such changes threaten these ethnically-based military factions, who now face strong incentives to defend their positions of privilege by deposing the new leaders and restoring their co-ethnics to power.

Given this level of pervasive insecurity, it is no wonder that many African leaders turned to ethnicity as a solution to their loyalty issues.

African countries have experienced many violent transitions of power, often at the hands of either military officers or rebel insurgents. In such circumstances, we have come to expect further violence. When leadership changes hands peacefully, however, whether by electoral or other constitutional means, we have greater hope for future stability. Moreover, countries undergoing such peaceful change may be democratizing or consolidating their initial steps toward liberal governance—particularly in Africa where military dictators and autocrats have long ruled. These, then, are contexts where we least desire, and indeed may least expect, a violent reaction from the military—for such a reaction would undermine the positive progress that has been made. After all, the ability to pass power from one leader to another via established rules, without violence or a succession crisis, is both a key indicator of democratic consolidation and one of the great promises of constitutional governance.

Yet, all too frequently, such a violent reaction from the military is exactly what occurs. Out of a documented 67 constitutional transitions of executive leadership (from decolonization until

2005), 21 of them were closely followed by an attempted military coup (within 4 years)—fully 31%. If we confine the data to electoral transitions only, then the prognosis is even worse: here 19 of 48 transitions, or approximately 40%, were followed by coup attempts (see Table 2).^{^[1]}African militaries thus seem to share much responsibility for the fact that constitutional politics, that democracy and the regular alternation of executive leadership that it entails, are having a rough time surviving in many African countries. If we could understand why theses militaries are reacting violently to changes in the head of state, then we could potentially figure out how to make democratization more likely to stick.

In this article, I will argue that ethnic dynamics play a critical role in motivating military officers to depose elected, or otherwise constitutionally appointed, leaders. In Africa, where most states are highly diverse and where no single ethnic group constitutes a majority, the opening of political competition increases the likelihood that executive power will rotate between individuals of different identities. In other words, peaceful transitions of power will often entail changes in the ethnic identity of the governing chief executive. For a diverse society, this is indeed the normative ideal: ethnic identity should neither privilege nor preclude an individual from attaining power.

The problem, as far as civil-military relations are concerned, stems from the historical actions of past leaders. Faced with uncertainty and insecurity, many post-independence African leaders chose to ensure military loyalty by recruiting co-ethnics into their officer corps’. In effect, they created ethnic armies tied through personal loyalty and identity-based patronage to the leader himself. Thus, changes in the ethnic identity of the chief-executive, no matter how peacefully attained, pose a real danger to stability—for such changes threaten these ethnically-based military factions, who now face strong incentives to defend their positions of privilege by deposing the new leaders and restoring their co-ethnics to power.

To evaluate this hypothesis, I employ a mixed-methods framework incorporating both cross national quantitative analyses and qualitative country-case studies. Data was collected from a wide variety of primary, secondary, and tertiary sources on all constitutional leadership transitions in Africa (from decolonization to the present), the ethnicity of pre- and post-transition leaders, subsequent coup attempts, and control variables such as relative wealth and ethnic diversity. Both the entire set of constitutional transitions and the sub-set of those transitions that occurred through elections were analyzed. In both cases, where the ethnic identity of the chief executive changed, the risk of a coup attempt increased significantly: in all models, the ethnic change variable was positive and statistically significant. Moreover, the predicted probability of a coup attempt increased by an average of 47.45 (full data set) and 58.9 (elections only) percentage points when moving from no ethnic change to ethnic change—from a coup risk of under 10% to 50-70%. Indeed, in only two instances did a coup attempt occur after a constitutional transition in leadership without that transition involving a change in ethnic identity (S˜ao Tom´e and Pr´ıncipe in 1991 and Sudan in 1986).

These broad, cross-national patterns are further supported by case study evidence. Following Lieberman’s suggestions for model-testing small-N analysis, embedded within a large-N framework, trace the causal mechanisms of the theory through three “on the line” country-cases: Cameroon and Kenya both experienced a coup attempt after a constitutional turnover in leadership which involved ethnic change, while Botswana experienced changes in leadership not followed by coup attempts where there was also ethnic continuity. In both of the positive cases, we see evidence of an ethnic patronage system reacting to the threat of change, while in the negative case no such ethnic patronage system existed nor could one be threatened by the changes in leadership (due to ethnic continuity).

