Executive turnovers in sub-Saharan Africa

Abstract

We use competing risks duration analysis to examine the role of ethnicity and governance institutions in executive turnovers of power in 47 sub-Saharan African countries between 1960 and 2008. Using measures that capture the politically important aspects of countries’ ethnic divisions, we find that as compared to countries characterized by one, and only one, potential majority ethnic grouping, leaders’ tenures are shorter in countries with a majority ethnic group that comprises a majority subgroup and that these leaders are at risk of losing power through both coups and elections. In contrast, leaders of countries with no ethnic majority group are better positioned to stave off coups, perhaps by undercutting rivals’ efforts to coordinate among different ethnic groups. Our results, however, suggest that the effectiveness of this strategy wanes over time. Although the results also suggest that the type of regime matters, they do not indicate a monotonic relationship between democracy and elections and authoritarianism and coups.

Keywords

Executive turnover ethnicity Africa coups democratization elections

A central premise to the study of political leadership is that office holding is valuable and leaders seek to hold office as long as possible. Among its benefits, office holding can provide access to patronage and graft and opportunities to influence policies at home and abroad. In some contexts, holding office may be the difference between life and death not just for a leader, but also for his or her supporters and co-ethnics. Power, of course, can be lost through a variety of means including assassinations, coups, election losses, resignations, or term limits.

In sub-Saharan Africa (hereafter, Africa), where different types of executive turnovers occur side-by-side and democratization and economic development have brought profound, but uneven changes, maintaining power requires leaders to navigate complex demographic and political environments. In this regard, Africa provides a rich context to examine the factors influencing executive turnovers of power. African leaders are known for both their long tenures in office and their rapid turnover (Bienen and van de Walle, 1989), often through extra-legal means. Moreover, given that the majority of African countries experienced some type of transition to democracy starting in the 1990s and many have developed multiparty electoral regimes, the drivers and types of executive turnovers vary spatially and temporally. As a consequence, the contours of African leadership—in terms of both longevity and the process by which leadership transitions occur—are ripe for examination.

In this paper we seek to untangle the dynamics underlying executive turnovers in Africa by developing a model assessing how ethnic demography and governing institutions affect the length of African leaders’ terms and the likelihood that they will lose power via regular (e.g. election losses) or irregular (e.g. coups) means. In making this contribution our research addresses a number of considerations that have been the source of a great deal of contention in the literature. Most notably, the role of ethnicity in African political life remains highly contested. Osaghae (1999), for instance, contends that in “Sub-Saharan Africa, ethnic cleavages are the most consequential for politics” (57). In contrast, Elischer (2013) suggests that the influence of ethnicity has been overstated. With respect to executive turnovers, some studies find that ethnic politics heighten the likelihood of certain forms of irregular turnovers of power (e.g. Harkness, 2016; Jenkins and Kposowa, 1992; Roessler, 2011), while others find no such effects (e.g. Arriola, 2009; Collier and Hoeffler, 2005). The influence of ethnicity on voting behavior and elections is similarly debated (cf. Bratton et al., 2012; Carlson, 2015; Ferree, 2006; Posner, 2005).

Our purpose in this study is to help illuminate the relationship between ethnic demography, political institutions and executive turnovers in Africa. The excellent research of scholars such as Ferree (2010), Scarritt and Mozaffar (1999) and Baker et al. (2016), who conceptualize ethnic identity as having different levels of aggregation, is our starting point. Because ethnicity is a choice variable, we expect the salience of the different levels of aggregation to vary depending on the context. We apply and expound on the logic of coordination failure employed by Ferree (2010, 2012) in her examination of the forces behind legislative seat volatility in Africa and Horowitz’s (1985) conception of the attritional coup to theorize about the relationship between different ethnic configurations and executive turnover in Africa.

Following Ferree (2010, 2012), we hypothesize that leaders of countries with a majority ethnic group that has nested within it a subgroup that also constitutes a majority will be at greater risk of losing power through elections than leaders of countries with only one potential majority ethnic group because members of this group may have difficulty coordinating their votes at election time. We also expect these leaders to be more vulnerable to coups because leaders will be tempted to narrow those included in government to members of their subgroup and exclude erstwhile allies from the overarching group. Those excluded will likely want to correct the situation and may therefore attempt to oust the leader. Leaders of countries with only minority groups are expected to be more vulnerable to lose power through an election than leaders of countries with only one potential majority group because the multiethnic coalitions that brought them to power are likely to fall apart. In contrast, we expect them to be less vulnerable to coups because conducting a coup is likely to necessitate inter-ethnic coordination.

Using competing risks duration analysis to analyze this relationship in 47 African countries between 1960 and 2008, our results indicate that as compared to countries characterized by only one potential majority ethnic grouping, leaders’ tenures are shorter in countries with a majority group with a majority subgroup and that these leaders are at risk of losing power through both coups and elections. In contrast, leaders of countries with no ethnic majority group are better positioned to stave off coups, by undercutting rivals’ efforts to coordinate among different ethnic groups. Our results, however, suggest that the effectiveness of this strategy wanes over time.

Our research also bears on our understanding of the linkages between the lengths of leaders’ tenures, the processes by which executive turnovers occur, and the quality of life provided by a regime. While, as Bueno de Mesquita (hereafter, BDM) and Smith (2009: 168) note, “the longer a leader lasts in ofﬁce, the poorer the welfare of the incumbent’s average subject,” others suggest that political instability in the form of coups has had deleterious effects on Africa’s economic growth and development (e.g. Fosu 1992, 2002). We also know that executive turnovers and democratization are integrally linked and that democracy is closely related to the quality of life offered by a country (e.g. BDM et al., 2003; BDM and Smith, 2010; Lake and Baum, 2001). Although the results of our analysis also suggest that the type of regime matters, they do not indicate a monotonic relationship between democracy and elections and authoritarianism and coups.

The remainder of the paper is organized as follows. In the next section we develop our theoretical expectations regarding the influence of ethnicity and governance arrangements on different types of executive turnovers. This is followed by a discussion of the data and methods used to evaluate our hypotheses. After presenting our results, the paper concludes by summarizing our findings and placing them within the context of the literature.

Theory and hypotheses

Underlying our expectations about the drivers of executive turnovers in Africa are two assumptions. First, we assume that political leadership is valuable and that leaders seek to hold office as long as possible. In the African context, where the perks of office can be boundless and the consequences of losing power can be life determining, the desire to maintain power may be particularly acute (e.g. see Bayart, 1993). Second, we assume that leaders adjust their behavior in response to the ethnic composition of their societies and to the institutions in which they operate. Specifically, we assume that ethnicity operates as a choice variable that leaders utilize strategically and that leaders will seek to adjust governing institutions to further their survival.

Ethnic politics and executive turnovers

While the study of ethnicity has long been a staple of the African politics literature, there is little consensus about the effects of ethnicity on leadership survival.¹ Some studies find that ethnic politics fuel coups in Africa (Harkness, 2016; Jenkins and Kposowa, 1992). In contrast, Arriola (2009) finds that ethnic diversity is unrelated to the likelihood of coups in Africa. In a different vein, Omgba (2009) argues that ethnic fractionalization has a negative effect on the length of an African executive’s tenure in office. Adding to the discord, Collier and Hoeffler contend that neither civil wars (2004) nor coups (2005) are grounded in ethnic divisions, while Cederman et al. (2010) and Roessler (2011) find that ethnic exclusion and competition are drivers of political violence.

An evaluation of the effects of ethnicity on electoral politics is similarly muddled. Some studies have found ethnicity to be decisive in vote choice (e.g. Carlson, 2015; Ferree, 2006), while other studies have reported more nuanced results. In countries where ethnicity does influence vote choice, prior research suggests that ethnicity is not the only factor at play (Adams and Agomor, 2015; Long and Gibson, 2015) and may not be the most important one (Bratton et al., 2012; Hoffman and Long, 2013).

Why is there so much discord regarding the role of ethnicity in African politics? One reason may be that ethnic salience varies across countries (Basedau et al., 2011; Elischer, 2013) and subnational units (Kuenzi and Lambright, 2015) and its effects can be moderated by contextual factors (Ichino and Nathan, 2013), including access to non-ethnic centric information (Conroy-Krutz, 2013). Thus, instead of operating as a constant, ethnicity may function as a choice variable (Chandra, 2012) as leaders selectively accentuate ethnic solidarity or exacerbate ethnic tensions in a purposeful manner depending upon circumstances and context. For example, Kuenzi and Lambright (2015) find that gubernatorial candidates in Nigeria are less likely to make ethnic appeals in highly diverse districts, suggesting that overt ethnic appeals may be perceived, in some instances, as detrimental to a candidate’s electoral prospects.

Ethnic-based voting also could be muted by the fact that not everyone may have a co-ethnic contesting an election. For instance, based on an exit poll for the 2007 election in Kenya, Long and Gibson (2015) find support for their hypothesis that voters with a co-ethnic candidate weigh ethnicity much more heavily than those without one. In addition, when elites form alliances, voters who are not co-ethnics of the candidates are much less likely to honor the alliances (Long and Gibson, 2015).

There also may be operational reasons for these differences. Scholars have employed different measures of varying validity to assess ethnicity. For example, because of social desirability bias, survey-based measures from data sources such as the Afrobarometer may underestimate the level of ethnic salience and ethnic-based voting (Carlson, 2016). Measurement issues may also account for the contradictory findings seen in the literature on coups. Many studies have failed to find a relationship between ethnic factors and coup activity and as a consequence, a number of scholars omit ethnicity from their analyses of coups (Harkness, 2016: 589). Harkness (2016) argues that the null findings are not surprising given the broad social indicators used to capture ethnic considerations. According to Harkness (2016) “there is no compelling theoretical story linking such measures to the actions of military officers” (591).

Thus, to better illuminate the linkages between ethnicity and outcomes such as coups or elections necessitates measuring the politically and contextually salient aspects of ethnicity. We do so by building upon the research of scholars such as Ferree (2010, 2012), Scarritt and Mozaffar (1999), and Baker et al. (2016) who have refined broad measures of ethnic demography to better capture the ethnic divisions that are likely to be politically significant.

Central to this approach is the notion that there can be different levels of aggregation of ethnic identity. That is, one overarching ethnic group may comprise different subgroups, with smaller groups “nested” in a larger group. For example, the Mande in Mali comprise three smaller groups: the Bamabara, Malinke, and Dyula (Scarritt and Mozaffar, 1999: 103). Ferree (2010) likens such groups to Russian matryoshka dolls because “the larger groups perfectly contain the smaller ones, so identities do not cross-cut” (764). In order to capture the reality of nested ethnic groupings, Scarritt and Mozaffar (1999) categorize three levels of ethnopolitical groupings for each country.

As Ferree (2012) notes, the strengths of Scarritt and Mozaffar’s African ethnicity data are that they consider the identities that have been politically salient and the multidimensional nature of ethnic identity in Africa. Ferree (2010) uses the Scarritt and Mozaffar (1999) data to classify countries into three categories: countries “where one, and only one, group forms a majority,” which we also refer to as “countries with a majority that has only minority subgroups”; countries where “a majority group contains within itself a second smaller majority group,” which we refer to as “countries with a majority group with a majority subgroup”; and countries where “no social group is large enough to form a majority on its own,” which we refer to as “countries with all minority groups.”

In countries that have a majority ethnic group with a majority subgroup, there is more than one potential majority ethnic group. To illustrate what is meant by a country that has a majority ethnic group with a majority subgroup,² Ferree (2010, 2012) uses the example of Benin where the southerners are the largest politically salient ethnic grouping with about 70% of the population falling into this grouping. The Fon are one of the ethnic groupings nested within the southerners and constitute about 56% of Benin’s population. The Fon ethnic grouping includes subgroups such as the Gun and Azizo.

The research examining cabinet duration in parliamentary systems and coordination failure inform Ferree’s (2010) theorizing of how ethnic configurations influence legislative seat volatility in Africa. She likens countries with one, and only one, majority ethnic group to one-party majority governments which tend to be stable. She compares countries with all minority groups to multiparty or minority cabinets which tend to be less durable. When majority groups with majority subgroups are possible, voters in a nested ethnic group may fail to coordinate their votes to win an election. Ferree’s (2010) results support her expectations, and Long and Gibson’s (2015) findings about the instability of multiethnic coalitions are consistent with Ferree’s (2010) theorizing. Weghorst and Bernhard’s (2014) research supports the expectation that countries with a majority group that has a majority subgroup are likely to have higher levels of legislative seat volatility than countries with only one potential ethnic majority group. However, they do not find evidence that countries with only minority groups experience higher legislative seat volatility.

We extend this research by considering how different ethnic configurations influence executive turnovers through elections and coups. With regard to elections, our theorizing is consistent with Ferree’s (2010). We expect leaders in countries that have a majority ethnic group with only minority subgroups to remain in power as only one potential majority group is positioned to win elections. In countries where more than one majority is possible, we expect there to be less leadership stability as citizens in the subgroups may fail to coordinate their votes or support. As Ferree (2010: 765) explains, some citizens in these groups may be attracted to the smaller potential majority group because smaller groups mean there are fewer people with whom one must share the benefits of power. That is, some may be driven to maximize the per capita benefits of electoral victory. Others may support the largest potential majority grouping to maximize the possibility of electoral victory. The situation is therefore primed for coordination failure. Although leaders in countries with only minority groups may seek to maintain power by dividing the opposition, we would expect the multiethnic electoral coalitions on which the leaders depend to be unstable resulting in a greater risk that leaders will lose power through elections. This logic leads us to the following two hypotheses.

Hypothesis 1A: Leaders of countries with a majority group that comprises a majority subgroup are at greater risk of losing power through an electoral defeat than are leaders of countries with a majority group that has all minority subgroups.

Hypothesis 1B: Leaders of countries with only minority groups are at greater risk of losing power through an electoral defeat than are leaders of countries with a majority group that has all minority subgroups.

With regard to losing power through a coup, we expect those countries with a majority ethnic group with a majority subgroup to produce a riskier environment for an executive as loyalties could easily shift and potential rivals would have a basis for a new ethnic majority. The source of instability in countries with a majority group that comprises a majority subgroup is not just a matter of a coordination problem, however. Horowitz’s (1985) concept of the attritional coup also helps to explain why we would expect more instability with this ethnic configuration. As Horowitz (1985: ch. 12) describes, with each coup, one sees the attrition of an ethnic group or ethnic groups, resulting in more ethnic groups being excluded from the ruling clique. Moreover, broader identities, such as regional identities come to be disregarded. As Horowitz notes, “Narrowing progressively diminishes overarching identifications of this kind and demands greater ethnic precision” (1985: 498). This process is consistent with the conception of ethnicity as a choice variable. The circumstances and interests at play help determine how narrowly ethnic identity will be defined.

Thus, those in the overarching group but who are not in the leader’s subgroup may doubt the leader’s commitment to include them in the ruling clique, often for good reason. The leader may not trust those outside of his/her subgroup and therefore seek to eliminate them from the ruling elite. Moreover, those within the leader’s subgroup might want to see their per capita benefits maximized and therefore want to exclude those outside the subgroup from the clientelistic network. Intense loyalty within the subgroup could potentially be of greater import to the leader than a shallower level of loyalty from a larger group. Indeed, an axiom of group politics is that small groups tend to be more unified in purpose and are better able to coordinate their actions to achieve desired outcomes (Olson, 1964). The leader may therefore choose to emphasize the subgroup identity. Members of the larger group who are eliminated from positions of power in this winnowing process are likely to want to correct the situation. Because they were once part of the ruling clique, they are likely to be in positions that allow them to challenge the leader.

We expect leaders of countries with only minority groups to fare better with regard to coups. In countries with no majority ethnic group, leadership survival may depend upon the level of inter-group coordination among the opposition. Clark (2007) argues that government legitimacy helps to stave off coups. The narrower the ruling elite, the smaller the proportion of people who are likely to see the government as legitimate. If rival groups work together, a leader’s hold on power may be perilous. Absent these alliances, however, leaders may be able to extend their tenures by dividing their opponents or pitting rival leaders against one another.

Indeed, ethnic alliances tend to be fragile while the challenges involved in ousting the leader are likely to be substantial. In particular, Horowitz (1985) argues that governments controlled by small ethnic minority groups tend to be very vigilante in identifying and eliminating threats. We expect that leaders of countries with only one potential majority ethnic group will be more vulnerable to coups than leaders of countries with only minority groups whether someone from the majority group is in power or not. Coordination within an ethnic group is easier than coordination between ethnic groups. If someone of the only potential majority group is not in power, those in this group will have many people with whom they can coordinate to overthrow the sitting leader. If someone from the majority group is in power, members of their own group may wish to take their place and would be positioned to coordinate with others to oust the leader. Or put differently, leaders of large ethnic groups may garner less loyalty from their co-ethnics than leaders of small groups (e.g. see Gallego and Pitchik, 2004; Londregan et al., 1995). We, therefore, have the following two hypotheses about the effects of a country’s ethnic demography on leadership alternations through coups.

Hypothesis 2A: Leaders of countries with a majority group that comprises a majority subgroup are at greater risk of losing power through a coup than are leaders of countries with a majority group that has all minority subgroups.

Hypothesis 2B: Leaders of countries with only minority ethnic groups are at less risk of losing power through a coup than are leaders of countries with a majority group that has all minority subgroups.

We hypothesize that leaders of countries with a majority group that has a majority subgroup are more likely to lose power through both elections and coups than their counterparts in countries with a majority group that has only minority subgroups and therefore only one potential majority group. In contrast, we hypothesize that leaders of countries with no majority ethnic group are more likely to lose power through elections but less likely to lose power through coups. Why the difference in these expectations? A leader of a country with only minority groups depends on coordination with different ethnic groups at election time. Without such coordination, the leader may be more likely to lose power. For example, in Kenya, which does not have a majority ethnic group, President Kibaki was able to assemble an alliance of different ethnic groups that brought him to power in 2003. By the 2007 election, that alliance had fallen apart and significant ethnic violence ensued. Many African countries have a majority runoff system requiring that a president be elected with a majority of the vote, a structural constraint that necessitates significant coordination. In contrast, a leader in a country with only minority groups depends on members of different ethnic groups not coordinating to overthrow him or her through a coup. Given the costs associated with coordination, we would expect the lack of any majority ethnic grouping to increase a leader’s risk of losing power through an election but decreases the likelihood of losing power through a coup.

Governance institutions

We also anticipate that a country’s governance institutions will affect leaders’ tenures. To assess these effects, we draw on selectorate theory (e.g. see BDM et al., 1999, 2003; BDM and Smith 2009, 2010), which argues that leaders’ ability to hold power is shaped by the relative sizes of the winning coalition (those whose support a leader must have to maintain office) and the selectorate (those who could potentially affect the selection of the leader). To maintain the support of the winning coalition, leaders supply a mix of private and public goods. When the winning coalition is small and the selectorate is large, such as in an authoritarian state, leaders emphasize private goods in hopes of decreasing challenges from within the minimum winning coalition. For instance, African leaders may alter the composition of the minimum winning coalition by expanding their patronage coalitions to include rivals in their cabinets (Arriola, 2009).³ Roessler (2011) finds that there is a negative relationship between the number of ethnic groups represented in the central government and the likelihood of a successful coup.⁴

However, as the size of the winning coalition expands, the leader may be unable to placate coalition members with private goods because such goods must be divided among more interests. Eventually, when the winning coalition constitutes a majority of the adult population (e.g. a presidential democracy), a leader may seek to maintain power through the provision of public goods and sound public policy (BDM and Smith, 2010). This, in turn, may lead to leadership challenges from members of the winning coalition that, all else being equal, should result in shorter tenures for leaders of more democratic regimes (BDM and Smith, 2010).

As applied to Africa, selectorate theory yields two hypotheses about the risks associated with different types of executive turnovers.

Hypothesis 3A: Leaders of countries where the minimum winning coalition approaches the selectorate (e.g. as a country becomes more democratic) are at greater risk of losing power through an election loss and are less at risk of losing power through a coup.⁵

Hypothesis 3B: Leaders of countries with smaller minimum winning coalitions (e.g. authoritarian regimes) are at greater risk of losing power through irregular turnovers such as coups and are less at risk of losing power through an election loss.

Although democracy’s mitigating effect on the likelihood of coups has been taken as a forgone conclusion by some, the empirical evidence supporting this relationship is unclear. The two studies that have established this relationship in the context of Africa, Lindberg and Clark (2008) and Posner and Young (2007), do not control for the effects of other variables. As Powell and Chacha (2016: 534) note, subsequent studies employing appropriate controls do not find that democracies are less vulnerable to coups, and Powell and Chacha (2016) find no relationship between democracy and coup attempts. Bell (2016: 1167) finds that the rate at which coups are attempted is about the same in democracies and non-democracies, but that coups are more likely to succeed in democracies than in non-democracies because democratic leaders are constrained in their ability to use repression in the face of threats to their power. In Powell et al.’s (2016) study of coups in Africa, democracy is never a significant predictor. In the models of coups in Africa presented by Rabinowitz and Jargowsky (2018), democracy is either not significant or associated with a higher likelihood of coups. Finally, Hiroi and Omori (2013: 39) find that hybrid regimes are much more at risk of coups as compared to fully autocratic regimes.

Thus, while our expectations about the relationship between governance institutions and executive turnovers may seem tautological, the literature suggests that these effects are not always monotonic. Moreover, African governance institutions can be malleable. Leaders may seek to tweak the rules of the game in hopes of sustaining their tenures by “reforming” electoral processes or pushing to extend or remove term limits.⁶ In some of Africa’s strongest democracies—Botswana, Namibia, and South Africa—elections do not engender turnovers of power in the ruling parties. Although alternations in political parties are not the same as alternations of leaders, we would expect democracy to increase the likelihood of both phenomena. These anomalies suggest that the effects of regime type on different types of executive turnovers are complicated and merit further examination.

Data and methods

To evaluate the hypotheses developed above, we use data from 47 African countries.⁷ Since our interest is in explaining the length of time that leaders hold power and how time interacts with the manner in which leaders lose power, our data are duration in nature. We start the clock in 1960, the year widely heralded as the year of African independence.⁸ For countries gaining independence after 1960, time begins with their year of independence. Time stops in 2008. Leaders who were in power through 2008 are right-censored. In instances when a country has multiple leaders in the same year, we follow the decision rule of Londregan et al. (1995) and use the leader who held power at the start of the year.

We use the leaders identified by Posner and Young (2007) and augment those data with information from rulers.org and other sources. Specifically, we disaggregate Posner and Young’s data into annual duration format and then extend these data to 2008. Thus, our dependent variable is coded as a dichotomous measure that equals zero for years in which a leader remained in power and one for the year that a leader lost power. In total, our data include 222 leaders and 164 failures (excluding natural deaths) from 47 countries for a total time at risk of 2,023 country years. However, because of missing data for some of the independent variables, not all of these observations are used in the analyses presented below (see Table A.1).

Since leaders can be simultaneously at risk of losing power through a variety of means including assassinations, constitutional limits, coups, election losses, and resignations, we estimate our analysis using a competing risks duration model. An extension of the Cox proportional regression model developed by Fine and Gray (1999, also see Cleves et al., 2010), the model is appropriate to assess processes where a subject is at risk of failure for more than one reason. If these risks are orthogonal, then the competing events can be treated as censorings and each outcome can be estimated independently. However, when the competing risks are correlated, proper estimation requires that the event of interest be modeled while also accounting for the risks associated with experiencing the competing event.

Specific to our analysis, we estimate separate models for coups and elections such that for each model the other means by which a leader could lose power are treated as the competing event. That is, for the elections model, the competing event is assassinations, constitutional limits, coups, and resignations, and for the coups model, the competing event is assassinations, constitutional limits, election losses, and resignations.

To assess how governance institutions affect the length of leaders’ tenures and exits, we include variables for governance institutions and cabinet size (Cabinet Size). Following Powell (2012: 1028), we created three regime dummy variables, Democracy, Authoritarian, and Anocracy, with Anocracy serving as the referent category. To gauge the influence of ethnicity and ethnic politics on executive turnovers we include dummy variables for countries that have only minority groups (All Minority Groups) and countries that have a majority group with a majority subgroup (Majority Group with Majority Subgroup). Countries characterized by a majority groups with only minority subgroups (Majority Group with All Minority Subgroups) are the referent category. We also include Years Since Last Coup to account for previous coup activity. Table A.2 summarizes the ethnic categorization of each country in the dataset.

We include controls (not reported in the tables below) for economic performance (GDP Change and GDP Per Capita). Economic growth decreases both the likelihood of regime change (Przeworski and Limongi, 1997) and the frequency of coups (Johnson et al., 1984), while declining growth rates and recessions can lead to executive interruptions (Kim and Bahry, 2008) and coups (Galetovic and Sanhueza, 2000). Specific to the African context, Londregan et al. (1995) and Omgba (2009) find that that economic growth lessens the likelihood that an African leader will lose power. Also included (not reported) are controls for population (Population) and dummy variables for leaders who obtained power through an appointment (Appointed) or who were their country’s first post-independence executives (First Leader). Table A.1 details the measurement of the independent variables and their data sources and summarizes missing data.

Results

We begin by presenting the results for our analysis of coups in Table 1. The values presented in the middle column are the coefficients.⁹ Robust standard errors (clustering on country) are presented in the parentheses below the coefficients and asterisks are used to indicate statistically significant coefficients (p ⩽.05, one-tailed). Positive coefficients increase the risks of an executive turnover, while negative coefficients decrease these risks. Because diagnostics indicate that the assumption of proportionality (e.g. the effects of the covariates are constant over time) was not met by some of the covariates, interactions between these covariates and the log of time are included to correct for this assumption violation. To assess the substantive effects of the independent variables, the right-hand column presents the changes in the subhazard rates (see note 9) for the statistically significant coefficients. Lastly, the bottom of the table reports descriptive information about the dependent variable (e.g. the number of successful coups, the total number of leaders at risk, and total time at risk).

Table 1.

Competing risks duration analysis of executive turnover by coups, 1960–2008.

	Coefficient(Robust Standard Error)	Change in subhazard rate
Majority group with majority subgroup	.544* (.184)	72%
All minority groups	−.027 (.169)
Democracy	−2.12* (.976)	−88%
Authoritarian	.208(.377)
Cabinet size	−.014(.019)
Years since last coup	−.237* (.071)	−22%
Failures/subjects	64/198
Time at risk	1,689
Log-likelihood	−235.34
Wald χ2	215.99*

Inspection of the results suggests that both demographic and institutional factors are significant predictors of coups. Consistent with expectations, the sign and significance for Majority Group with Majority Subgroup indicates that as compared to countries that have a majority group with all minority subgroups and therefore have only one potential majority ethnic group (the referent category), leaders of countries characterized by majority groups with majority subgroups are likely to have shorter tenures and lose power through a coup. The change in the subhazard rate for Majority Group with Majority Subgroup (72%) suggests that this effect is substantively significant as well. The coefficient for All Minority Groups, however, is statistically insignificant, suggesting no difference in the risks of a coup for leaders of countries where there is no majority ethnic group compared to countries where there is one potential majority group.

The results also suggest that prior coup activity is a significant predictor of a coup. Specifically, the negative sign for Years Since Last Coup indicates that the risk of a leader losing power by a coup decreases over time. Or put differently, the more recent a previous coup, the greater the risk that a leader will be deposed by another coup with the change in subhazard rate suggesting an annual decrease of 22%.

Consistent with expectations, the results indicate that democratic political institutions decrease the likelihood that a leader will be removed from power via a coup. Specifically, the sign and significance for Democracy indicates that a leader is at less risk of losing power through a coup in democratic states. The marginal change in the subhazard rate for Democracy (–92%) implies that there are important substantive differences between democracies and anocracies (the referent category) and the manner by which leadership turnovers occur. Leaders of authoritarian regimes, however, do not appear to be at any more or less at risk of losing power through a coup than leaders of anocracies. Although the results do not suggest a curvilinear relationship, they also do not suggest a continuous linear relationship between the level of democracy and coups.

Table 2 presents the results of the analysis examining executive turnovers resulting from election losses. Considering first the ethnicity variables, the sign and significance for Majority Group with Majority Subgroup suggests that leaders of countries that have more than one potential majority ethnic grouping are at greater risk of losing power through an electoral loss than leaders of countries that have only one potential majority ethnic group. The change in the subhazard rate for the coefficient suggests a strong substantive effect. At the same time, the coefficient for All Minority Groups is statistically insignificant, suggesting no difference in the risks of an election loss for leaders of countries where there is one potential majority group as compared to countries where there is no ethnic majority group. These findings dovetail nicely with those of Weghorst and Bernhard (2014) who find that countries with Majority Group with Majority Subgroup have higher rates of legislative seat share volatility than countries with a single potential ethnic majority group but do not find that countries with All Minority Groups have higher rates of volatility.

Table 2.

Competing risks duration analysis of executive turnover by elections, 1960–2008.

	Coefficient(Robust Standard Error)	Change in subhazard rate
Majority group with majority subgroup	2.16* (.470)	764%
All minority groups	.194(.571)
Democracy	−.482(.727)
Democracy x ln(t)	1.55* (.758)	373%
Authoritarian	−17.07* (.785)	−3,880%
Cabinet Size	.109* (.036)	12%
Years since last coup	.028(.181)
Failures/subjects	18/198
Time at risk	1,689
Log-likelihood	−51.10
Wald χ2	4,488.02*

p <.05 (one-tailed). Time varying covariates are interacted with the natural log of time. Also included in the model (not reported) are controls for Appointed, First Leader, First Leader x ln(t), GDP Change, GDP Per Capita, GDP Per Capita x ln(t), and Population.

The results also suggest that, initially, leaders of democratic states are no more at risk of an election loss as compared to their counterparts in anocracies and leaders of authoritarian countries have essentially no risk of losing power through an election. In democracies though, the risk of an election loss increases the longer a leader remains in power as indicated by the positive, statistically significant coefficient for Democracy x ln(t) and the large change in the subhazard rate. This result makes sense. Over time, the opposition can utilize the political rights and liberties democracies provide to organize an effective challenge at the polls and, as democracy coalesces, leaders may be more constrained in using extra-legal means to limit the ascendency of political rivals.

Also note that the positive sign and significance for Cabinet Size indicates that efforts by leaders to expand their cabinets in hopes of placating potential rivals and perhaps reducing the risk of a coup may increase their vulnerability to losing at the polls. Of course, if leaders expand their cabinets to accommodate potentially powerful rivals, then a large cabinet suggests a leader who is facing numerous competitors. By granting rivals cabinet status, a leader may increase their legitimacy in the eyes of the public.

The results in Tables 1 and 2 are largely consistent with our theoretical expectations and suggest that ethnic divisions and a country’s governance arrangements affect the likelihood of executive turnovers via coups and election losses. Yet, by pooling data from the pre and post-Cold War periods, these analyses may obscure differences in these relationships stemming from significant macro-level changes in Africa. The end of the Cold War is widely accepted as a critical juncture in the trajectory of democratization in Africa (e.g. Englebert and Dunn, 2013: 186, inter alia; Wahman, 2014) and elsewhere.¹⁰ In 1989, 29 African countries had constitutions that provided for single-party rule. By 1994, no African countries were under “de jure” one-party rule (Bratton and van de Walle, 1997: 8). Although coups still occur, the majority of African countries have experienced some type of transition to democracy and have developed multiparty electoral regimes. Moreover, many of the “founding fathers,” traditional monarchs, or independence leaders who once dominated their countries’ politics are no longer in power.

As Andersen and Ross (2014) note, most casual relationships identified by political scientists are not likely to have temporal stability and therefore, political scientists should be attentive to the “potential changes in the conditions that affect the causal relationship” (1016). Given the major changes in the macro-political environment outlined above, the longevity of African leadership and the process by which leadership transitions occur are likely to differ in the pre-1990 period and post-1990 period. To assess these differences Table 3 presents the results of the coup model separately for the two periods and for the elections model for the post-1990 period (the elections model for the pre-1990 period cannot be estimated reliably due to insufficient observations).

Table 3.

Competing risks duration analysis of pre and post-Cold War executive turnover, 1960–2008.

	Coups				Elections
	Pre-1990		Post-1990		Post-1990
	Coefficient(Robust Standard Error)	Change in subhazard rate	Coefficient(Robust Standard Error)	Change in subhazard rate	Coefficient(Robust Standard Error)	Change in subhazard rate
Majority group with majority subgroup	.403* (.227)	49%	−.992(1.07)		2.02* (.494)	654%
Majority group with majority subgroup x ln(t)			1.04* (.442)	182%
All minority groups	−.162(.254)		−3.42* (1.41)	–97%	.197(.615)
All minority groups x ln(t)			1.99* (.605)	638%
Democracy	−.478(.737)		−18.12* (.779)	–1340%	−659(1.11)
Democracy x ln(t)					1.59* (.774)	390%
Authoritarian	−.027(.533)		1.41* (.723)	307%	−16.78* (.912)	−4,130%
Cabinet Size	−.001(.019)		−.026(.054)		.079* (.036)	8%
Years since last coup	−.248* (.067)	–22%	−.179* (.099)	–16%	.023(.019)
Failures/subjects	43/111		21/127		17/127
Time at risk	842		847		847
Log–likelihood	−143.08		−42.00		−46.71
Wald χ2	94.96*		3,110.89*		5,160.73*

p <.05 (one-tailed). Time varying covariates are interacted with the natural log of time. Also included in the model (not reported) are controls for Appointed, Appointed x ln(t) (pre-1990 Coups), First Leader, First Leader x ln(t) (all models), GDP Change, GDP Change x ln(t) (post-1990 Coups), GDP Per Capita, GDP Per Capita x ln(t), Population, and Population x ln(t) (post-1990 Coups).

The results presented in Table 3 suggest some interesting period differences in patterns of executive turnover in Africa with respect to coups. Most notably, as compared to the pooled analysis presented in Table 1, the magnitude and timing of the impact of the ethnicity variables differ during and after the Cold War. For both periods, the likelihood of a leader losing power via a coup is greater in countries where there is a majority group with a majority subgroup. In the post-1990 period, however, it is only over time that Majority Group with Majority Subgroup increases the likelihood of a coup. Specifically, the sign and significance for Majority Group with Majority Subgroup x ln(t) indicates that a leader becomes more at risk of being ousted the longer s/he stays in power in contexts where there is a majority group that has a majority subgroup. Inspection of the change in the subhazard rate suggests that this effect is slightly stronger in the post-1990 period. Another significant difference is in the performance of All Minority Groups across the two periods. In the pre-1990 period the coefficient is insignificant, while in the post-1990 period the likelihood of a coup decreases initially in countries where there is no majority ethnic group. However, the sign and significance for All Minority Groups x ln(t) indicates that the longer a leader of a country with such an ethnic configuration remains in power, the more at risk the leader is of being deposed by a coup. In situations where there is no ethnic majority group, it is likely to take time for groups to organize to oust a leader. The fact that this relationship is significant after 1990 but not before 1990 makes sense. Overall, the level of democracy was much lower in the pre-1990 period and thus coordination would have been difficult. The post-1990 period offers more opportunities to coordinate successfully with others. Across both time periods, however, the effect of Years Since Last Coup is consistent and negative suggesting that the risk of a coup decreases as the time since the last coup increases.

Far and away though, the most important difference between the periods for the coup model is in the performance of Democracy and Authoritarian. In the pre-1990 period, Democracy and Authoritarian are insignificant, suggesting that regardless of a country’s governance institutions, a leader was at risk of losing power through a coup. The finding is consistent with Jenkins and Kposowa’s (1992: 271) observation that coups were “the institutionalized method for changing governments in postcolonial Africa.” In contrast, in the post-1990 period, the results indicate that democratic governance institutions effectively buffer leaders from coups. Moreover, the positive, significant coefficient for Authoritarian suggests that, for leaders of authoritarian states, their likelihood of being replaced via a coup increases in the post-1990 period. In both periods, the coefficient for Cabinet Size fails to achieve significance in the models.

The results for the elections model for the post-1990 period are consistent with those for the pooled model. This, of course, is to be expected as all but one of the executive turnovers in our dataset resulting from an election loss occur after the end of the Cold War.

Discussion and conclusions

In this paper we use a competing risks framework to examine patterns of executive turnovers in post-independence sub-Saharan Africa. Central to our analysis is the contention that office holding is valuable and maintaining power requires leaders to adapt their behavior in response to the ethnic composition of their societies and to the governance institutions in which they operate. Because our effort includes a large number of years and countries, accounts for the wave of democratization after the end of the Cold War, and considers both coups and elections, we have significant analytical leverage to assess the drivers of these processes.

Our findings regarding the role of ethnicity are particularly intriguing. As we note above, the role of ethnicity in African politics is unsettled and highly contentious. Theoretically, we conceptualize ethnicity as a choice variable that leaders will accentuate or downplay depending upon the context in which they are operating. Empirically, we have sought to capture these dynamics by using measures that highlight the politically salient aspects of ethnicity. Our approach departs from the indicators used by other scholars who have tended to rely on broad measures of ethnicity with no obvious political relevance, and consequently, have failed to find evidence that ethnic cleavages are a predictor of successful coups and to a lesser extent, election losses.

More specifically, we find that leadership instability is greatest in countries where there is a majority group with a majority subgroup as compared to countries characterized by a majority group with all minority subgroups (i.e. where there is only one potential majority ethnic group). Our results suggest that leaders of countries with the former type of ethnic divisions are at risk of losing power through either a coup or an election defeat. In contrast, leaders who rule over countries with only minority groups are better positioned to stave off coups perhaps by undercutting rivals’ efforts to coordinate among different ethnic groups. At the same time, in the post-1990 period our results suggest that the effectiveness of this strategy may be limited as the longer a leader of a country with no majority ethnic group remains in power, the greater the risk of a successful coup.

The analysis also provides nuance to our understanding of the relationship between democracy and executive turnovers. Our work supports the idea that democratization in Africa is working as indicated by the large increase in the number of executive turnovers resulting from election losses starting in 1990 and a corresponding decrease in the number of successful coups after the Cold War. It does not, however, indicate a simple linear relationship between democracy and leadership turnovers through coups or elections. Although leaders of democratic regimes are generally less likely to lose power through coups, the likelihood of meeting such a fate differs little for leaders of authoritarian regimes versus leaders of anocracies. Only during the post-Cold War period does authoritarianism seem to boost the likelihood of coups. Indeed, regime type had no effect at all on the likelihood of leaders losing power through coups in the pre-Cold War period. With regard to electoral turnovers, authoritarian regimes are much less likely to have leadership turnovers through elections than are anocracies. Initially, democracies are no more likely to have such turnovers than anocracies. Over time, however, this risk increases in democracies probably because members of the opposition are able to avail themselves of the civil liberties, such as the right to free speech and association, that enhance their ability to successfully contest the power of incumbents.

Moreover, there remain opportunities for leaders to adjust governance institutions in the short-term to perpetuate their tenures. Even as the general trajectory across Africa is for increased democratization, countries continue to oscillate along the authoritarian/democracy continuum. As a consequence, African leaders remain at risk of losing power through regular and irregular processes even as elections across the continent are increasingly perceived as “free and fair.” We have sought to identify the forces that influence executive turnovers and give rise to either extreme executive longevity or extreme executive instability given that both types of tenures can undermine economic development and the quality of life of many Africans.

Normatively, our effort suggests a mixed bag with respect to the impact of governance institutions and ethnic politics on leadership tenure and executive turnovers. On the one hand, our results add to the body of evidence indicating that many of the anticipated effects of democratization in Africa have come to fruition. Our results also clearly support the notion that ethnic politics play a vital role in defining the contours of political leadership in Africa. Thus, the degree to which political leaders devote their efforts to navigating these cleavages and use state resources to appease ethnic rivals may affect their ability to promote economic development and better the lives of all of their constituents.

Still, we would be remiss not to note potential limitations to our work that should be explored in future research. Consistent with prior research, we conceptualize ethnicity and governance institutions as separate factors affecting patterns of executive turnovers. It may be that these two sets of factors are strongly correlated such that a country’s ethnic divisions shape its governance institutions. Methodologically this is not problematic for the results presented here since both sets of factors are treated as exogenous predictors. At the same time, additional analyses (not reported) indicate that there is a statistically significant link between the composition of a country’s ethnic demography and its governance institutions such that countries where there is a majority ethnic group with a majority subgroup tend to be more democratic as compared to countries with one or no majority ethnic group; a pattern that is consistent with the work of Baker et al. (2016). Thus, an obvious extension to our effort here would be to assess why countries with particular ethnic divisions tend to be more-or-less democratic.

A second extension of this effort would be to examine how leaders use ethnic-based appeals depending upon their country’s political demography. Drawing on Chandra (2012), we argued that ethnicity operates as a choice variable that leaders utilize strategically to extend their tenures. Understanding how leaders exploit ethnic identities by either aggravating or alleviating ethnic tensions for political gain and under what conditions they choose to do so are topics that are open to further investigation.

Footnotes

Appendix

Table A.2.

Coding of ethnic composition independent variables.

Majority group with majority subgroup	All minority groups^a	Majority group with all minority subgroups
Benin Guinea-Bissau Mauritius	Angola CAR Cameroon Cape Verde Chad Congo Republic Cote d’Ivoire DRC Ethiopia Gabon Gambia Ghana Guinea Kenya Lesotho Liberia Madagascar Mali Mozambique Namibia Sao Tome and Principe Seychelles Somalia Togo Zambia	Botswana Burkina Faso Burundi Comoros Djibouti Equatorial Guinea Eritrea Malawi Mauritania Niger Nigeria Rwanda Senegal Sierra Leone South Africa Sudan Tanzania Uganda Zimbabwe

Weghorst and Bernhard (2014) referred to the category All minority groups as no ethnic minimum winning coalition. They included the homogenous countries of Cape Verde, Lesotho, Sao Tome and Principe, and Seychelles in this category presumably because Scarritt and Mozaffar (1999) find that these countries have no ethnopolitical political groupings at all. Thus, there is not a majority ethnopolitical group that can mobilize politically in these countries. Since these countries do not fit neatly into any of the categories, we ran the models excluding these four cases. The only notable difference in the results from those reported here is majority group with majority subgroup is not significant in the model of coups in the pre-1990 period.

Data from Weghorst and Bernhard (2014) based upon coding rules developed by Ferree (2010) using data from Scarritt and Mozaffar (1999)

Acknowledgements

We wish to thank Daniel Posner, Andrew Young, Leonardo Arriola, and Kimuli Kasara for kindly sharing their data and Jo Andrews for her helpful comments and suggestions. We also thank Brendan Morris, Mallory Waters, and Kate Eugenis for their assistance with data collection.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

ORCID iD

Michelle Kuenzi

References

Adams

Agomor

(2015) Democratic politics and voting behaviour in Ghana. International Area Studies Review 18(4): 365–381.

Andersen

Ross

(2014) The big oil change a closer look at the Haber–Menaldo analysis. Comparative Political Studies 47(7): 993–1021.

Arriola

(2009) Patronage and political stability in Africa. Comparative Political Studies 42(10): 1339–1362

Baker

Scarritt

Mozaffar

(2016) Ethnopolitical demography and democracy in sub-Saharan Africa. Democratization 23(5): 838–861.

Basedau

Erdmann

Lay

et al . (2011) Ethnicity and party preference in sub-Saharan Africa. Democratization 18(2): 462–489.

Bayart

(1993) The State in Africa: The Politics of the Belly. London: Longman.

Bell

(2016) Coup d’état and democracy. Comparative Political Studies 49(9): 1167–1200.

Bienen

Van De Walle

(1989) Time and power in Africa. American Political Science Review 83(1): 19–34.

Bogaards

(2008) Dominant party systems and electoral volatility in Africa: A comment on Mozaffar and Scarritt. Party Politics 14(1): 113–130.

10.

Bratton

van de Walle

(1997) Democratic Experiments in Africa. New York: Cambridge University Press.

11.

Bratton

Bhavnani

Chen

T.-H.

(2012) Voting intentions in Africa: Ethnic, economic or partisan? Commonwealth & Comparative Politics 50(1): 27–52.

12.

Bueno de Mesquita

Morrow

Siverson

et al . (1999) Policy failure and political survival : The contribution of political institutions. Journal of Conflict Resolution 43(2): 147–161.

13.

Bueno de Mesquita

Smith

Siverson

et al . (2003) The Logic of Political Survival. Cambridge, MA: MIT Press.

14.

Bueno de Mesquita

Smith

(2009) Political Survival and Endogenous Institutional Change. Comparative Political Studies 42(2): 167–197.

15.

Bueno de Mesquita

Smith

(2010) Leader survival, revolution, and the nature of government finance. American Journal of Political Science 54(4): 936–950.

16.

Carlson

(2015) Ethnic voting and accountability in Africa: A choice experiment in Uganda. World Politics 67(2): 353–385.

17.

Carlson

(2016) Identifying and interpreting the sensitivity of ethnic voting in Africa. Public Opinion Quarterly 80(4): 837–857.

18.

Cederman

L-E

Wimmer

Min

(2010) Why do ethnic groups rebel? New data and analysis. World Politics 62(1): 87–119.

19.

Chandra

(ed.) (2012) Constructivist Theories of Ethnic Politics. New York: Oxford University Press.

20.

Clark

(2007) The decline of the African military coup. Journal of Democracy 18(3): 141–155.

21.

Cleves

Gutierrez

Gould

et al . (2010) An Introduction to Survival Analysis Using Stata, 3rd ed. College Station, TX: Stata Press.

22.

Collier

Hoeffler

(2004) Greed and grievance in civil wars. Oxford Economic Papers 56(4): 563–595.

23.

Collier

Hoeffler

(2005) Coup traps: Why does Africa have so many coups d’etat? CSAE Working Papers.

24.

Conroy-Krutz

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

25.

Easterly

Levine

(1997) Africa’s growth tragedy: Policies and ethnic divisions. Quarterly Journal of Economics 112(4): 1203–1250.

26.

Elischer

(2013) Political Parties in Africa: Ethnicity and Party Formation. Cambridge University Press.

27.

Englebert

Dunn

(2013) Inside African Politics. Lynne Rienner Publishers, Incorporated.

28.

Fine

Gray

(1999) A Proportional Hazards Model for the Subdistribution of a Competing Risk. Journal of the American Statistical Association 94(June): 496–509.

29.

Ferree

(2006) Explaining South Africa’s Racial Census. Journal of Politics 68(4): 803–815.

30.

Ferree

(2010) The Social Origins of Electoral Volatility in Africa. British Journal of Political Science 40 (4): 759–779.

31.

Ferree

(2012) How Fluid is Fluid? The Mutability of Ethnic Identities and Electoral Volatility in Africa. In Chandra

Kanchan

, ed. Constructivist Theories of Ethnic Politics. New York: Oxford University Press, 312–339.

32.

Fosu

(1992) Political Instability and Economic Growth: Evidence from Sub-Saharan Africa. Economic Development and Cultural Change 40(4): 829–41.

33.

Fosu

(2002) Political Instability and Economic Growth: Implications of Coup Events in Sub-Saharan Africa. American Journal of Economics and Sociology 61(1): 329–48.

34.

Galetovic

Sanhueza

(2000) Citizens, Autocrats, and Plotters: A Model and New Evidence on Coups d’E’ tat. Economics and Politics 12: 183–204.

35.

Gallego

Pitchik

(2004) An Economic Theory of Leadership Turnover. Journal of Public Economics 88(2): 2361–2382

36.

Gören

(2014) How Ethnic Diversity Affects Economic Growth. World Development 59(July): 275–297.

37.

Harkness

(2016) The Ethnic Army and the State Explaining Coup Traps and the Difficulties of Democratization in Africa. Journal of Conflict Resolution 60(4): 587–616.

38.

Hiroi

Omori

(2013) Causes and Triggers of Coups d’état: An Event History Analysis. Politics & Policy 41(1): 39–64.

39.

Hoffman

Long

(2013) Parties, ethnicity, and voting in African elections. Comparative Politics 45(2): 127–146.

40.

Horowitz

(1985) Ethnic Groups in Conflict. Berkeley: University of California Press.

41.

Ichino

Nathan

(2013) Crossing the Line: Local Ethnic Geography and Voting in Ghana. American Political Science Review 107(2): 344–361.

42.

Jenkins

Kposowa

(1992) The political origins of African military coups: Ethnic competition, military centrality, and the struggle over the postcolonial state. International Studies Quarterly 36(3): 271–292.

43.

Johnson

Slater

McGowan

(1983) Explaining African military coups d’etat, 1960–1982. American Political Science Review 78(3): 622–640.

44.

Kim

Bahry

(2008) Interrupted presidencies in third wave democracies. The Journal of Politics 70(3): 807–822.

45.

Kuenzi

Lambright

(2015) Campaign appeals in Nigeria’s 2007 gubernatorial elections. Democratization 22(1), 134–156.

46.

Lake

Baum

(2001) The invisible hand of democracy: Political control and the provision of public services. Comparative Political Studies 34(6): 587–621.

47.

Lindberg

Clark

(2008) Does democratization reduce the risk of military interventions in politics in Africa? Democratization 15(1): 86–105.

48.

Londregan

Bienen

van de Walle

(1995) Ethnicity and leadership succession in Africa. International Studies Quarterly 39(1): 1–25.

49.

Long

Gibson

(2015) Evaluating the roles of ethnicity and performance in African elections evidence from an exit poll in Kenya. Political Research Quarterly 68(4): 830–842.

50.

Marinov

Goemans

(2014) Coups and democracy. British Journal of Political Science 44(4): 799–825.

51.

Olson

(1964) The Logic of Collective Action. Cambridge, MA: Harvard University Press.

52.

Omgba

(2009) On the duration of political power in Africa: The role of oil rents. Comparative Political Studies 42(3): 416–436.

53.

Osaghae

(1999) Conflict research in Africa. International Journal on World Peace 16(4): 53–72.

54.

Posner

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

55.

Posner

(2005) Institutions and Ethnic Politics in Africa. New York: Cambridge University Press.

56.

Posner

Young

(2007) The institutionalization of political power in Africa. Journal of Democracy 18(3): 126–140.

57.

Powell

(2012) Determinants of the attempting and outcome of coups d’etat. Journal of Conflict Resolution 56(6): 1017–1040.

58.

Powell

Chacha

(2016) Investing in stability: Economic interdependence, coups d’etat, and the capitalist peace. Journal of Peace Research 53(4): 525–538.

59.

Powell

Lasley

Schiel

(2016) Combating coups d’état in Africa, 1950–2014. Studies in Comparative International Development 51(4): 482–502.

60.

Przeworski

Limongi

(1997) Modernization: theories and facts. World Politics 49(2): 155–183.

61.

Rabinowitz

Jargowsky

(2018) Rethinking coup risk: Rural coalitions and coup-proofing in sub-Saharan Africa. Armed Forces & Society 44(2): 322–346.

62.

Roessler

(2011) The enemy within: Personal rule, coups, and civil war in Africa. World Politics 63(2): 300–346.

63.

Scarritt

Mozaffar

(1999) The specification of ethnic cleavages and ethnopolitical groups for the analysis of democratic competition in contemporary Africa. Nationalism and Ethnic Politics 5(1): 82–117.

64.

Svolik

(2009) Power sharing and leadership dynamics in authoritarian regimes. American Journal of Political Science 53(2): 477–494.

65.

Wahman

(2014) Democratization and electoral turnovers in sub-Saharan Africa and beyond. Democratization 21(2): 220–243.

66.

Weghorst

Bernhard

(2014) From formlessness to structure? The institutionalization of competitive party systems in Africa. Comparative Political Studies 47(12): 1707–1737.