Literature review

Electoral violence affects electoral participation. Systematic studies of the impact of electoral violence on electoral participation show a negative association between the two. A cross-sectional study revealed that voters who have experienced threats of election violence at the polls were less likely to vote in Nigeria (^{Bratton, 2013}, p. 129). A cross-national study also found the same globally (^{Norris, 2014}, p. 143). In this case, it can be argued that electoral violence is used by the incumbent to prevent the opposition supporters from turning out (^{Hafner-Burton et al., 2014}).

Another cross-national study showed that “electoral violence affects willingness to (not) vote, but this effect differs by partisan attachment and specific electoral context” (^{Burchard, 2015}, p. 143). In cases where opposition supporters were exposed to electoral violence they were more willing to vote. Although not significant, in cases where the incumbent supporters were exposed to violence they were less likely to vote (^{Burchard, 2015}, p. 139). Depending on the type of violence (whether strategic or incidental, the identity of the perpetrators and or the message that it is meant to convey to voters) and how it affects directly individuals, electoral violence can be used to mobilize voters to vote (see Travaglianti, 2014), to prevent them to turnout or to punish victors (^{Bekoe & Burchard, 2017}; ^{Burchard, 2015}).

These studies measured electoral violence and electoral participation only by single items. ^{Burchard (2015}) and ^{Bratton (2013}) measured electoral violence by threat of violence at the polls. ^{Norris (2014}) did the same but her measurement of electoral violence was integrated in the scope of an electoral malpractice index together with other items. On electoral participation, they measured it by voting participation (^{Bratton, 2015}; ^{Burchard, 2015}; ^{Norris, 2014}).

The quality of elections, in general, has an impact on electoral participation. “Elections work well (when) they can provide the main opportunity for most people to participate in politics” (^{Powell, 2000}). ^{Norris (2014}) found that voter turnout tended to be high when there was electoral integrity - that is, when votes were counted fairly, elections were fair, voters were offered genuine choice in the elections and journalists provided fair coverage of elections. It tended to be low when there was electoral malpractice, for example, rich people bought elections, television news favoured the governing party, voters were bribed, opposition candidates were prevented from running and voters were threatened with violence at the polls (p. 143).

^{Norris (2014}) notes that “too often elections around the globe are deeply flawed” meaning that they do not have integrity. In Africa the scenario is also of concern. As ^{Bleck and Walle (2019}) described it: “extensive intimidation of the opposition” and suspension of the opposition newspaper took place, respectively, in the 2016 Gambian and Zambian elections (pp. 2-3). Districts were gerrymandered in Zimbabwe ahead of the 2008 election (^{Cheeseman & Klaas, 2018}, p. 41); and candidates in Nigeria, Uganda, Ghana and Kenya were found to distribute largesse and buy votes (^{Cheeseman & Klaas, 2018}, pp. 61-92).

Performance evaluation (of institutional actors, such as: the incumbent or opposition parties and individual actors like the incumbent or opposition candidates) affect voting behaviour. It is based on rational choice theory that assumes that voters

act rationally - that is, before deciding on a course of action (such as buying a particular product) they weigh up the costs and benefits of the various alternatives and they will take the decision which maximizes the benefits and minimizes the costs to themselves. (^{Denver, 2007})

Thus, voters vote and make a party choice based on the performance evaluation they make of political parties and candidates by looking at current political issues, campaign events and personalities of party leaders or candidates.

^{Bratton et al. (2005}) found a negative relationship between performance evaluation of the president and voting. But Bratton et al. (2013) discovered a positive connection of prospective sociotropic economic evaluation and economic policy performance with an intention to vote for the ruling party (p. 92). In the African context of electoral violence, ^{Collier and Vicente (2012)} found that voters that were exposed to violence and were the targets of an institutional actor (ActionAid International Nigeria’s anti-violence campaign) were more likely to vote than others. As is shown later in this study we analyse the effect of evaluation of government performance on election violence.

Electoral participation can be also predicted from the habit of participating in politics in general. ^{Bratton et al. (2005}) found that the act of participating in politics after elections by contacting and communing led Africans to participate on election day and even in other forms of political participation like protesting. They also found that participating on election day led voters to contacting and communing (^{Bratton et al., 2005}, p. 297). ^{Bechtel et al. (2018}) found that the long-term and spillover effects of compulsory voting in the Swiss canton of Vaud (1900-1970) increased turnout in federal referendums by 30 percentage points.

