Fear of Electoral Violence and its Impact on Political Knowledge in Sub-Saharan Africa

Abstract

The literature on electoral violence has focused on its causes as well as its scope, ignoring the implications for citizens trying to practice their political citizenship. Informed citizens are a central part of a functioning democracy. The emotive response to violence may play an important role here. This article contributes to a deeper understanding of how the voter responds to violence. Recent work on the role of emotions in politics has demonstrated its positive role for cognition, as fear can increase the individual’s propensity to amass information and reevaluate attitudes and behavior. This is tested in a hitherto unexamined context, namely, Sub-Saharan Africa, using the Afrobarometer survey (20 different countries). In general, political fear is a significant predictor of political knowledge, but in the opposite direction compared to the hypothesis. The affective intelligence hypothesis only receives partial support using this data, namely, in countries with high levels of political violence.

Keywords

political knowledge fear violence elections Sub-Saharan Africa

Introduction

The ability of citizens to participate in elections and make considered judgments is crucial for holding elected representatives accountable, but also in terms of forward-looking electoral decisions. Informed voters are central to the workings of democracy, as captured by the emphasis put on “enlightened understanding” by Dahl (1989). Recent research has demonstrated that uninformed voters make different evaluations of candidates and parties, and in the long run make different choices during elections as they have a harder time translating their preferences at the poll (Fowler and Margolis, 2014; Singh and Roy, 2014). This suggests that political knowledge is central to whether or not the outcome of an election reflects the will of the people. To put it bluntly, political knowledge matters for the realization of democracy. If political knowledge is unequally distributed between groups in society, and recent work suggests that there are systematic differences in the distribution of political knowledge (Fowler and Margolis, 2014: 109; see also Jennings, 1996: 229), this is worrisome. Political knowledge and cognitive capacities are important as a resource that enables political participation in general and matter for the functioning of democracy in several ways (Verba et al., 1995: 304–333; Knight and Johnson, 1997: 299).

This article argues that one context where these questions are brought to the fore is the context of violence. Not all groups in society are equally exposed to violence when it does occur, which could result in further unequal distribution of political knowledge if violence is found to impact levels of political knowledge. Homicide rates in Africa are more than double the global average and 36% of the homicides in the world took place in Africa in 2010 (United Nations Office on Drugs and Crime (UNODC), 2011: 11), and the highest global rate of major assault is also found in West, Central, and Southern Africa even if data are scarce (Harrendorf et al., 2010: 22). Particularly relevant to a voter should be violence occurring in relation to elections, or political violence more generally. The experience of electoral violence is widespread in Africa, and many elections are characterized by harassment, intimidation, destruction of property, assaults, deaths, and displacement especially in the period leading up to the election (Bekoe, 2012; Bratton, 2008; Dercon and Gutiérrez-Romero, 2012; Höglund, 2009; Inokoba and Maliki, 2011). Based on the African Electoral Violence Dataset (AEVD),¹ electoral violence is a regular occurrence for nearly half the countries included in the dataset, and about 20% of the elections resulted in severe levels of violence (Bekoe, 2012: 8; Straus and Taylor, 2012: 17). Research pertaining to electoral violence has not paid attention to how it impacts the voter and her ability to practice her citizenship, nor consequences in general of electoral violence (see, for example, Bekoe, 2012; Collier and Vicente, 2014; Dercon and Gutiérrez-Romero, 2012; Dunning, 2011; Fjelde and Höglund, 2014; Guelke, 2001; Hafner-Burton et al., 2014; Hickman, 2009; Höglund, 2009; Straus and Taylor, 2012; Wilkinson, 2004; note, however, one exception, Getmansky and Zeitzoff, 2014). This article attempts to rectify this, by examining whether violence can impact levels of political knowledge, in particular through the emotional response induced by political violence.

Recently, however, there has been an effort to understand how violence in general influences politics, both at the macro-level and at the micro-level. For instance, some works examine the history of war and its influence on the ensuing party system and electoral performance of former warring parties (see, for example, Ishiyama, 2014; Ishiyama and Widmeier, 2013; Levitsky and Way, 2012; Sindre, 2014). Vera Mironova and Sam Whitt (2016: 2) note that “researchers are increasingly examining the impact of violence on micro-level norms and preferences,” but that the impact of violence is mixed so far. Within this field, there is great variation in terms of what specific violence is studied, as well as what norms, preferences, or micro-level behavior are studied. For instance, a recent study finds that mass uprisings or colored revolutions lead to a decrease in trust (Ishiyama and Pechenina, 2016). In a study of Kenya, the authors examined whether violence hardens ethnic identities, but they were unable to demonstrate such effects, even when considering the frequency of the violence, its intensity, or the specific group involved in the act (Ishiyama et al., 2016). In contrast, Mironova and Whitt (2016) do find that cues related to in- and outgroup matter for how violence shapes social norms. What is clear is that the political consequences of violence are a new and expanding field of research, and that the consequences thus far are both varied and sometimes surprising.

In the search for the informed and rational voter, many have been concerned with the increased appeal to emotions in politics. Emotion has historically often been portrayed as the opposite of rationality, even its enemy. For a long time, the role of emotions in politics was largely ignored in terms of empirical scrutiny. However, recent work has now convincingly shown how emotion and rationality are in fact very complementary, and that emotional responses are central to increased attention and in the end political learning (Erisen et al., 2014; Marcus, 2003; Marcus et al., 2000; Marcus and Mackuen, 1993; Redlawsk, 2006; Valentino et al., 2008; see also Bickford, 2011, for a normative argument about the role of emotions for political judgment). As has been succinctly formulated: “experiencing strong emotions is not necessarily at odds with achieving the normative ideals of democratic citizenship” (Valentino et al., 2008: 249). This research has demonstrated a strong association between emotions (particularly fear) and political learning, based on both experimental and cross-sectional data. This research is however limited in that it has only been examined in relatively safe contexts (mainly the US), where the object that causes fear is fairly removed (for a notable exception, see Getmansky and Zeitzoff, 2014).² This spurs the question, what happens in an environment where the cause for fear is more proximate and real? Testing the external validity of these claims (moving from experiments to real world experiences) as well as examining the scope of this effect (moving from the US to Sub-Saharan Africa) more thoroughly are the next steps. In sum then, there are good reasons to believe that violent electoral landscapes impact levels of political knowledge in important ways. Crucially, we do not know whether electoral and political violence spurs vicious circles decreasing political knowledge, producing less accountability, less forward-looking choices, and ultimately deteriorating democracy. Or whether the experience of electoral and political violence has other side effects on the citizenry, improving political knowledge which in the end might improve democracy and counteract the violence.

The research question is therefore as follows: Does fear of electoral violence impact levels of political knowledge? In order to answer this question, this article uses Afrobarometer data from 2008/2009. After presenting the research hypotheses in full, the article details the data used and how the central variables are measured and coded. Proceeding to the analysis, this article contributes to an understanding of how the voter responds to violence through focusing on the emotive response to violence. The findings in this article concerning the impact of fear do not support the hypothesis. In fact, fear is statistically associated with a decrease in political knowledge, and the analysis does not yield evidence in support of the proposed mechanisms. However, in contexts of high levels of social conflict, fear increases political knowledge, that is, the impact is context dependent. This is particularly striking, as previous research has been cautious about predicting the impact of fear in contexts of more serious threats.

Violence and Political Knowledge

The centrality of informed citizens for the functioning of democracy has spurred a lot of research on political knowledge in particular. Research related to political knowledge has for instance been conducted to explain its occurrence (both in relation to individual and structural variables) (see, for example, Grönlund and Milner, 2006; Henderson, 2014; Jerit, 2009), on its distribution with respect to gender differences (Dolan, 2011; Dow, 2009; Pereira et al., 2015), on its impact on political participation and voter turnout in particular (Howe, 2006; Jung et al., 2011), and in relation to ideas about the lack of well-informed citizens (see, for example, Baron, 1994; Bartels, 1996; Bullock, 2011; Gilens, 2001; Norris, 2011: 142–168). Much of this literature, however, takes the division between the rational and irrational for granted, and leave out the role of emotions in shaping cognition and political knowledge.

In research attempting to understand the development of democracy and the nature of elections, there has been an increased focus on electoral violence. This literature has in general been concerned with the causes of electoral violence as well as its scope (see, for example, Bekoe, 2012; Collier and Vicente, 2014; Dercon and Gutiérrez-Romero, 2012; Dunning, 2011; Fjelde and Höglund, 2014; Guelke, 2001; Hafner-Burton et al., 2014; Hickman, 2009; Höglund, 2009; Straus and Taylor, 2012; Wilkinson, 2004). However, in trying to understand the conditions under which citizens today practice their political citizenship, we need to understand what violence does to the voter. Getmansky and Zeitzoff’s recent article is an example of work that tries to do that, by focusing on the experience of terrorism or the threat of terrorism. In their work, they show how political choices are influenced by the threat of violence, causing voters in Israel to vote for more right-wing and hardline parties, potentially inducing a vicious circle where conflict dynamics are reproduced (Getmansky and Zeitzoff, 2014). This article looks at a more limited part of that chain, focusing on the voter’s ability to make a reasoned choice rather than the ballot choice itself: the experience of violence and its impact on political knowledge.

Fear and its Cognitive Role

Our cognitive functions and ability to process information are central to our ability to practice our democratic citizenship. The question of what influences our cognitive abilities is a large field. If we execute our democratic citizenship in an environment of violence, what happens then? Based on current research on emotions, I argue that a violent environment triggers emotional responses, which in turn may trigger increased attention and learning.

Marcus, Neuman, and Mackuen suggest that certain emotional responses are likely to encourage attention and ability to process information, and have shown that political learning is linked to levels of anxiety (fear), where those voters who feel uneasy about the election also think more critically of the electoral choices before them (Marcus, 2003; Marcus et al., 2000; Marcus and Mackuen, 1993). Emotional responses are said to stimulate rational thought. Emotions are therefore not seen as an obstacle to reason, but an important element of thoughtful reflection, and indeed “emotion enhances the ability of voters to perform their citizenly duties” (Marcus and Mackuen, 1993: 681). The emotional response of fear acts as a cue to trigger attention and reflection. Anxiety, or fear, among citizens leads to an increase in attention, motivation to learn, and a reevaluation of their positions in relation to that which caused the anxiety (Marcus et al., 2000: 3f, 63, Marcus and Mackuen, 1993: 681). The importance of anxiety and experiencing threat in politics, on both behavioral choices and opinions, is well-established, in particular the role anxiety plays for information seeking and political learning (see, for example, Huddy et al., 2005, 2007; Hunter et al., 2012: 4; Taber, 2003: 461; Valentino et al., 2008, 2011; see also Erisen et al., 2014).

However, there are still questions as to the impact of serious threat (to life and self-identity) and the emotional response related to that (Marcus et al., 2000: 95). For instance, Nadeau et al. (1995) suggest with their work that while permanent threat may induce anxiety, it also leads to hopelessness, which undercuts motivations for political learning. This article tries to gauge just how fear in the face of more serious threats impacts the voter, thereby testing the scope of this theory.

Thus, if the violence we experience and see in our community is of a political nature, the fear resulting from such violence may directly impact our attention to the origins of that violence: politics. Given the recent findings concerning the impact of fear or anxiety in politics and on political learning in general, we should expect that fear associated with politics should increase attention to politics and thereby increase levels of political knowledge. The following hypothesis is formulated:

Hypothesis 1. Political fear increases people’s attention to politics, resulting in higher political knowledge.

We should therefore also observe that the attention to politics should increase as a result of political fear. So interest in politics should increase as should the individual’s time devoted to discussing politics. These mechanisms are also tested in the analysis.

Finally, this literature as a whole also motivates a further exploration of whether specific context traits matter for the impact of political fear on political knowledge. Environments that are more volatile may be more likely to give rise to the relation expected between political fear and political knowledge. In a context where electoral and political violence is more common, the experience of political fear is likely to be more credible, that is, the respondent has good grounds to be afraid and the emotional and cognitive response in such environments may differ from more calm and stable contexts. To some degree, one could think of this as a priming effect, where increased exposure to something makes it easier for memory retrieval and the cognitive response to an event kicks in faster and more strongly if you have been exposed to it in the past. This is sometimes discussed as the salience and accessibility mechanism behind priming effects (see, for example, Shehata and Falasca, 2014):

Hypothesis 2. High occurrence of political violence is likely to prime individuals and heighten the impact of political fear on political knowledge.

Data and Method

The data used are the Afrobarometer survey, Round 4, which covered 20 countries,³ as this round included questions that can be used to measure political knowledge. All surveys were conducted in 2008, except in Zambia and Zimbabwe where they were conducted in 2009. In total, the round includes 27,713 respondents. The analysis uses across-country weights to adjust for the sampling procedures used by the Afrobarometer, and multi-level modeling as the data are hierarchically structured.

Dependent Variable: Political Knowledge

The dependent variable political knowledge is measured using the two political knowledge questions that were asked in Round 4 of the Afrobarometer. The dependent variable political knowledge is measured through the respondent’s ability to identify (name) their Member of Parliament and the Minister of Finance correctly. This means that the variable focuses on what is called the veracious dimension of political knowledge, that is, factually correct knowledge, and in particular the type of knowledge that is usually termed “surveillance facts” rather than “textbook facts” or static facts (Barabas et al., 2014; Jennings, 1996: 229). Factually correct knowledge can belong to one of four types, defined by their temporal dimension (static or surveillance) and their topical dimension (general or policy-specific) (Barabas et al., 2014: 841). The surveillance facts tapped in this article relate to the individual’s capacity to retain information regarding new events, rather than information concerning the functioning of government as detailed in the constitution for instance (temporally invariant). However, it does not capture what Barabas et al. call policy-relevant knowledge, but rather general surveillance facts. Policy relevant surveillance knowledge should in general be even harder to obtain than surveillance knowledge about institutions and people (Barabas et al., 2014: 841f).

The variable is coded as the percent of questions the respondent answered correctly (percent correct). This results in a dependent variable with three possible values, neither correctly identified (0), one correctly identified (50), and two correctly identified individuals (100). Importantly, since the variable does not differentiate between which questions were answered correctly, the distance between knowing nothing to knowing one answer and between knowing one answer and two answers is deemed the same.⁴ A linear regression model is deemed appropriate for the analysis, as this is also simpler to use and interpret (many have made calls for more parsimonious and simple models; see, for example, Achen, 2002; Schrodt, 2014: 292f).⁵ Using the entire sample, as many as 45.17% could not identify either individual, and 41.47% could identify one, and only 13.36% of the sample could correctly identify both.

In order to gauge this variable’s validity, it was correlated with a number of items. First, it is positively correlated with education (0.295***) and voting (0.081***), variables typically positively correlated with political knowledge. In addition, percent correct was also negatively correlated with an index variable (−0.115***), a variable which measures whether or not the respondent had heard enough about a number of political actors and institutions to answer a question about trust. The index variable was created through adding up the times the respondent used the response category “Haven’t heard enough to say” when asked to rate their trust for a number of political actors/institutions: president, national assembly, national election commission, elected local council, ruling party, opposition, police, and courts of law.⁶ The negative correlation between percent correct and have not heard about is as expected, as the higher the respondent scored on the not heard about index, the fewer political actors the respondent felt they knew enough about to comment, and this in turn corresponds to rating low on the percent correct variable. Percent correct is a better measure of political knowledge, but the subjective rating of the respondent whether they know enough about certain actors to comment on them further is an indication of the degree to which the respondent feels familiar and knowledgeable about these actors and institutions (thus to some degree tapping internal efficacy).

The measure used here for political knowledge is limited, and the needs of the voter in order to make an informed choice are much larger than what is captured by general surveillance facts. What kind of knowledge a voter should primarily have is of course a much larger normative and empirical question (see, for example, Dolan, 2011, for a discussion on measures that may lead to a gender bias, and Luskin and Bullock, 2011, for a discussion of how response categories are valued). Yet, name recognition and attention to new events in politics are useful and in no way unimportant facets of the multidimensional concept of political knowledge. A larger problem may be that these questions focus on the more formal aspects of politics, as they pertain to elected representatives rather than local big men which might be even more important local political knowledge in many contexts (see, for instance, Chabal and Daloz, 1999; Utas, 2012). Nonetheless, formal power is not irrelevant in these contexts either, and as one of the questions focus on the Member of Parliament representing their district the data used certainly has a local focus as well.

Fear of Political Violence and Intimidation

The main independent variable is fear of political violence and intimidation. In addition, potential intervening variables are also specified below. Fear is measured with the help of the survey item “During election campaigns in this country, how much do you personally fear becoming a victim of political intimidation or violence?” (four possible answer categories: not at all, a little bit, somewhat, and a lot, 0–3) (Table 1).

Table 1.

Distribution of Fear Political Intimidation or Violence.

	Percent
Not at all	46.34
A little bit	20.55
Somewhat	14.54
A lot	18.57
Total	100.00

Estimated using sampling weights. Total number of observations 26,959.

The underlying assumption of the fear hypothesis is that it should increase our attention to politics specifically, which in turn leads to higher levels of political knowledge. However, an experimental study focused on immigration noted that while fear may trigger attention and information seeking, it may not lead to better information processing (Gadarian and Albertson, 2014: 133). Hence, fear may simply lead to more attention to politics without better knowledge. Or indeed, we see both steps: first increased attention followed by increased knowledge. This is tested using variables measuring political interest (How interested would you say you are in public affairs? Not at all interested, Not very interested, Somewhat interested, and Very interested, 0–3) as well as tendency to discuss politics (When you get together with your friends or family, would you say you discuss political matters: Never, Occasionally, and Frequently, 0–2).

In contrast with previous research on the role of emotion for information seeking, this article is not based on experimental data. This means of course that endogeneity cannot be avoided; it is possible that strong emotional reactions to politics are the result of a tendency to seek out information about politics, that is, the more well-informed are more afraid of politics due to them knowing more about politics. However, experimental data elsewhere have shown that causality runs mainly in the direction discussed in this article and not the other way around. This problem can, however, not be ruled out entirely.

In order to test the priming hypothesis, the fear variable is interacted with two different measures of political violence, as fear may have been primed in such contexts. I suggest that this may be the case in regions with high and common occurrences of political violence. Using the Social Conflict in Africa Database (SCAD), there are two measures of degree of violence and conflict: deaths and events (Salehyan et al., 2012; Salehyan and Linebarger, 2013). This dataset contains information about protests, riots, strikes, and other social disturbances, for 19 of the 20 countries in the Afrobarometer data (Cape Verde not included). The data thus measure political violence more broadly, and are not limited to election violence (for a discussion about these distinctions, see, for example, Höglund, 2009; Ishiyama et al., 2016: 305). Events records the total number of events ongoing during the year of the survey in each region, whereas deaths records the total number of persons killed at such events in each region.

Control Variables

In addition, a number of control variables were added to the model, particularly as several of these variables are likely to influence political knowledge, but also be associated with exposure to violence. Not including these control variables would then bias the coefficient estimates.

Voters who identify with the incumbent party should be less likely to be exposed to fear and intimidation, yet more likely to be knowledgeable about those elected. Therefore, the models also control for whether the respondent voted for the incumbent or not with a dummy variable, coded one if the respondent answered that they would vote for the incumbent party, zero otherwise.⁷ Another group that may be targeted or feel more at risk is ethnic minorities. Using the Afrobarometer survey, one question asks whether the respondent feels that their ethnic group (if they have identified with one) is treated unfairly by the government. The variable consists of four categories, from never, sometimes, often, to always, 0–3.

Countries that have just gone through an election are different in two important ways. First, political discussions and debates are likely to have been more common recently, and as a result, people may be more likely to know more about politics, but as any new minister or Member of Parliament was recently announced as such, people may not have learned about this. Second, the fear of electoral and political violence is reasonably different during different parts of the electoral cycle. Therefore, the analysis identified countries that, during the time of the survey was collected, had gone through an election the year before, that is, post-election countries. Only Ghana had an election the same year, but several months after the survey was conducted. Nine countries were identified as post-election countries (1 year): Benin, Burkina Faso, Kenya, Lesotho, Madagascar, Mali, Nigeria, Senegal, and Zambia.

In addition, a number of control variables are used as they are generally recognized as important predictors of political knowledge (Grönlund and Milner, 2006; Henderson, 2014; Jerit, 2009). The respondent’s age (continuous scale, in years), level of education (10 categories, from no formal schooling to post-graduate), gender (male = 1), political interest (“How interested would you say you are in public affairs?” four categories, from not at all interested to very interested, 0–3), and access to news were therefore included in the model. In order to measure the respondent’s access to news, items indicating access to news through radio is used. Access to news through radio should be the most important, as radio should be the most accessible channel and not dependent on the respondent’s literacy.⁸ The variable is measured on a 5-category scale (“How often do you get news from the radio?” from never, less than once a month, a few times a month, a few times a week, to every day, 0–4). Poverty or low income should undermine people’s attention to political events and thus result in less political knowledge. Lack of resources is measured through how often the respondent has gone without food and cash income. The survey questions related to these resources were measured as follows: “Over the past year, how often, if ever, have you or anyone in your family gone without: a cash income/enough food to eat.” The variables were measured on a 5-category scale (never, just once or twice, several times, many times, and always, 0–4). For more details on the variables used in the analysis, see Table A1 of Appendix 1, with descriptive statistics on all variables.

Analysis

The analysis proceeds in several steps; first, an examination of the relationship between fear and the dependent variable political knowledge, followed by an examination of the intervening variables (Tables 2 and 3). In Table 2, the dependent variable is percent correct (the results are robust to alternative model specifications).⁹ Model 1 is simply a bivariate model of fear and political knowledge, whereas Model 2 includes all individual level control variables. The fear variable is reported as significant and negative in both models, if somewhat attenuated when the controls are introduced. The results remain the same when adding other control variables such as other sources of news, urban/rural, and when dropping political interest, ethnic group treated unfairly (the significance level drops to 90%), and voting for incumbent. The next two models add the region-level control variables, and again the fear variable is reported as significant and negative.¹⁰ Model 3 uses the number of events of social conflict as a measure of political violence, whereas Model 4 uses the number of deaths due to social conflict as a measure of political violence. While the fear variable is reported as statistically significant in all of these models, it is consistently reported as negative (the more fearful, the less knowledgeable), that is, in the opposite direction of what the hypothesis stated (even if the magnitude of the coefficient is not large). Interestingly, the number of events of social conflict has a negative and significant relationship with political knowledge, whereas the number of deaths is positively associated (however when Kenya is dropped as a case, the variable deaths is reported as negative and significant).

Table 2.

Political Knowledge, Percent Correct (0–100).

	Bivariate	All individual level controls	All individual, region and country level controls (events and deaths, respectively)		Interaction with political violence (events and deaths, respectively)
	Model 1	Model 2	Model 3	Model 4	Model 5		Model 6
Individual-level variables
Fear of intimidation or violence	−0.583** (0.220)	−0.438* (0.209)	−0.446* (0.215)	−0.507* (0.252)	−0.730** (0.248)		−0.485** (0.222)
Food shortage		−0.900*** (0.205)	−0.920*** (0.210)	−0.926*** (0.285)	−0.917*** (0.210)		−0.926*** (0.209)
Lack of cash income		−0.507* (0.244)	−0.431 (0.253)	−0.404 (0.411)	−0.436 (0.253)		−0.402 (0.252)
Age		0.212*** (0.020)	0.210*** (0.021)	0.210*** (0.028)	0.210*** (0.021)		0.210*** (0.021)
Education		4.693*** (0.192)	4.671*** (0.197)	4.627*** (0.411)	4.672*** (0.197)		4.626*** (0.197)
Male		6.557*** (0.516)	6.735*** (0.535)	6.756*** (1.086)	6.744*** (0.535)		6.756*** (0.536)
Political interest		1.967*** (0.254)	1.839*** (0.259)	1.852*** (0.343)	1.843*** (0.259)		1.852*** (0.259)
Access to radio news		2.019*** (0.183)	1.982*** (0.184)	1.969*** (0.260)	1.980*** (0.184)		1.970*** (0.184)
Incumbent voter		−0.328 (0.499)	−0.829 (0.487)	−0.840 (0.846)	−0.807 (0.487)		−0.845 (0.490)
Ethnic group treated unfairly		−0.144 (0.161)	−0.078 (0.190)	−0.123 (0.237)	−0.073 (0.190)		−0.123 (0.189)
Intercept	33.739*** (1.165)	0.906 (1.739)	9.257*** (2.030)	8.160*** (4.472)	9.501*** (2.025)		8.135*** (1.959)
Cross-level interactions
Fear × deaths							0.000	(0.001)
Fear × events					0.162*	(0.074)
Region-level variables
Deaths due to social conflict				0.023*** (0.009)			0.023*** (0.002)
Events of social conflict			−1.585*** (0.455)		−1.793*** (0.473)
Post-election country			−9.257*** (2.030)	−14.934*** (5.329)	−9.839*** (1.759)		−14.938*** (1.897)
Variance (intercept)	291.856	254.562	205.006	188.965	203.546		189.070
Variance (residuals)	932.934	819.550	818.403	818.537	818.305		818.519
N (respondents)	26,911	24,115	22,982	22,982	22,982		22,982
N (regions)	246	246	241	241	241		241
AIC	226,453.9	199,982	189,620.5	189,605.7	189,618.6		189,607.4
Log pseudolikelihood	−113,222	−99,978.011	−94,795.241	−94,787.83	−94,793.276		−94,787.678
Wald chi-square	7.02	1018.52	1050.38	1322.16	1054.78		1322.89
Prob. > chi-square	0.008	0.000	0.000	0.000	0.000		0.000

AIC: Akaike information criterion.

Across-country weights are used in all models. There is no data on social conflict (deaths or events) for Cape Verde, hence the total number of observations drop considerably when these variables are used. Standard errors within parentheses.

Significance levels: *p < 0.05; **p < 0.01; ***p < 0.001.

Table 3.

Predicted Values for Percent Correct, Based on Model 5.

	Fear = min.	Fear = max.	Δ Percent correct
At no events	40.75	38.56	−2.19
At 5 events	31.80	32.04	+0.24
At 10 events	22.85	25.52	+2.67
At 15 events	13.90	19.00	+5.10
Δ Percent correct	−26.85	−19.56

All other variables set at their mean (food shortage, lack of cash income, age, education, access to radio news, ethnic group treated unfairly) or modal value (male, political interest, incumbent voter, post-election country).

However, perhaps the impact of fear is primed by the level of violence in a country. The analysis therefore proceeds to testing the idea that in some areas the effect of fear may be dependent on priming, following Hypothesis 2. This is done in Models 5 and 6, where fear is interacted with the two measures of political violence, respectively. In both models, the fear variable is reported as significant, albeit still negatively associated with political knowledge. The coefficients for events and deaths are also similar to the estimations without the interaction. The interaction itself with events is significant and positive, whereas the interaction with number of deaths is not significant. This means that the overall impact of fear is partially dependent on the level of political violence, in terms of events of social conflict. In fact, in regions with a high frequency of social conflict events, the overall impact of fear changes sign. This can be seen in Table 3, where the predicted values of percent correct are shown for various levels of fear and events, when the other control variables are set at their means or modal values.

As we see, the impact changes from being negative to positive. Comparing a region with no events of social conflict with a region with 15 events, Model 5 predicts that the impact of fear, as we move from the minimum value of fear (0) to its maximum value (3), changes. In the first case, their political knowledge score is likely to be 2.19 percentage points lower, whereas with many events of social conflict, the expected difference is an increase with 5.10 percentage points.

Thus, in countries with a high occurrence of political violence,¹¹ that is, where fear is more associated with actual risk, there is some indication that the variable has a positive impact on political knowledge. What is clear from this analysis is that the impact of political fear on political knowledge is not necessarily identical across cases, countries, and with respect to the traits of the violence, but can vary, suggesting that further case studies would be useful.

Considering the proposed mechanism, if we turn to examine potential intervening variables, we do not see evidence in support of the hypothesis either (Table 4). The impact of fear is not significant for political interest or the tendency to discuss politics (Model 7 and 8). Interacting fear with measures of political violence and social conflict does not change the result.

Table 4.

Intervening Variables for Fear.

	Political interest (0–3)	Discuss politics (0–2)
	Model 7	Model 8
Individual-level variables
Fear of intimidation or violence	−0.010 (0.010)	0.001 (0.005)
Food shortage	−0.002 (0.009)	0.005 (0.005)
Lack of cash income	0.020* (0.009)	−0.001 (0.004)
Age	0.006*** (0.001)	−0.000 (0.000)
Education	0.055*** (0.006)	0.045*** (0.004)
Male	0.242*** (0.017)	0.110*** (0.010)
Political Interest		0.298*** (0.009)
Access to radio news	0.102*** (0.008)	0.032*** (0.004)
Incumbent voter	0.155*** (0.019)	0.031** (0.011)
Ethnic group treated unfairly	−0.009 (0.008)	0.004 (0.004)
Intercept	0.967*** (0.055)	0.051 (0.028)
Region level variables
Events of social conflict	−0.030*** (0.008)	0.011* (0.005)
Post-election country	0.123** (0.040)	0.022 (0.025)
Variance (intercept)	0.072	0.025
Variance (residuals)	1.059	0.368
N (respondents)	23,008	22,904
N (regions)	241	241
AIC	57,804.09	36,733.24
Log pseudolikelihood	−28,888.043	−18,351.621
Wald chi-square	902.33	3450.14
Prob. > chi-square	0.0000	0.0000

AIC: Akaike information criterion.

Across-country weights are used in all models. Standard errors within parentheses.

Significance levels: *p < 0.05; **p < 0.01; ***p < 0.001.

In general, political fear is a significant predictor of political knowledge, but in the opposite direction compared to the hypothesis. In contrast with previous work on fear and emotive responses, survey data from Sub-Saharan Africa do not indicate the same cognitive benefits. The larger tendency here is that political violence (number of events) and the fear of political violence reduce political knowledge. The affective intelligence hypothesis does not explain the observed negative relation between fear and political knowledge. This suggests that how political violence and political knowledge are associated need to be studied further, in order to uncover both the underlying mechanisms as well as whether there are specific traits of the violence that matter in one way rather than another. Previous work has explored such aspects as the longevity of the violence, as well as its intensity, origin, and target (Ishiyama et al., 2016). The scope of this article does not allow for an investigation of these issues, but these are promising avenues. As some contextual differences also were discovered, it would be appropriate to further scrutinize the relationship between fear of electoral violence with case studies, where the data may also allow the analysis to further unravel the mechanisms at play.

Conclusion

While much research is conducted on the extent and scope of electoral violence and explanations for electoral violence, less work has been conducted on how the individual voter is affected when trying to exercise their democratic citizenship in violent political landscapes. To scrutinize how electoral and political violence impacts the voter, available data from the Afrobarometer were used to test how political knowledge is affected by such experiences. In this article, the effect on political knowledge resulting from fearing political violence and intimidation has been tested. Fear of political violence was hypothesized to heighten levels of political knowledge following the affective intelligence hypothesis, but the results show that fear actually decreases political knowledge. What mechanisms drive this negative association between fear of violence and political knowledge remain to be determined. Fear of violence thus seems to overall impede the voter in practicing their citizenship; making informed choices at the ballot box is more difficult in the midst of violence. Yet, fear in the context of numerous events of political violence and social conflict increases political knowledge.

This article has contributed to understanding the scope and relevance of the affective intelligence hypothesis, particularly through using non-experimental data from Sub-Saharan Africa. And these data suggest that its scope cannot be extended to election and political violence in Sub-Saharan Africa. Only partially is the hypothesis relevant, namely, in areas with a lot of political violence. Here, the respondent may experience a priming effect, where the previous experience of violence makes the individual more prone to feel fear as well as that fear having a larger cognitive importance. In such cases, political fear has a positive impact on political knowledge.

One of the main limitations of this article is how variable measures are limited to the format of the survey, particularly in terms of the measure of political knowledge (but also in terms of measuring other emotional responses to violence). As the concept of political knowledge is multidimensional, the trends noted here may play out differently had other dimensions of political knowledge been tested. What kind of political knowledge is most central to the African voter, particularly in contexts of electoral violence, is also an open question. Name recognition of elected representatives is not the only relevant form of knowledge. We need to test these relationships with other forms of measure of political knowledge. Other extensions were also not possible due to data limitations, as differentiating the kind of violence that inspires fear further (such as its intensity, longevity, targets, and origins). In addition, whether actual and direct exposure to political violence has the same impact on political knowledge as the fear of violence has also requires further scrutiny.

Despite these limitations, this article offers an expansion and deeper understanding of how violence shapes the ability of citizens to make informed choices in politics. Importantly, this article demonstrates that fear of electoral violence spurs vicious circles, where political knowledge decreases, which in turn may produce less accountability, and less forward-looking choices among the electorate. Similarly, the experience of political violence in the region of the respondent is associated with lower levels of political knowledge. Whether or not electoral and political violence (and individual citizens’ different experiences and reactions to such violence) overall decreases levels of political knowledge and undermines the political process and decreases the degree to which political choices reflect the will of the people requires more research.

Footnotes

Appendix 1

Table A1.

Descriptive Statistics.

	Mean	Mode	Standard error	Min.	Max.	Observations
Individual-level variables
Access to radio news	2.973	4	0.090	0	4	27,669
Age	36.586	30	0.609	18	110	27,380
Discuss politics	0.916	1	0.033	0	2	27,468
Education	3.045	4	0.192	0	9	27,669
Ethnic group treated unfairly	1.090	0	0.208	0	7	25,320
Fear intimidation or violence	1.053	0	0.109	0	3	26,959
Food shortage	1.186	0	0.088	0	4	27,651
Incumbent voter	0.411	0	0.043	0	1	27,713
Lack of cash income	2.080	3	0.126	0	4	27,551
Male	0.497	0	0.002	0	1	27,713
Not heard about political actors index	0.492	0	0.068	0	8	27,713
Political Interest	1.811	3	0.048	0	3	27,439
Percent correctly answered	34.092	0	3.552	0	100	27,663
Region level variables
Events of social conflict	1.82	0	2.46	0	15	241
Deaths due to social conflict	36.52	0	221.66	0	1572	241
Country level variables
Events of social conflict	6.47	1	2.02	0	30	19
Deaths due to social conflict	136.78	0	93.91	0	1629	19
Post-election country	0.45	0	0.11	0	1	20

Acknowledgements

I would like to thank Enzo Nussio, Markus Holdo, Kristine Höglund, Katrin Uba, Julia Jennstål, Gina Gustavsson, Ben Oppenheim, the anonymous reviewers and the editors for their helpful comments on earlier versions of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research for this article was partially funded by the Swedish Research Council [grant number 421-2011-1438].

Notes

ORCID iD

Johanna Söderström

Author Biography

Johanna Söderström (PhD 2011) is a EURIAS Research Fellow at the Netherlands Institute for Advanced Study in the Humanities and the Social Sciences (NIAS), Associate Professor at the Department of Comparative Politics, University of Bergen, Researcher at the Department of Government, Uppsala University, and member of the research program Varieties of Peace at Umeå University. Her work is concerned with the intersection between democratization, peacebuilding and statebuilding. Her most recent book was entitled Peacebuilding and Ex-Combatants: Political Reintegration in Liberia (2015), and she recently edited a special issue on “Understanding Armed Groups and Party Politics” (in Civil Wars, 2016).

References

Achen

(2002) Toward a New Political Methodology: Microfoundations and ART. Annual Review of Political Science 5: 423–450.

Ariffin

Coicaud

J-M

Popovski

(2016) Emotions in International Politics: Beyond Mainstream International Relations. Cambridge: Cambridge University Press.

Barabas

Jerit

Pollock

et al . (2014) The Question(s) of Political Knowledge. American Political Science Review 108 (4): 840–855.

Baron

(1994) Electoral Competition with Informed and Uniformed Voters. American Political Science Review 88 (1): 33–47.

Bartels

(1996) Uninformed Votes: Information Effects in Presidential Elections. American Journal of Political Science 40 (1): 194–230.

Bekoe

(ed.) (2012) Voting in Fear: Electoral Violence in Sub-Saharan Africa. Washington, DC: United States Institute of Peace Press.

Bickford

(2011) Emotion Talk and Political Judgment. Journal of Politics 73 (4): 1025–1037.

Bleiker

Hutchison

(2014) Introduction: Emotions and World Politics. International Theory 6 (3): 490–491.

Branton

Cassese

Jones

et al . (2011) All Along the Watchtower: Acculturation Fear, Anti-Latino Affect, and Immigration. Journal of Politics 73 (3): 664–679.

10.

Bratton

(2008) Vote Buying and Violence in Nigerian Election Campaigns. Electoral Studies 27 (4): 621–632.

11.

Bullock

(2011) Elite Influence on Public Opinion in an Informed Electorate. American Political Science Review 105 (3): 496–515.

12.

Chabal

Daloz

J-P

(1999) Africa Works: Disorder as Political Instrument. Oxford: International African Institute, James Currey and Indiana University Press.

13.

Collier

Vicente

(2014) Votes and Violence: Evidence from a Field Experiment in Nigeria. The Economic Journal 124 (574): F327–1F55.

14.

Costalli

Ruggeri

(2017) Politics Symposium: Emotions, Ideologies, and Violent Political Mobilization. PS: Political Science & Politics 50 (4): 923–927.

15.

Dahl

(1989) Democracy and Its Critics. New Haven, CT and London: Yale University Press.

16.

Dercon

Gutiérrez-Romero

(2012) Triggers and Characteristics of the 2007 Kenyan Electoral Violence. World Development 40 (4): 731–744.

17.

Dolan

(2011) Do Women and Men Know Different Things? Measuring Gender Differences in Political Knowledge. Journal of Politics 73 (1): 97–107.

18.

Dow

(2009) Gender Differences in Political Knowledge: Distinguishing Characteristics-Based and Returns-Based Differences. Political Behavior 31 (1): 117–136.

19.

Dunning

(2011) Fighting and Voting: Violent Conflict and Electoral Politics. Journal of Conflict Resolution 55(3): 327–339.

20.

Erikson

Stoker

(2011) Caught in the Draft: The Effects of Vietnam Draft Lottery Status on Political Attitudes. American Political Science Review 105 (2): 221–237.

21.

Erisen

Lodge

Taber

(2014) Affective Contagion in Effortful Political Thinking. Political Psychology 35(2): 187–206.

22.

Feldman

Stenner

(1997) Perceived threat and authoritarianism. Political Psychology 18(4): 741–770.

23.

Fierke

(2013) Political Self-sacrifice: Agency, Body and Emotion in International Relations. Cambridge: Cambridge University Press.

24.

Fjelde

Höglund

(2014) Electoral Institutions and Electoral Violence in Sub-Saharan Africa. British Journal of Political Science 46 (2): 297–320.

25.

Fowler

Margolis

(2014) The Political Consequences of Uninformed Voters. Electoral Studies 34: 100–110.

26.

Gadarian

Albertson

(2014) Anxiety, Immigration, and the Search for Information. Political Psychology 35 (2): 133–164.

27.

Getmansky

Zeitzoff

(2014) Terrorism and Voting: The Effect of Rocket Threat on Voting in Israeli Elections. American Political Science Review 108 (3): 588–604.

28.

Gilens

(2001) Political Ignorance and Collective Policy Preferences. American Political Science Review 95 (2): 379–396.

29.

Groenendyk

Banks

(2014) Emotional Rescue: How Affect Helps Partisans Overcome Collective Action Problems. Political Psychology 35 (3): 359–378.

30.

Grönlund

Milner

(2006) The Determinants of Political Knowledge in Comparative Perspective. Scandinavian Political Studies 29 (4): 386–406.

31.

Guelke

(2001) Violence and Electoral Polarization in Divided Societies: Three Cases in a Comparative Perspective. In: Rapoport

Weinberg

(eds) The Democratic Experience and Political Violence. London: Frank Cass, pp.78–105.

32.

Hafner-Burton

Hyde

Jablonski

(2014) When Do Governments Resort to Election Violence? British Journal of Political Science 44 (1): 149–179.

33.

Harrendorf

Heiskanen

Malby

(2010) International Statistics on Crime and Justice. Helsinki: European Institute for Crime Prevention and Control, and United Nations Office on Drugs and Crime (UNODC).

34.

Heller

(2004) The Ministry of Fear. Social Research 71 (4): 849–862.

35.

Henderson

(2014) Issue Publics, Campaigns, and Political Knowledge. Political Behavior 36 (3): 631–657.

36.

Hickman

(2009) Is electoral violence effective? Evidence from Sri Lanka’s 2005 presidential election. Contemporary South Asia 17 (4): 429–435.

37.

Höglund

(2009) Electoral Violence in Conflict-Ridden Societies: Concepts, Causes, and Consequences. Terrorism and Political Violence 21 (3): 412–427.

38.

Howe

(2006) Political Knowledge and Electoral Participation in the Netherlands: Comparisons with the Canadian Case. International Political Science Review 27 (2): 137–166.

39.

Huddy

(2004) Fear and How It works: Science and the Social Sciences. Social Research 71 (4): 801–805.

40.

Huddy

Feldman

Weber

(2007) The Political Consequences of Perceived Threat and Felt Insecurity. The ANNALS of the American Academy of Political and Social Science 614 (1): 131–153.

41.

Huddy

Feldman

Taber

et al . (2005) Threat, Anxiety, and Support of Antiterrorism Policies. American Journal of Political Science 49 (3): 593–608.

42.

Hunter

Murray

Robbins

(2012) Political Violence and Voter Turnout. Chicago, IL: MPSA.

43.

Inokoba

Maliki

(2011) Youths, Electoral Violence and Democratic Consolidation in Nigeria: The Bayelsa State Experience. Anthropologist 13 (3): 217–225.

44.

Ishiyama

(2014) Civil Wars and Party Systems. Social Science Quarterly 95 (2): 425–447.

45.

Ishiyama

Pechenina

(2016) Colored Revolutions, Interpersonal Trust, and Confidence in Institutions: The Consequences of Mass Uprisings. Social Science Quarterly 97 (3): 748–770.

46.

Ishiyama

Widmeier

(2013) Territorial Control, Levels of Violence, and the Electoral Performance of Former Rebel Political Parties After Civil Wars. Civil Wars 15 (4): 531–550.

47.

Ishiyama

Gomez

Stewart

(2016) Does Conflict Lead to Ethnic Particularism? Electoral Violence and Ethnicity in Kenya 2005–2008. Nationalism and Ethnic Politics 22 (3): 300–321.

48.

Jennings

(1996) Political Knowledge over Time and Across Generations. Public Opinion Quarterly 60 (2): 228–252.

49.

Jerit

(2009) Understanding the Knowledge Gap: The Role of Experts and Journalists. Journal of Politics 71 (2): 442–456.

50.

Jost

Glaser

Kruglanski

et al . (2003) Political Conservatism as Motivated Social Cognition. Psychological Bulletin 129 (3): 339–375.

51.

Jung

Kim

de Zúñiga

(2011) The Mediating Role of Knowledge and Efficacy in the Effects of Communication on Political Participation. Mass Communication and Society 14 (4): 407–430.

52.

Knight

Johnson

(1997) What Sort of Political Equality Does Deliberative Democracy Require? In: Rehg

Bohman

(eds) Deliberative Democracy: Essays on Reason and Politics. Cambridge, MA and London: The MIT Press, pp.279–319.

53.

Levitsky

Way

(2012) Beyond Patronage: Violent Struggle, Ruling Party Cohesion, and Authoritarian Durability. Perspectives on Politics 10 (4): 869–889.

54.

Lupia

Menning

(2009) When Can Politicians Scare Citizens Into Supporting Bad Policies? American Journal of Political Science 53 (1): 90–106.

55.

Luskin

Bullock

(2011) “Don’t Know” Means “Don’t Know”: DK Responses and the Public’s Level of Political Knowledge. Journal of Politics 73 (2): 547–557.

56.

Marcus

(2003) The Psychology of Emotion and Politics. In: Sears

Huddy

Jervis

(eds) Oxford Handbook of Political Psychology. New York: Oxford University Press, pp.182–221.

57.

Marcus

Mackuen

(1993) Anxiety, Enthusiasm, and the Vote: The Emotional Underpinnings of Learning and Involvement During Presidential Campaigns. The American Political Science Review 87 (3): 672–685.

58.

Marcus

Neuman

MacKuen

(2000) Affective Intelligence and Political Judgment. Chicago, IL: University of Chicago Press.

59.

Mironova

Whitt

(2016) Social Norms after Conflict Exposure and Victimization by Violence: Experimental Evidence from Kosovo. British Journal of Political Science. Epub ahead of print 26 April. DOI: 10.1017/S0007123416000028.

60.

Mouffe

(2000) Politics and Passions: the Stakes of Democracy. Ethical Perspectives 7 (2–3): 146–150.

61.

Nadeau

Niemi

Amato

(1995) Emotions, Issue Importance, and Political Learning. American Journal of Political Science 39 (3): 558–574.

62.

Norris

(2011) Democratic Deficit: Critical Citizens Revisited. New York: Cambridge University Press.

63.

Nussio

(2012) Emotional Legacies of War Among Former Colombian Paramilitaries. Peace and Conflict: Journal of Peace Psychology 18 (4): 369–383.

64.

Ottati

Steenbergen

Riggle

(1992) The Cognitive and Affective Components of Political Attitudes: Measuring the Determinants of Candidate Evaluations. Political Behavior 14 (4): 423–442.

65.

Pereira

Fraile

Rubal

(2015) Young and Gapped? Political Knowledge of Girls and Boys in Europe. Political Research Quarterly 68 (1): 63–76.

66.

Pyszczynski

(2004) What are we so afraid of? A terror management theory perspective on the politics of fear. Social Research 71 (4): 827–848.

67.

Redlawsk

(2006) Feeling Politics: Emotion in Political Information Processing. New York: Palgrave Macmillan.

68.

Rudolph

Gangl

Stevens

(2000) The Effects of Efficacy and Emotions on Campaign Involvement. Journal of Politics 62 (4): 1189–1197.

69.

Salehyan

Linebarger

(2013) Elections and Social Conflict in Africa. Available at: https://www.files.ethz.ch/isn/161758/CCAPS%20Research%20Brief%20No.%206_final.pdf (accessed 24 February 2017).

70.

Salehyan

Hendrix

Hamner

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

71.

Schrodt

(2014) Seven Deadly Sins of Contemporary Quantitative Political Analysis. Journal of Peace Research 51 (2): 287–300.

72.

Shehata

Falasca

(2014) Priming Effects During the Financial Crisis: Accessibility and Applicability Mechanism behind Government Approval. European Political Science Review 6 (4): 597–620.

73.

Sindre

(2014) Internal Party Democracy in Former Rebel Parties. Party Politics 22: 501–511.

74.

Singh

Roy

(2014) Political Knowledge, the Decision Calculus, and Proximity Voting. Electoral Studies 34: 89–99.

75.

Söderström

(2015) Peacebuilding and Ex-combatants: Political Reintegration in Liberia. London and New York: Routledge.

76.

Straus

Taylor

(2012) Democratization and Electoral Violence in Sub-Saharan Africa, 1990-2008. In: Bekoe

(ed.) Voting in Fear: Electoral Violence in Sub-Saharan Africa. Washington, DC: United States Institute of Peace Press, pp.15–38.

77.

Summers-Effler

(2002) The Micro Potential for Social Change: Emotion, Consciousness, and Social Movement Formation. Sociological Theory 20 (1): 41–60.

78.

Taber

(2003) Information Processing and Public Opinion. In: Sears

Huddy

Jervis

(eds) Oxford Handbook of Political Psychology. New York: Oxford University Press, pp.433–476.

79.

United Nations Office on Drugs and Crime (UNODC) (2011) Global Study on Homicide. Available at: https://www.unodc.org/gsh/ (accessed 24 February 2017).

80.

Utas

(ed.) (2012) African Conflicts and Informal Power: Big Men and Networks. London: Zed Books.

81.

Valentino

Brader

Groenendyk

et al . (2011) Election Night’s Alright for Fighting: The Role of Emotions in Political Participation. Journal of Politics 73 (1): 156–170.

82.

Valentino

Hutchings

Banks

et al . (2008) Is a Worried Citizen a Good Citizen? Emotions, Political Information Seeking, and Learning via the Internet. Political Psychology 29 (2): 247–273.

83.

Verba

Schlozman

Brady

(1995) Voice and Equality: Civic Voluntarism in American Politics. Cambridge, MA: Harvard University Press.

84.

Wilkinson

(2004) Votes and Violence: Electoral Competition and Ethnic Riots in India. Cambridge and New York: Cambridge University Press.