This article suggests that many African countries are, in a sense, trapped by their history— that democratization efforts and movements toward constitutional governance will continue to be undermined by African militaries so long as such efforts fail to account for the ethnic dynamics that have long infused African civil-military relations. Identifying the problem, however, is not in and of itself a solution. My work should, if anything, emphasize the difficulty and danger involved in transforming an ethnically stacked military into a broadly inclusive institution where rank, privileges, and compensation are based on merit rather than identity. Yet, given the multi-ethnic character of most African societies and the normative desirability of constitutional politics that allow for the rotation of power between identity groups, such military restructuring is ultimately necessary. Further analysis suggests that the international community could play a valuable role in diminishing the risks associated with this transformation—that the presence of foreign troops can protect and

shield struggling civilian governments while they restructure their militaries and extricate themselves from historic systems of ethnic patronage.

2 Literature Review

In writing this article, I have been driven to understand what I think of as an important empirical observation—that, in Africa, when power peacefully changes hands between one head of state and the next, a relatively rare event in and of itself, that transition is frequently undermined by military interventionism. Why do African militaries so often react violently to constitutional changes in leadership?

While this question is certainly connected to two extensive literatures, each of which contain important insights—works concerned with democratization and democratic consolidation and those that focus on explaining military coups—it is not directly addressed in either. Two pieces do come close, however, in that they specificaly address African military interventionism in the context of democratization: John F. Clark’s 1995 article, “The Decline of the African Military Coup,” and a small section in Michael Bratton and Nicolas van de Walle’s 1997 book, Democratic Experiments in Africa. Below, I discuss the insights of each of these literatures in fairly broad strokes while providing a more detailed look at the work of Clark and Bratton and van de Walle.

Within the literature on democratization and democratic consolidation, econometric studies generally agree on a simple and yet frustratingly difficult solution to the problem of autocratic regression: get rich.

These studies consistently find that development, and the wealth it generates, are “what makes democracies endure.”The policy prescription stemming from these findings— develop faster—seems singularly unhelpful, however, for the majority of countries struggling with democratization today. Consider that in 2008 the average annual per capita income in sub-Saharan

Africa was $1991 and that 43 countries world-wide are currently classified as low-income economies (with less than $975 annual per capita income). Even with incredibly strong growth rates, it will take decades (if not much longer) for many of these poor countries to reach income levels that are even remotely ‘safe’ for democracy. Are they thus largely doomed to autocracy and failed democratic experiments for the foreseeable future? Perhaps not: some quite poor countries do succeed in establishing and maintaining basic democratic practices—Kenya, Senegal, and Tanzania for example—which suggests that we should dig deeper than development levels; that we should try to understand the mechanisms of democratic breakdown and how even poor countries can avoid them.

The military coup literature is minimally helpful for understanding why militaries might intervene after peaceful transitions of power—mostly because such works are generally insensitive to context and rarely distinguish between interventions against civilian versus military governments or democratic versus autocratic regimes. Nonetheless, medium-N analyses and case studies do provide potentially significant insights relevant to the question at hand, such as the role that ethnic hostilities have played in inciting particular coup attempts. Their findings, however, have yet to be successfully generalized in a methodologically rigorous way. Large-N studies have found no consistent, significant relationship between the measures of ethnic diversity that they include and patterns of coup attempts. Indeed, only two findings seem generally robust across statistical studies: that poor countries are more susceptible to coups and that once a coup has occurred, more are likely in the future. We are thus left with a vague sense that ethnic politics might matter, at least in Africa, and a familiar frustration in the daunting challenge of poverty.

In his 1995 article, Clark seeks to examine the relationship between political liberalization and the degree of military interventionism in African domestic politics. He concludes that liberalization diminishes the overall risk of military intervention due to the increased legitimacy that democratic regimes possess. As a first attempt to analyze the behavior of African militaries in the specific context of democratization, Clark’s article is a significant contribution. Yet, his methodology— which eschews statistics and fails to compare liberalizing to non-liberalizing states—does not lend itself well to making broad generalizations. Moreover, his qualitative case study evidence points to the fact that many liberalizing African states (9 of 21 by his count) still struggle deeply with (often repetitive)

coup attempts.