Yet the context where people live is a predictor of people’s behaviours. Context matters (^{Goodin & Tilly, 2006}). Context affect people’s attitudes and behaviours. People living in dominant party systems will be expected to be supplied less with democratic goods like freedoms and equality that are necessary for them to participate in elections in particular and in politics in general. This is because dominant parties are detrimental to the quality of democracy (^{Walle & Buttler, 1999}) and the prospects of liberal democracy (du ^{Toit, 1999}). They pose the greatest threat to consolidation of third wave democracies (^{Huntington, 1996}) as they can degenerate into dominant authoritarian or hegemonic parties.

Whilst scholars have classified dominant parties by counting the effective number of parties, this is found to be problematic and flawed (^{Bogaards, 2004}). In this study we use Sartori’s counting rule of dominant party meaning “the party that have secured at least three consecutive parliamentary majorities and the presidency” (^{Bogaards, 2004}, p. 191). Thus, the dominant party systems in our sample are the following countries: Botswana, Burkina Faso, Cameroon, Gabon, Mozambique, Namibia, South Africa, Sudan, Tanzania, Togo, Uganda and Zimbabwe.

Access to news media is another driver of electoral participation. Based on studies suggesting that media exposure has a sizable impact on voting behaviour (^{Bratton et al., 2005}, p. 297; ^{Norris, 2001}; ^{Verma & Sardesai, 2014}), one would state that voters who often access news media will be more engaged in electoral participation than others. Due to recent development of news media, where we can distinguish traditional media³ from new news media⁴, different outcome could be expected: as the new news media tends not to be regulated across the world, including in Africa, and thus with less ethics on its use (^{Martens et al., 2018}), scholars would expect it to have more a negative influence than traditional ones.

The social structure also shapes people’s behaviours. Modernization theory suggests, for instance, that urban dwellers will be more likely to participate in elections, as they are more exposed to the information necessary to know the role of elections and how society functions. In contrast, in the African context, rural residents are more likely to vote (^{Bratton et al., 2005}); and they do so for the ruling parties (^{Bratton et al., 2013}, p. 92). With respect to gender, women are less likely to be interested in public affairs and discuss politics (^{Afrobarometer & CPGD, 2013}; ^{Pereira et al., 2005}; ^{Shenga & Pereira, 2009}), to be represented in parliament (^{Shenga, 2014}, p. 113) and more likely to be victim of violence (^{Bachman, 1994}). Although initial evidence from Afrobarometer found no evidence of the impact of gender on voting, African women were less likely to commune and contact (^{Bratton et al., 2005}).

In addition to electoral participation, it is also important to review the relationship between the quality of elections and democratic consolidation, as the former has implications on the later. The quality of elections has an effect on democratic legitimacy. ^{Norris (2014}) found that perceived electoral integrity had a positive impact on satisfaction with democracy while perceived malpractice had a negative effect (^{Norris, 2014}, p. 125). ^{Greenberg and Mattes (2013}) found that the perceived freeness and fairness of elections affected positively the supply of democracy, indicated by satisfaction with democracy and the extent of democracy (p. 247).

^{Greenberg and Mattes (2013}) measured the quality of elections only by free and fair elections; but ^{Norris (2014}) included a battery of items. For electoral integrity, she used the following items: votes are counted fairly, election officials are fair, voters offered genuine choice in an election, and journalists provide fair coverage of elections. For electoral malpractices, she used: rich people buy elections, television news favours the government party, opposition candidates are prevented from running and voters are threatened with violence at the polls.

Data and methods for explaining electoral participation

We perform our analysis and tests using Afrobarometer⁵ public opinion surveys conducted between 2016 and 2018 in 34⁶ African countries that have made democratic political reform⁷. The Afrobarometer surveys includes a set of questions on election violence, the quality of elections, government performance preventing election violence, political participation, media exposure, demographics, and democratic consolidation.

