

(2012) Social conflict in Africa: A new dataset. International Interactions 38(4): 503–511.

75.

Straus

Scott

(2011) ‘It’s sheer horror here’: Patterns of violence during the first four months of Côte D’Ivoire’s post-electoral crisis. African Affairs 110(440): 481–489.

76.

Straus

Scott

Taylor

Charlie

(2012) Democratization and election violence in sub-Saharan Africa, 1990–2008. In: Berkoe

Dorina

(ed.) Voting in Fear: Electoral Violence in Sub-Saharan Africa. Washington, DC: USIP, 15–38.

77.

Taylor

Charles F

Pevehouse

Jon CW

Straus

Scott

(2017) Perils of pluralism: Election violence and incumbency in sub-Saharan Africa. Journal of Peace Research 54(3): 374–411.

78.

von Borzyskowski

Inken

(2019) The Credibility Challenge: How Democracy Aid Influences Election Violence. Ithaca, NY: Cornell University Press.

79.

von Borzyskowski

Inken

Kuhn

Patrick M

(2020) Dangerously informed: Voter information and pre-electoral violence in Africa. Journal of Peace Research 57(1): 15–29.

80.

von Borzyskowski

Inken

Wahman

Michael

(forthcoming) Systematic measurement error in election violence data: Causes and consequences. British Journal of Political Science. https://doi.org/10.1017/S0007123418000509.

81.

Wahman

Michael

(2017) Nationalized incumbents and regional challengers: Opposition- and incumbent-party nationalization in Africa. Party Politics 23(3): 309–322.

82.

Wahman

Michael

Boone

Catherine

(2018) Captured countryside? Stability and change in sub-national support for African incumbent parties. Comparative Politics 50(2): 189–216.

83.

Weidmann

Nils

(2015) A closer look at reporting bias in conflict data. American Journal of Political Science 60(1): 206–218.

84.

Weidmann

Nils B

Schutte

Sebastian

(2017) Using night light emissions for the prediction of local wealth. Journal of Peace Research 54(2): 125–140.

85.

Young

Lauren

(2020) Who dissents? Self-efficacy and opposition action after state-sponsored election violence. Journal of Peace Research 57(1): 62–76.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB