Patterns of Protest in Contemporary Africa: An Empirical Investigation of Regional Trends Employing Multiple Imputation

Literature Review and Hypotheses

Mass demonstrations in Africa are not a novel phenomenon, so much so that today we can speak about a Third, or even a Fourth Wave of African protests (Branch and Mampilly, 2015; Mueller, 2018; Rodrigues Sanches, 2022). While the First and Second Waves have been successfully framed around a unifying topic, this has not yet happened for more recent events unfolding in the last 15 years (Branch and Mampilly, 2015). Broadly speaking, the First Wave was connected with the national liberation movements of the 1950s and 1960s (Mueller, 2018). On the contrary, the massive Second Wave demonstrations that unfolded during the 1990s focused more on political institutions and political rights after the economic hardship that had been imposed by the harsh economic policies of the 1970s and 1980s (Bratton and Van De Walle, 1992; Mueller, 2018; Seddon and Zeilig, 2005). ¹

In contrast, the Third Wave has taken place in a completely different environment, from both a technological and material standpoint. The awesome spread of modern communications devices, the Internet and social media helped to form hashtag movements like #FeesMustFall or #ZumaMustFall in South Africa (Bosch, 2019), mobilized protesters in Burkina Faso and DRC (Mateos and Erro, 2021), or spread information on protests internationally through #SudanRevolts in Sudan (Branch and Mampilly, 2015). Although one would be hard-pressed to find a single topic for the last wave, it is clear that Balai Citoyen and Y ʼen a marre had a similar anti-authoritarian focus, and protests in Kinshasa were influenced by demonstrations in Burkina Faso (Polet, 2022). Similarly, the snowball effect that helped spread the Arab Spring across the MENA region was rooted in economic and political grievances, directed toward a ruling class that ruled with an iron fist, and was enabled by technology (Hussain and Howard, 2013).

As a consequence of a changing environment, today’s scholarship on African protests contemplates a diversity of frameworks and approaches. There is, for example, a broad literature that uses class analyses (Engels, 2015; Seddon and Zeilig, 2005), often with a particular focus on the centrality of middle-class protest participation (Daniel et al., 2023; Noll and Budniok, 2023). Other scholars use frameworks rooted in social media mobilization (Bosch, 2019; Mateos and Erro, 2021), resource mobilization (Demarest, 2016), or even political opportunity structures, as was the case for a recent edited volume (Rodrigues Sanches, 2022). These different frames and explanatory approaches to the Third Wave of African protests have proved fundamental in understanding the dynamic diversity of the phenomenon, yet an attempt must be made to consider them jointly.

In this article, we focus on individual reasoning behind protests and sort out them into three broad camps: (1) political protests demanding democracy; (2) protests related to material needs and food insecurity; (3) protests focusing on bad governance, particularly, anti-corruption protests targeted at national or subnational institutions. We elaborate on each of these three sets of factors in the following paragraphs and then use them to operate a separation of the African continent into five different regions, each with its own peculiarities when it comes to protest participation.

While a yearning for democracy was fundamental for the Second Wave of protests of the early 1990s, in contemporary Africa anti-authoritarian sentiments are not considered primary for protest participation. This remains true even if in more developed contexts there has been resistance to democratic backsliding as seen in Zambia (Rakner, 2021). Similarly, protest movements such as Y ʼen a marre in Senegal or Balai Citoyen in Burkina Faso targeted the undemocratic extension of term limits (Touré, 2017). In addition, a tentative democratization process continued in some localities, visible, for example, in the Cameroonian demonstrations against the Biya regime (Manga and Mbassi, 2017). Especially interesting have been some recent pro-democratic protests in unlikely locations, such as those that challenged absolute monarchy in Eswatini in 2018–2019, doomed by limited support in rural areas (Mthembu, 2022) and those unfolding in Angola during the Lourenço presidency, where leadership change had not transformed the regime (de Almeida et al., 2022).

Yet, a connection between some recent protest movements and democratic deterioration does not imply a preference for democracy among protesters. It is well known that European academia tends to attribute to all African protest movements a transformative, democratizing character even when the reality is different (Siméant, 2013). Moreover, within the struggle against colonialism—old and new—the very meaning of democracy can become polluted and assume negative connotations (Brooks et al., 2020; Mignolo, 2011; Engels, 2022a). Dissatisfaction with democracy might imply that those who participate in street demonstrations harbor simultaneously anti-European and anti-democratic sentiments. Moreover, a rejection of authoritarianism and coups d’état is certainly more common among government officials, intellectuals, and outside observers, than among common people (Engels, 2022b). When democracy is inefficient in providing societal change and catering to the lower classes’ needs, non-democratic agents can also have compelling arguments, as recently happened in Burkina Faso. Finally, the African middle class does not have to follow the Global North’s pro-democratic ideas (Daniel et al., 2023).

Taking these two conflicting assumptions seriously results in the stipulation of two opposite hypotheses concerning the democratic attitudes of African protesters which can be formulated as follows:

A second set of factors, assessing the material conditions behind public protests, also implies exploring components that clash with each other. It is uncontroversial that across Africa poor living conditions, unmet personal needs, and unemployment can push people to join public demonstrations in the hope of improving their situation. Many protests involve lower-class citizens whose specific reasons for participation are often built on top of heavy economic grievances (Branch and Mampilly, 2015; Mueller, 2018). Research shows that food insecurity and lowered cereal production in particular, are associated with higher protest intensity (Sánchez and Namhata, 2019). ² This is especially valid for the most recent Third Wave protests, where valence issues related to income inequality and poor service delivery were connected with regime change (Harris and Hern, 2019). The example of the early 2000s South African demonstrations speaks for itself, as neoliberal reforms heavily impacted economically poor segments of society and left them as the losers of democratic rule (Alexander, 2010).

In parallel, the literature also shows that it is generally more affluent, more informed, and more educated people who have the time, energy, and resources to protest, with findings that go beyond the now-old insight of resource mobilization theory (Jenkins, 1983). This was the most prominent mechanism behind the West African “Occupy” movements. In the case of “Occupy Ghana”, the middle-class protest was unconcerned with democracy but was rather connected with how electricity shutdowns endangered social status (Noll and Budniok, 2023). Similarly, in Nigeria, it was middle-class citizens who revolted against the end of fuel subsidies in 2012 (Resnick, 2015). Consistently with this vision, massive geospatial survey data for the whole continent also associates protests with urban localities that have higher average levels of education (Dahlum and Wig, 2019).

The hypotheses for these two diverging trends can be stated as:

Third, political corruption is considered a trigger for protests, with the term itself having become a semantic umbrella for all sorts of government-related grievances (Bauhr, 2017). Yet, the literature shows that high-level corruption does not have a deterministic impact on protest participation, because two effects can be present. One motivates people to protest as a reaction to corruption, in the hope of changing the government’s officials’ behavior. Across Africa, there is evidence that the personal-level experience of petty corruption increases one’s willingness to protest, but also raises reliance upon bribes to get things done (Monyake and Hough, 2019). Corruption and police brutality were the highlights of the #EndSARS (Special Anti-Robbery Squad) protest movement in Nigeria (Omeni, 2022). Similarly, an anti-corruption discourse was a salient component of Burkinabé protests (Touré, 2017).

The opposite effect causes African people to lose hope for real change because of widespread perceptions of impunity for officials who violate their mandate, making protests feel worthless. At the local level, corruption is often accepted as a fact of life, a deviation from a formal, not a social norm (Nicaise, 2019). Furthermore, a recent statistical analysis found that perceptions of elite corruption across the continent have a clear impact on protest participation, while police corruption does not (Lewis, 2021). Another notable study found that officials’ corruption affects protest participation in Senegal, while presidential corruption does not (Inman and Andrews, 2015). The fact that people protest only when they feel there is a chance to alter their realities and against certain types of corruption is consistent with the idea of impunity and corruption as a “fact of life.”

This discussion results in two opposite hypotheses concerning corruption:

These explanatory camps can intersect, as demonstrations are often motivated by a combination of grievances. As recently seen in Tunisia, when economic insecurity comes to the fore, the urban poor can join and sustain anti-democratic protests hoping to improve service delivery (Ash, 2023). Protests also need to include different segments of the population from both cities and rural areas, as well as civil society, and the “urban underclass” to sustain themselves (Branch and Mampilly, 2015). ³ In fact, protests in Africa often result from the involvement of both “generals of revolutions” and “foot-soldiers of revolutions” (Mueller, 2018). Simply put, current protests often appear on the surface pro-democratic because the speakers belong to the middle class. In this regard, Branch and Mampilly (2015) aptly argue that “[t]he presence or absence of multiparty elections in different countries may be less important in determining where protest occurs than the common experience of violent state power and economic deprivation” (p. 73).

Too often, works focusing on African politics either refer to a single country or the whole continent, and one could be satisfied with testing these hypotheses over an Africa-wide sample. Another intriguing possibility is to test these hypotheses over regional samples to test for convergence or for significant discrepancies. The question then becomes: why do we see divergent patterns without a unified response to challenges? And, empirically, which factors separate the African regions in terms of protest participation?

Following the established scholarship on the subject (Bratton and Van De Walle, 1994; Branch and Mampilly, 2015; Cheeseman, 2015), we believe that the overarching cause for different protest trends lies in the way that neopatrimonialism was established across different regions, and then in the long-term regional performance of democratic governance. The overarching idea is that where neopatrimonial regimes were successful in monopolizing political power, protests could only come from the most disenfranchised strata of society, that had been left outside corruption networks. As a consequence, protest in these localities does not have a classically transformative goal, but expresses instead the frustration of citizens from different classes against a system that excludes them, either economically or politically. These protesters do not necessarily believe in democracy nor in the possibility of eliminating corruption, in fact, they would like to participate in the patronage networks for personal gain. Basically, the pragmatism of being out of the neopatrimonial system directs them to protest against the regime (Bratton and Van De Walle, 1994). In contrast, where neopatrimonialism mattered less, the power of trade unions, urban middle classes and more educated classes meant that protesters came from a more affluent background and could therefore afford to have faith in democracy and believe in restoring the rule of law.

Against this initial backdrop, the Third Wave of protest has had an intervening factor in democratization. Regions that have had strong examples of successful transition in the 1990s will generally trust the power of democracy to transform the current political order, while disappointment with democracy will lead to disillusionment and the absence of a democratic transition to a desire for democracy. Therefore, to understand the nature of protests, one has to account for how much the democratic transition has failed to deliver on its expectations. Figure 1 summarizes the whole mechanism.

OPEN IN VIEWER

Figure 1.

Regional Patterns of Neopatrimonialism and Democracy in Africa.

To illustrate how the mechanism works, one can contrast West and East Africa as the most extreme cases and then move on to the rest of the continent. South Africa would generally be on the same “positive” side as the West of the continent, except for the emergence of widespread discontent toward democracy and the appeal of non-democratic options.

From a theoretical point of view, Western Africa constitutes our most positive example, where all three hypotheses should be present in their active version. Across the region, the decolonization process was comparatively peaceful, a feature that has had profound consequences on the permanence of some countries (Senegal, Ghana) as regional democratic leaders in recent years. ⁴ This part of the continent saw the emergence of powerful trade unions that improved the economic position of the working class (Cheeseman, 2015). Concerning corruption, the creation and survival of independent political institutions outside of the patronage systems were fundamental for the creation of viable political opposition against corrupt elites. In fact, the literature review above shows corruption as a strong protest driver in Nigeria, Burkina Faso, and Ghana, so we expect West African responders to be motivated by a desire to solve it. Finally, we also know that the West African middle classes have taken a leading role, which drives a positive expectation for H2 related to resource-led mobilization.

Conversely, in East Africa, societal inequality and the unchallenged continuation of neopatrimonial regimes has created a much more dramatic situation. Here, democratic development has been stifled, while also lacking a country to absolve the role of regional democratic leader, and with the region even experiencing political problems in the more democratic countries. Kenya serves as a prime example. Even though since the 1990s we have had multi-party elections, they are often violent and divisive events (Branch and Cheeseman, 2009). When this rushed democratization failed, it resulted in a strong elite-controlled state influenced by colonialism and rapid state formation (Branch and Cheeseman, 2006). We, therefore, do not expect protesters to be attached to democracy as a superior form of government. Moreover, the centrality of land and poverty questions for protests in Ethiopia (Abebe, 2020; Dias and Yetena, 2022) and neighboring countries such as Kenya (Asingo, 2018) supports the idea that East African protesters are underprivileged. Finally, corruption is not generally the target of protests in this region, because of the necessity to protect the patron–client status quo, so that protesters should not be motivated by a belief in solving it.

Central Africa can then be characterized similarly to the East, with dominant neopatrimonial structures worsened by consolidated exclusionary political regimes. Yet, historically there has been a push toward more democratic politics. While having few experiences with democracy, for example, the Central African Republic and its democratic experiment in the 1990s, we can hardly talk about strong or complete democratization (Mehler, 2005; Vlavonou, 2016). Central Africa has seen some of the longest-serving presidents: Paul Biya (Cameroon), Omar and Ali Mbongo (Gabon), José Eduardo dos Santos (Angola), Teodoro Obiang Nguema Mbasogo (Equatorial Guinea) or Denis Sassou-Nguesso (Republic of Congo) and show constantly lower levels of democracy than the other sub-Saharan regions. This has influenced protests and their composition. In the wave of Anglophone demonstrations that rocked Cameroon in 2016–2017, leading to Ambazonian separatism, people with local power positions and from the cultured strata of society were vocal about implementing Constitutional provisions (Pommerolle and De Marie Heungoup, 2017). In Angola, the 2010s saw a protest resurgence after 20 years of silence, animated by young urban male activists (Pearce, 2015). Later on, with the new president the expectations were high, if unfulfilled, and protesters had the opportunity to push for political liberalization and economic changes (de Almeida et al., 2022). In both cases, democratization was at the center of the demonstrators’ demands, even if the baseline contexts diverged, while corruption did not appear to be a primary driver.

Then, Southern African countries were initially more similar to West Africa in that they had a historical experience of political activism (Cheeseman, 2015), less successful neopatrimonial regimes and experienced a democratic transition. They also have a history of broad social movements fighting for civil and liberal rights, with the South African and Namibian anti-Apartheid movements at the forefront. Southern African protesters used to be motivated by a belief in democracy. Unfortunately, things have changed over the past 15 years, as the economy in South Africa and in the neighboring countries has deteriorated and new political actors have questioned the worth of a democracy that is unable to deliver prosperity. There is also an expectation that protests in Southern African countries are led by the urban poor, establishing a negative relationship in H2. Last, protests in Southern Africa have generally had a hopeful, transformative character vis-à-vis corruption and impunity, as a result of the successful historical fights for civil rights.

Differently from the rest of the continent, Northern Africa saw state-led development projects linked to a mix of populist and authoritarian styles of politics. Furthermore, we expect Northern Africa to exhibit a diverging pattern from sub-Saharan Africa due to the recent weight of the Arab Spring, which followed the lack of democratization in the 1990s, despite some political liberalization (Schneider and Schmitter, 2004). We expect Northern African protesters to be more affluent and educated, not especially pro-democratic, and not motivated by corruption, because of the betrayed democratic expectations of a decade ago (Massoud et al., 2019) and the movement’s urban and technology-related nature (Hussain and Howard, 2013).

Table 1 summarizes how the three hypotheses apply to African regions, keeping in mind that regions themselves can be contentious and that there is some disagreement among different international institutions. In our work, this separation is made keeping in mind that countries such as Angola and Zimbabwe are attributed, respectively, to Central and Eastern Africa for reasons of coherence with the theoretical discussion above. In doing so, we have followed the division made by the United Nations in their statistical reports on the continent.

OPEN IN VIEWER

Table 1.

The Hypotheses in Regional Context.

Hypotheses	Description	North Africa	Western Africa	Central Africa	East Africa	Southern Africa
H1	Protesters believe in democracy	No	Yes	Yes	No	No
H2	Protesters have more resources	Yes	Yes	No	No	Yes
H3	Protesters believe you can solve corruption	No	Yes	No	No	Yes

Following data availability in Afrobarometer 8 (2019–2021), countries were assigned to regions as follows.

North Africa: Morocco, Tunisia, Sudan.

West Africa: Benin, Burkina Faso, Cape Verde, Cote d’Ivoire, Gambia, Ghana, Guinea, Liberia, Mali, Mauritania, Nigeria, Niger, Senegal, Sierra Leone, Togo.

East Africa: Ethiopia, Kenya, Malawi, Mozambico, Tanzania, Uganda, Zambia, Zimbabwe.

Central Africa: Angola, Cameroon, Gabon.

Southern Africa: Botswana, eSwatini, Lesotho, Namibia, South Africa.

Operationalization and Methodological Approach

Moving now from the conceptual to the empirical part of this work, the following paragraphs outline the variables included in the model, justify the choice of a specific methodological approach, and offer guidance for the interpretation of coefficients. Summary statistics—mean, variance, observations, missing values—for variables are also presented in a table to offer a quick quantitative assessment of the sample.

Following our theoretical discussion, the dependent variable reflects responders’ self-reported attendance of a protest march or demonstration. As in Afrobarometer this was evaluated on a scale, it was re-coded as a simple yes/no binary variable, after seeing in a trial run that different intensities of protest intention and participation yielded almost identical coefficients. This transformation also makes the probability interpretation of regression coefficients easier, expressing changes in the likelihood of participating in a demonstration.

Starting with a survey of the independent variables from the more explicitly political factors, the model includes a barrage of factors that measure whether citizens’ attitudes toward democracy and elections influence their propensity to protest (H1). The first two concern the personal satisfaction with democracy in one’s country and whether the responder (1) considers democracy to be the only acceptable form of government, (2) approves of other ruling arrangements, or (3) is indifferent between these two alternatives. Note that the wording of this question is soft, since it does not ask whether one would endorse an authoritarian government. Assuming that indifference is not a middle ground between more and less democratic preferences, but an independent opinion, we treat this as a categorical variable.

Some control variables ensure that the contingency of current politics remains independent from the main variable. Within the democratic camp, two control variables identify whether responders consider elections in their country free and fair, and whether they believe the official results to reflect the real outcome. In this work’s context, both are meant to measure whether electoral grievances increase responders’ protest potential and they are expected to be negative everywhere. They were chosen among election-related questions because of their clarity and reference to universally recognized issues related to representative democracy. In addition, we have included the reported frequency of one’s political discussions, to ensure that participation outcomes are not simply driven by an individual’s general political interest does not drive. Connected to it is a simple binary variable measuring whether people declared that they voted in the last election, to capture whether voting and protests act as complements or substitutes.

Then, to evaluate the validity of the statements in H3, we included two variables that assess the impact of corruption perceptions on protest. One captures whether responders believe that corruption has increased or decreased in their country, and the other whether officials who break the rules are punished. Higher corruption and perceptions of impunity are correlated with the positive reading of the hypothesis, which states that protesters are motivated to defend the “rule of law.” Other corruption-related variables present in the survey specifically targeted some specific power position or asked about whether it is dangerous to report corruption. They were not chosen here because of high missing value rates in the first case and high correlation between missing values and the level of democracy in the second. Then as a control variable for this theoretical group, the model also includes whether responders trust the current president, which makes sure that results will not depend on whether one is a supporter or an opposer of the current administration.

Last, a series of variables act as tests of the standard resource-based hypothesis for protest participation in H2. An important caveat is in order: these factors might not perfectly align with each other at the regional level, showing instead a mixed picture, where, for example, less affluent but also more educated people protest. This said, for each responder, this set of variables covers urban or rural location, education level, satisfaction with living conditions and whether one has a salaried form of employment. Two control variables related to age and gender supplemented this set of factors, following best practices in the discipline. Altogether, this combination of factors should be wide enough to cover different aspects, while also allowing us to keep the empirical explanation sufficiently parsimonious. Here, variables that were coded too granularly in the original survey, such as age and education levels, were recast into tighter categories—labeled cohort and schooling—to simplify the coefficients’ reading and reduce noise, as is shown in Appendix 1.

This article’s statistical model is a logit regression, with data tables reporting both odds ratios and raw coefficients. Odds ratios simplify interpretation, as instead of being centered upon 0, the null hypothesis of no effect corresponds to a “1” value. Values larger than 1 report the probability increase resulting from a one-unit increase in the independent variable, while values below 1 indicate a decrease in the likelihood of protesting. Robust standard errors accounted for the presence of heteroskedasticity in the multivariate model. Beyond these specifications, we adopted a procedure to increase the dataset’s reliability and completeness. As one can see in the summary statistics table at the end of this section, the inclusion of all variables in the model would imply the loss of 27.9% of all observations in the Afrobarometer dataset due to missing values. Precisely because of the presence of missing values across the variables of interest, multiple imputation was employed to run probability estimates and create 10 parallel complete datasets, over which then the model was computed. ⁵ A series of correlated auxiliary variables contributed to the estimation, with Appendix 1 showing the full multiple imputation specification.

In particular, missing values are problematic for capturing protest participation, because people who avoid answering political questions tend to be less politically active. If one adopts the original data without applying multiple imputation, the use of variables such as “corruption perceptions” in combination with protest participation erases from the dataset the responses of many people who did not join public demonstrations. Table 2 below makes this calculation explicit: among the 1967 people who did not answer the corruption question, only 172 also reported that they participated in protests. The difference between the two groups is striking when looking at percentages: among non-responders, the intensity of protest participation is around half. Using corruption perceptions as an individual-level explanatory factor for protest without correcting for missing value bias therefore systematically overestimates protest participation. The same effect can be seen for three other variables with significant missing values, such as perceptions of electoral fairness, perceptions that officials are punished for corruption and personal trust in the president.

OPEN IN VIEWER

Table 2.

Examples of Missing Value Distortion.

Attended demonstration	Corruption response missing in dataset		Free election response missing in dataset		Officials punished response missing		Presidential trust response missing
	No	Yes	No	Yes	No	Yes	No	Yes
No	37,432	1795	35,602	3625	36,880	2347	38,015	1212
	82.56%	91.26%	82.40%	88.35%	82.53%	89.65%	82.73%	89.31%
Yes	7908	172	7908	172	7809	271	7,935	145
	17.44%	8.74%	17.60%	11.65%	17.47%	10.35%	17.27%	10.69%

Notice how, while multiple imputation is infrequent in political science work, it has been previously employed in a major study of education spending under multiparty elections in Africa (Stasavage, 2005). Its use should be encouraged whenever missing values are either too high—for example, above 20%—after the inclusion of all variables of interest and also when a systematic correlation between missing values and the outcome of interest is likely. Those interested in criticizing multiple imputation by arguing that one is introducing data that are not real should consider the following: (1) the inclusion of observations with missing values for one of the variables allows the reinstatement into the model of the true values of all the other variables; (2) if missing values on one variable are not specifically correlated to other factors, the imputation process will use its distribution to attribute random values: any estimate using a largely incomplete variable under these conditions is, therefore, unlikely to produce statistically significant coefficients; (3) the common praxis of running models on survey data with lots of missing values skews results and estimates in unpredictable, unmeasured dimensions, often overestimating the impact of affluent or educated responders.

The previous table (Table 3) presents summary statistics for all variables included in the regression model. Summary statistics are here presented pre-multiple imputation, keeping in mind that for variables with many missing values, the mean might change and the standard deviation will grow, as estimated datasets include more variation.

OPEN IN VIEWER

Table 3.

Summary Statistics and Missing Values for All Variables.

Variable	μ	σ	Min	Max	Observations	Missing values	% missing
Protest	0.758	1.131	0	4	47,307	777	1.62
Cohort	2.669	1.552	1	6	48,072	12	0.02
Gender	1.500	0.500	1	2	48,084	0	0.00
Schooling	3.494	2.305	0	9	47,909	175	0.36
Formal employment	0.338	0.473	0	1	47,908	176	0.37
Urban location	0.453	0.498	0	1	48,084	0	0.00
Living conditions	2.655	1.276	1	5	47,956	128	0.27
Democracy support	1.556	0.737	0	2	46,596	1488	3.09
Democracy satisfaction	2.285	1.029	0	4	46,621	1463	3.04
Free elections	2.970	1.123	1	4	43,798	4286	8.91
Free result	2.987	1.083	1	4	43,080	5004	10.41
Vote	0.723	0.448	0	1	44,759	3325	6.91
Political discussions	0.819	0.729	0	2	47,878	206	0.43
Presidential trust	1.621	1.158	0	3	46,632	1452	3.02
Officials punished	1.349	1.061	0	3	45,267	2817	5.86
Corruption trend	3.672	1.325	1	5	45,992	2092	4.35
		Combined			34,668	13,416	27.90

Results

The results from the quantitative model are presented in the table below (Table 4), divided by region (columns) and by the three hypotheses (rows), with each group featuring explanatory variables and their specific controls. While statistically significant results were present for different factors across the board, the data from the North African region offered the best model fit for protest participation (quasi-R² = 17%). In the following paragraphs, the reader must keep in mind that odds ratio percentages related to variations of more than one unit in an independent variable are to be calculated exponentially. ⁶ This said, we discuss the results following the theory section’s order.

OPEN IN VIEWER

Table 4.

Results of Logit Regression Analysis by Region, Odds Ratios Reported.

		Africa, whole		North Africa		West Africa		East Africa		Central Africa		Southern Africa
		Coeff.	Odds ratio	Coeff.	Odds ratio	Coeff.	Odds ratio	Coeff.	Odds ratio	Coeff.	Odds ratio	Coeff.	Odds ratio
H1	Indifferent to democracy	–0.3022***	0.7391	–0.3517**	0.7034	–0.2412**	0.7857	–0.3120**	0.7319	–0.4913***	0.6575	–0.1882	0.8284
		–0.0337		–0.0951		–0.0632		–0.0847		–0.0701		–0.0831
	Prefers democracy	–0.3234***	0.7234	–0.0621	0.9398	–0.2958***	0.7439	–0.3685***	0.6918	–0.6114***	0.5426	–0.2016*	0.8174
		–0.0259		–0.0937		–0.0438		–0.0581		–0.0499		–0.0726
	Satisfaction with democracy	0.0391*	1.0399	–0.1552***	0.8562	0.0269	1.0273	0.0098	1.0099	0.2956***	1.3439	0.0579	1.0596
		–0.0159		–0.0378		–0.0255		–0.0331		–0.0610		–0.0388
	Belief: elections are free	–0.0502**	0.9510	0.1713**	0.8425	–0.0420	0.9589	–0.0664*	0.9357	0.0710	1.0736	–0.0744	0.9283
		–0.0150		–0.0417		–0.0251		–0.0309		–0.0454		–0.0386
	Belief: results are fair	–0.0736***	0.9290	–0.0921	0.9121	–0.0988**	0.9059	0.0681	0.9342	–0.0139	0.9861	–0.0544	0.9470
		–0.0165		–0.0506		–0.0262		–0.0332		–0.0460		–0.0388
	Voted last election	0.1761***	1.1927	0.1169	1.1241	0.1214*	1.1291	0.2659***	1.3047	0.2641**	1.3023	0.1099	1.1161
		–0.0383		–0.1060		–0.0588		–0.0936		–0.1106		–0.0924
	Frequency of political discussions	0.4338***	1.5432	0.5767***	1.7803	0.4317***	1.5399	0.3293***	1.3900	0.4930***	1.6371	0.4023***	1.4952
		–0.0281		–0.1020		–0.0457		–0.0535		–0.0876		–0.0683
H2	Living standards	0.0342**	1.0348	0.1128**	1.1195	0.0004	1.0004	0.0651	1.0673	0.0612	1.0631	–0.0123	0.9877
		–0.0111		–0.0380		–0.0173		–0.0245		–0.0349		–0.0271
	Urban location	–0.0368	0.9638	0.1326	1.1419	–0.1506**	0.8601	0.0649	1.0671	–0.2609**	0.7703	0.3363***	1.3998
		–0.0279		–0.1025		–0.0388		–0.0658		–0.0700		–0.1072
	Formal employment	–0.0237	0.9766	–0.0395	0.9613	0.0157	1.0159	–0.1000	0.9048	0.0265	1.0268	–0.0070	0.9931
		–0.0280		–0.0923		–0.0463		–0.0530		–0.0896		–0.0779
	Level of schooling	–0.0301	0.9704	0.1331**	1.1424	–0.0163	0.9838	–0.0027	0.9973	–0.2440***	0.7835	–0.0588	0.9428
		–0.0157		–0.0553		–0.0237		–0.0369		–0.0417		–0.0473
	Age cohort	–0.0796***	0.9235	–0.2539***	0.7757	–0.0686***	0.9337	–0.0751***	0.9277	–0.0330	0.9675	–0.0381	0.9626
		–0.0084		–0.0262		–0.0130		–0.0185		–0.0283		–0.0218
	Gender (female = 1)	–0.2122***	0.8088	–0.4351***	0.6356	0.2199***	0.8026	–0.1917**	0.8256	–0.1780*	0.8369	–0.1249	0.8826
		–0.0214		–0.0545		–0.0344		–0.0470		–0.0630		–0.0609
H3	Corruption trend	0.0102	1.0103	–0.0935	0.9107	0.0413*	1.0422	0.0395	1.0403	–0.0130	0.9871	–0.0396	0.9611
		–0.0116		–0.0304		–0.0194		–0.0257		–0.0312		–0.0301
	Belief: crooked officials punished	0.0351**	1.0357	0.1103*	1.1166	0.4022*	1.0431	0.0324	1.0329	0.0011	1.0011	0.0471	1.0482
		–0.0136		–0.0529		–0.0216		–0.0299		–0.0391		–0.0342
	President opinion	–0.0782***	0.9247	–0.0295	0.9709	–0.0554**	0.9461	–0.1501***	0.8606	–0.1381**	0.8710	–0.0697*	0.9327
		–0.0121		–0.0403		–0.0198		–0.0242		–0.0347		–0.0324
	Constant	–0.5148***	0.5976	–0.2488	0.7797	–1.3807***	0.2514	–0.4396***	0.6443	–1.0466***	0.3511	–1.6419***	0.1936***
		–0.0616		–0.2356		–0.0447		–0.1376		–0.0908		–0.0586
	R2	0.0717		0.1721		0.0506		0.0755		0.0721		0.0563
	N	46,664		4175		19,527		13,030		4643		6289

Logit model with heteroskedasticity robust standard errors. Logarythmic coefficients are also featured as odds ratios to facilitate their reading. An increase in two units should be read through an exponential transformation. All values with p<.05 are marked in bold. Significance levels are expressed as *p<.05, **p<.01, ***p<.001. Country-based dummy variables were excluded from this table for readability and are available in the full tables featured in Appendix 1.

Starting from our first hypothesis (H1), probably the biggest surprise came from the impact of different standpoints toward democracy. Both across the whole of Africa and regionally, people who do not fully support democracy—declaring that they might contextually endorse non-democratic regimes—are also inclined to protest more. Our data show that openness to non-democratic governance makes people around 34% more likely to demonstrate in Western Africa and 84% in Central Africa. Overall, people who question democracy protest significantly more everywhere except for North Africa, where it is those who are indifferent to democracy who appear to be significantly correlated.

This can be explained consistently with the earlier theoretical discussion. While protests in Eswatini, #FreeSenegal or Balai Citoyen had a clear pro-democratic spirit, pro-democratic claims were made by middle-class protest leaders and speakers, while the rest focused on other topics (Mueller, 2018). Protests mostly need to raise support across different parts of the population. Therefore, while the most visible and vocal segment could be pro-democratic, a majority of the people protesting simply “ride with the crowd” even if they are not pro-democratic or just indifferent.

Fully pro-democratic citizens were also less likely to protest in Southern Africa perhaps due to the popularity of service-related demonstrations and the space gained by populists. People participating in such demonstrations could feel betrayed by the current government and democratization. Most of the transition days’ promises are left unfulfilled and “rainbowism” has been replaced with populism (Horáková, 2018; Turner, 2019). A good example of this trend is represented by the Economic Freedom Fighters who established a base among those who feel disappointed and “forgotten by the current administration” and have a populist agenda with anti-democratic overtones (Fölscher et al., 2021: 547). ⁷ Support for democracy decreased over the years, as people can increasingly imagine a re-authoritarianization of South Africa (Afrobarometer Data, 2019/2021). It is very probable the participants in Economic Freedom Fighters (EFF) protests would not mind experimenting with less-democratic governance.

While those who are dissatisfied with the current workings of democracy protest in Northern Africa, in Central Africa those who are actually satisfied with national democracy join demonstrations. As Cameroon, Gabon, and Angola are clearly non-democratic, this could signal that people simply protest to oppose sitting presidents, as confirmed by the significant coefficient on that variable. Also notice that in the general sample for all Africa, the coefficients for election-related variables are negative as expected, even if in the regional regressions, the effect is only present in North and East Africa for electoral freedom, and West Africa for suspicion that electoral results are manipulated. As for voting and political engagement, we find a clear correlation between them and protest participation. Those who frequently discuss politics also demonstrate more across the whole of Africa, with a large effect, while voting is clearly a complement and not a substitute for these other forms of participation, with positive coefficients statistically significant in West, East, and Central Africa.

Concerning now the broad hypothesis expressing how the personal availability of resources should influence mobilization (H2), we have a mixed picture both at the general and the regional levels. Standards of living appear to be a differentiating factor only in Northern Africa, where those who belong to the highest income group are 57% more likely to protest than those in the lowest living standard category. On the contrary, having or lacking formal employment does not have any significant role in the protest model for any of the regions, meaning that its impact is subsumed in the other categories. This can be quite surprising, considering the important role that the urban lumpenproletariat is generally thought to have in African demonstrations.

Schooling is also not statistically significant in three of the five regions, exhibiting a positive relationship with the likelihood of protesting in North Africa, and a negative one in Central Africa. A most interesting finding concerns urban and rural location, which changes its impact radically depending on the region. While in Southern African countries, city dwellers are almost 40% more likely to participate in protests, the opposite is true for Central Africa and Western Africa, where those who reside in the countryside are, respectively, 30% and 16% more inclined to attend demonstrations.

Control variables capturing responders’ age and gender show the expected effects, with the more likely participation of young or male citizens, albeit at different levels. Interestingly, neither of these factors is statistically significant in Southern Africa, while Northern Africa shows the strongest effects, with men being on average 57% more likely to demonstrate and a huge gap in favor of the youngest citizens, who are estimated to be around 3.5 times more inclined to join demonstrations than the most elderly cohort.

Finally, the discussion of H3, questioning whether corruption perceptions drive protest across African regions, is limited to the North and West of the continent, since in the other three groups of countries neither variable is statistically significant. Our analysis confirms the anti-corruption nature of West African protests covered by the qualitative literature: protest participation is made more likely by both a belief that corruption is growing in one’s country and faith in punishing abuses of power. As for Northern African responders, only the corruption belief factor is positive and statistically significant. Importantly, the control variable checking for presidential opinion is statistically significant and negative across the four regions of sub-Saharan Africa, confirming that those who oppose a sitting president are more likely to protest. The effect size is especially large in Central and East Africa, where the difference between the highest and lowest categories of presidential approval is well over 50%.

Finally, these results overwhelmingly support the disaggregation of the initial dataset into regional samples. This is because factors that appeared relevant in the all-Africa regression analysis of the first column did not matter across the board once the analysis became granular. The consistent picture that can be traced for the whole continent is one where protesters are more likely to be male, young, and not particularly convinced that democracy is always the best system of government. The reasoning behind these and other findings is grounded in experiences of authoritarian rule, problematic connections between state and society, the urban-rural divide and generational gaps.

Conclusion

Overall, we believe that our work successfully arbitrated some questions related to political protest participation across the African continent. Significant differences appeared between regions, revealing profound variation in what motivates North, West, East, Central, and Southern African citizens to demonstrate. The three main hypotheses concerned democratic beliefs, material conditions, and corruption as drivers of protest. As our literature review showed, there was some uncertainty concerning the sign of the relationships between these factors and citizen participation in demonstrations. We had made some assumptions related to neopatrimonialism and democratization, regarding how these links could work in different regions, and our results did not disappoint. In this conclusive section, we highlight our most important findings, discuss limitations, and offer suggestions for further research.

First, our results concerning democracy were quite sobering and need careful interpretation to avoid some dangerous misunderstandings. While indifference toward democracy or a preference for authoritarian solutions under certain conditions did correlate with protest participation, demonstrators still were more likely to have voted and to discuss politics more with their peers, highlighting a healthy ambivalence toward political institutions. As for resource-based mobilization, our evidence shows moderate support, but only for the continent as a whole and for North Africa. In parallel, the urban–rural divide cuts in different directions in different regions, with Southern African protesters being more concentrated in cities, and their West African counterparts more likely to come from rural areas. To complete the overall picture, corruption-related factors were not a strong determinant of protest participation outside of West Africa, where we found them to increase the likelihood of joining demonstrations, in line with our initial expectations.

While these results confirmed the validity of our endeavor, our study naturally has its own limitations. One is connected to the meaning of democracy. Why is it that while most people prefer democratic rule and reject authoritarianism, those who would not mind some kind of non-democratic rule protest more? The model is silent in this regard, lacking a mechanism coming from the data. Yet, this does not necessarily mean that protests in Africa are anti-democratic, but only that people who are not fully against non-democratic rule are more prone to protest. This finding further seems to confirm previous research that claims that it is important to differentiate between more pro-democratic speakers of protests with the rest of the movements (Mueller, 2018).

Another limit regards the possibly arbitrary division of some regional samples, and the absence of some countries from the survey (e.g. Egypt). Lacking stable definitions of what constitutes Central, Eastern, and Southern Africa, we had to make some decisions that included and excluded some countries. We encourage other scholars to replicate our approach to test for the validity of our results. Our research design also presents the restrictions inherent to a specific kind of theory-testing exercise—a survey-based quantitative analysis—which does not trace new causal patterns, but only tests for the presence of known ones.

Finally, a few suggestions for further research will seek ways to partially remedy these shortcomings, inspired by our reflections during the elaboration of this article. As a general work with a joint focus on several countries, our article can be expanded through more granular research practice. This can happen in several ways, incorporating evidence from different kinds of source materials. For instance, one important contemporary aspect of demonstrating has to do with traditional media coverage, social media usage, and the general discourse around the protesters’ identity and protest legitimacy. A positive societal outlook toward street demonstrations may lead people who usually are politically inactive to join a movement, thus changing group characteristics for a specific location.

Another easy expansion for this study would be to cover a longer time span, as acknowledged above. Especially crucial for results related to perceptions and opinions about democracy would be to capture their evolution over time in the most crucial localities for our study, where people are disillusioned today but also protest. This is because excessive reliance on quantitative methods does not allow researchers to open the “black box” of causality, having to rely instead upon evidence from case-study–focused publications.

Last, let us terminate on a positive note. The results showed an iron-clad relationship between the frequency of political conversations and the likelihood of joining a protest event, valid throughout the African continent. Taking the fully imputed data, talking about politics with one’s peers makes a huge difference: only 10% of those who never discuss politics demonstrate, while 25% of those who discuss politics frequently take part in protest activity. As many political discussions take place through social media, this study should increase interest in studying the online and offline networks conducive to protests to their continued viability, especially in less democratic contexts. Perhaps this should be the focus of a future study of protest participation among young African citizens, who can exploit communication platforms that the older generation could have only dreamed of.