We measure violence by using five items of which two tap electoral violence. Fear of election violence is indicated by the survey question: “During election campaigns in this country, how much do you personally fear becoming a victim of political intimidation or violence? A lot, somewhat, a little bit or not at all?”. The last four violence items are indicated by the following questions: “… Please tell me whether, in the past two years, you have ever personally feared any of the following types of violence: 1) Violence among people in your neighbourhood or village? 2) Violence at a political rally or campaign event? 3) Violence of public protest? 4) An armed attack by political or religious extremists? [If yes:] Have you actually personally experienced this type of violence in the past two years?” Reponses ranged from: no - never; yes - feared but did not experience it; and yes - feared and experienced it.

In the model of electoral participation, the quality of elections is measured by trust in the Electoral Management Body (EMB), freeness and fairness of elections, and whether the opposition was prevented from running in elections. Trust in EMB is measured by the following Afrobarometer question: “How much do you trust EMB?” Responses could be given varying from not at all to just a little, somewhat and a lot. Freeness and fairness of elections is indicated by question: “On the whole, how would you rate the freeness and fairness of the last national election, held in XXXX? Was it: not free and fair, free and fair with major problems, free and fair but with minor problems or completely free and fair?” Opposition prevented from running is measured by: “In your opinion, during this country’s elections, how often are opposition candidates prevented from running for office?” Answers varied from never to sometimes, often and always.

But in the model of democratic consolidation, the quality of elections is boosted. Besides being specified into considerations that taps both electoral integrity and electoral malpractice, it also includes the malpractices of electoral violence. At this point, electoral integrity includes measures of trust in the EMB and freeness and fairness of elections; and electoral malpractice includes aspects related to electoral violence covered in this study and opposition being prevented from running.

To measure government performance preventing election violence we use the following question: “How well or badly would you say the current government is handling ‘Preventing political violence during election campaigns’? Very badly, fairly badly, fairly well or very well?”

Participation after election is measured by membership in a community association development group; attending a community meeting, and participating in a demonstration or protest march; joining others to raise an issue, and to request government action; and contacting the media, contacting an official for help and contacting officials. Membership in a community association development group is specified by: “Could you tell me whether you are an official leader, an active member, an inactive member, or not a member of voluntary association or community group?”

Attending a community meeting and participating in a demonstration or protest march are indicated by the questions: “Please tell me whether you, personally, have 1) Attended a community meeting [and] 2) Participated in a demonstration or protest march, during the past year?” Joining others to raise an issue, and to request government action is specified by: “Please tell me whether you, personally, have 1) Got together with others to raise an issue [and] 2) Joined others in your community to request action from government during the past year?”

Contacting the media and officials for help are specified by: “Please tell me whether you, personally, have contacted 1) the media, like calling a radio program or writing a letter to a newspaper [and] 2) a government official to ask for help or make a complaint during the past year?” Contacting officials is indicated by: “During the past year, how often have you contacted: 1) A local government councillor, 2) A member of Parliament, 3) An official of a government agency, 4) A political party official, 5) Traditional leaders, and 6) Religious leaders about some important problem or to give them your views?”

Participation on election day - that is, voting is measured by: “Understanding that some people were unable to vote in the most recent national election in XXXX, which of the following statements is true for you? You were not registered to vote. You voted in the elections. You decided not to vote. You could not find the polling station. You were prevented from voting. You did not have time to vote. You did not vote because you could not find your name in the voters’ register. You were too young to vote?” Participation before election day is specified by attending a campaign rally and working for a candidate or party “Thinking about the last national election in XXXX, did you: 1) Attend a campaign rally? and 2) Work for a candidate or party?”

Media exposure is measured by: “How often do you get news from: 1) Radio, 2) Television, 3) Newspapers, 4) Internet and 5) Social media? Never, less than once a month, a few times a month, a few times a week or every day?” Two social structure indicators are used: residential location (rural/urban) and gender.

Democratic consolidation is specified by three items: First, support for democracy: “Which of these three statements is closest to your own opinion? Statement 1: Democracy is preferable to any other kind of government. Statement 2: In some circumstances, a non-democratic government can be preferable. Statement 3: For someone like me, it doesn’t matter what kind of government we have”. Second, extent of democracy: “In your opinion, how much of a democracy is [COUNTRY] today?” Third, satisfaction with democracy: “Overall, how satisfied are you with the way democracy works in [COUNTRY]?” At public opinion individual level, democracies consolidate when, on the one hand, citizens demand it, saying ”democracy is preferable than any other forms of government”, and reject authoritarian forms of government, such as: one party rule, military rule and one-man rule. On the other hand, when they perceive that their country is being supplied of democracy - that is, saying [COUNTRY] is a full democracy, and they are satisfied with democracy (see ^{Bratton et al., 2005}; ^{Greenberg & Mattes, 2013}).

All of these variables are measured with individual level data and they will be analysed at that level. However, the analysis will be performed at a multi-level to take into account the contextual-country level dominant party variable that has been created from the countries being analysed. Since Eswatini does not have party system, the number of observations drops from 34 to 33 countries.

We test the effect of electoral violence on electoral participation controlling for other factors simultaneously, and then the impact of the quality of elections on democratic consolidation, using binary logistic multiple regression analysis. All items in the regression were coded 1-0. Items tapping supposedly the same thing were merged into a single item to create a multidimensional concept or index. In order to merge them into an index we used factor and reliability analyses⁸. As reliability analysis did produce a valid reliable index for items of demand for democracy and supply of democracy, we employ them separately. We dropped items indicating rejection of authoritarian rule to avoid many items.

Findings

The effects of electoral violence on electoral participation, controlling for other factors, are presented in Table 1. As an unexpected finding, the results show that the malpractice of electoral violence leads individuals to participate in elections. All electoral violence effects are positive. Fear of violence in an election campaign is positive and significant on participating in an election campaign rally. It also appears to be positive, but not significant, on voting. Fear of violence at a political rally or campaign event is positive and significant both on voting and attending an election campaign rally.

On the other hand, violence in general has a significant and negative effect on participation in elections. Fear of violence in a neighbourhood prevent voters from attending election campaign rallies. Fear of violence by extremist groups reduces the prospects of voting and working for parties or candidates.

The effects of other predictors of participation in elections appear mostly to support the literature. The quality of elections affects electoral behaviour. As expected, electoral integrity fosters participation in elections, whereas electoral malpractice reduces it. Africans who perceive that elections are free and fair tend to participate on election day by voting and during election campaigns by attending rallies and working for candidates or parties. Those who trust EMBs are also likely to vote. And those who view that the opposition is prevented from running are less likely to cast a vote and attend a campaign rally. But, unexpectedly, those who view that the opposition is prevented from running are likely to work for a candidate or a party.

The effect of government performance preventing election violence is significant on all of the three measures of electoral participation but is, unexpectedly, negative in one. Whilst Africans who rate positively the performance of their governments preventing violence during election campaign are likely to vote and attend an election campaign rally, they are unlikely to work for a candidate or a party.

Living in a context where “the same party secured at least three consecutive parliamentary majorities and the presidency” (^{Bogaards, 2004}) drives Africans to not cast a vote or work for a candidate or party, as expected. However, it does not with respect to attending an election campaign rally. Africans who live in a context of dominant parties are likely to attend an election campaign rally.

The habit of participation in politics in general tends to affect participation in elections. Participation after elections leads individuals to participate in elections either on election day or during the election campaign. With only two exceptions - contacting the media and attending a demonstration or protest march. All indicators tapping participation after election that are significant have positive effects. The negative effect of contacting the media on electoral participation could be explained by the media highlighting negative aspects of elections. Too often the media highlights the negative and even sometimes make the negative to appear worse (^{Rosenthal, 2009}). With respect to the negative effect of attending a demonstration or protest march it is possible that those who reported attending a demonstration or protest march are more exposed to news media. The different dimensions of electoral participation affect positively one another. Those who say that they vote in elections also say that they attend election campaign rallies and work for a candidate or party and vice versa.

Moving to news media consumption, the results tend to indicate, although not consistently, that traditional news media drives individuals to participate in elections while new news media move them away from elections. Accessing radio news makes Africans to vote and attend campaign rallies. The same applies to accessing newspaper news on voting. Television news has mixed effects. It leads Africans to abstain from voting and abstain from attending an election campaign rally but surprisingly it does lead them to work for a candidate or a party. On the other hand, we observe that new news media have negative impacts, with exception to one case. Individuals who obtain news media from the internet are less likely to attend an election campaign rally and those who do so through social media are less likely to vote and work for a candidate or a party. But, unexpectedly, social media has a positive effect on attending a campaign rally.

Last but not least, rural residents tend more than urban residents to attend an election campaign and work for a party or candidate. This might be explained by the high incidences of poverty that characterize mainly rural areas in Africa. Rural dwellers may view that there could be gains for them to participate in an election campaign rally as it is in those periods that they may receive a free t-shirt, hat or fabrics (often promoting a political party), watch free concerts in rallies and even receive money or basic food stuff. Whereas on election day they get “nothing”. Note that although it is not statistically significant, living in a rural area is negatively associated with voting. By focusing on a subset of Round 6 of Afrobarometer surveys and using more extensive measures of electoral participation, ^{Shenga (2018}) found that poor Africans were less likely to vote but more likely to attend election campaign rallies, persuade others to vote for their party or candidate and work for a party or candidate. The same trend applies to rural residents. African females appear to be mainly disengaged in politics. They are less likely to vote, attend an election campaign rally and work for a party or candidate.

The implications of the quality of elections for democracy

The effect of the quality of elections on democratic consolidation, controlling for the context of a dominant party system, is presented in Table 2. Here, following ^{Norris’s model (2014}) of the quality of elections, election integrity and election malpractice, we integrate election violence measures into what we call the quality of elections. The measures of electoral integrity in our model are: trust the EMB and free and fair elections. The measures of electoral malpractice are: our two measures of electoral violence and the opposition prevented from running.

The logistic regression results of the effect of the quality of elections are all significant and continue to hold from initial (correlation) analysis. Election malpractice affects negatively dimensions of democratic consolidation. Africans who perceive a fear of violence at a political rally or campaign event are less likely to say they live in a full democracy and are satisfied with the way democracy is functioning in their country. But they still view democracy as the most preferable form of government. Those who fear violence during an election campaign rally are less likely to view that they are being supplied with democracy in their countries as well as less likely to support democracy. Yet those who view that the opposition is prevented from running become disillusioned with democracy to the extent that they do not prefer it because they view that they are already being supplied with democracy.

On the other hand, electoral integrity affects democratic consolidation positively. Africans who trust the EMB and rate elections free and fair tend more to prefer democracy and to view their country as a full democracy and are satisfied with it.

Africans who live in a dominant party system uncritically overrate the performance of their political system by reporting that they tend to be satisfied with democracy. Yet, in a dominant party system, democratic procedures of the rule of law, accountability, participation and competition tended to be subverted by life time parties. As one would expect, other coefficients of a dominant party system are negative but not significant.

Conclusions

Scholarly work using a systematic approach finds that electoral violence tends to drive Africans away from participating in elections (^{Bratton, 2013}; ^{Burchard, 2015}). By using the same approach, with more recent data, we found the opposite to be true. Africans who perceived electoral violence were more likely to attend an election campaign rally and vote. Instead of being used by the incumbent parties to prevent the opposition supporters from turning out (^{Hafner-Burton et al., 2014}), electoral violence in Africa appears to motivate those who perceive it to participate more. This finding, which is different from previous studies, is associated with the multidimensional approach to measuring both electoral violence and electoral participation and the use of more recent dynamics of African elections. It might be also associated with the climate of fear that persists from decades of colonial and post-colonial authoritarian rules.

No less important in Africa is the effect of violence in general. Violence has a negative effect on participation in elections. Africans who perceive a threat of violence in a neighbourhood also tended to report they do not attend election campaign rallies. And those who perceive a fear of violence by extremist groups tended not to work for parties or candidates. If a neighbourhood is perceived to be unsafe, voters might leave their residential location thus losing their prospect to vote in the neighbourhood where they were registered to vote.

Although this study points out that election violence is used to mobilize voters participating in elections, this malpractice is not conducive for democratic consolidation. ^{Norris (2014}) found that the perceived electoral integrity is associated with system support; and ^{Greenberg and Mattes (2013}), more specifically, found that free and fair elections is linked positively with the supply of democracy. On the other hand, ^{Norris (2014)} found electoral malpractice is negatively associated with system support. This study confirms these findings. Those who perceive the malpractice of electoral violence are less likely to support for, and be satisfied with democracy, and are less likely to evaluate their country as full democracy. Whilst those individuals who perceive the quality and integrity of elections are more likely to support, and be satisfied with, democracy.