Such cases indicate that more work is needed to theorize and test the conditions under which militaries do intervene against democratizing states.

Along similar lines, in a small section of their 1997 book, Bratton and van de Walle explore why some militaries intervene during democratic transitions and to what ends—to push democratization forward or to restore autocratic rule. While their sample size is small, and generalizations must be made with caution, the patterns Bratton and van de Walle observe are worthy of note. First, they find military organizations crucial to the ultimate success or failure of liberalization: where soldiers stepped in to support democratic processes, democracy prevailed; where soldiers moved against democratization, autocracy was restored. Second, they observe that their 6 cases of antidemocratic military interventions were largely driven by the perceived need to defend the existing patronage system. Third, Bratton and van de Walle find that a legacy of military interventionism in politics predisposed security forces to intervene in democratization efforts (whether for or against). Finally, they note that democratic transition itself was least likely to happen in countries where the incumbent political leader was the military’s ethnic patron. For example, in Nigeria, where northern Islamic Hausa-Fulani elites had long dominated the officer corps, General Abacha (of the samebackground) refused to step down from power and allow the democratically elected Abiola (a southern Yoruba) to accede to the presidency.

Taken together, these various works on military interventionism compel us to think critically about the role that military institutions play during processes of democratic transition and consolidation. They suggest that intra-military dynamics—such as the defense of patronage networks and the tensions produced by ethnic factionalism—may be critical in understanding when and why militaries intervene against constitutional governance. In the remainder of this paper, I will attempt to develop a theory of why ethnic factions within the military often move against democracy (in defense of patronage networks) and to test that theory against both cross-national and case-study evidence. Given the difficulty of measuring democratic transition and consolidation in their broadest sense, I focus on a single, yet particularly important, aspect of democracy: legal/constitutional changes in executive leadership. After all, the epitome of democracy is that power can, in fact, peacefully change hands via constitutional means.

3 Theory

Why do militaries seize power after constitutional changes of leadership? The theory presented here is one grounded in the actions of past leaders and how they structured the military in the interest of building and stabilizing loyalty. Under certain conditions, it will be argued, those past actions can severely compromise present and future civil-military relations, leading to violence by the military when power changes hands.

The period of decolonization was a rough time for many African states—in the 1960s, armed conflicts were waged in Algeria, Burundi, Chad, and the Sudan, among others; the Congo completely collapsed after a rank-and-file military revolt; and 34 coups were attempted in 18 different countries.

Given this level of pervasive insecurity, it is no wonder that many African leaders turned to ethnicity as a solution to their loyalty issues—even some, like Ghana’s Kwame Nkrumah, who initially decried such practices. Indeed, ethnic military recruiting is a well-documented phenomenon in Africa. Both Samuel Decalo and Boubacar N’Diaye, in their studies on long standing civilian regimes, discuss ethnic military recruiting as a successful strategy for stabilizing civil-military relations. The general idea is that by “ethnically matching” the military to the civilian regime, a leader can build military loyalty through both the mechanisms of ethnic identification and patronage, thereby avoiding military coups and strengthening their hold on power.¹⁵

While effective for many years, I argue that this historic source of stability has become a persistent cause of instability as long time autocrats finally let go of power (sometimes by dying in office) and, more generally, as Africa democratizes. The opening of political systems and the increasing use of democratic means of contestation have led to more and more peaceful transitions of power—wherein one leader hands over executive office to another via constitutional means. Given Africa’s multi-ethnic character, these developments have increased the likelihood that the ethnic identity of leadership also changes during such transfers of power. In other words, the odds that the pre- and post- transition leaders share the same ethnic heritage has diminished.

A change in the ethnicity of a country’s leader, whether by democratic means or not, is a real problem in places where the military was stacked with the previous leaders’ co-ethnics. Under a new leader who no longer shares their identity, military officers may fear or even immediately experience a rapid decline in their power and privilege. In these circumstances, ethnically based factions in the military have strong incentives to mutiny or try and seize government power for themselves.

Both being tactics to restore the previous status quo. This situation shares many traits with a classic security dilemma—both civilian and military personnel fear that the other will eventually move against them, both may then take actions to protect or defend themselves while they still can, both may view the others’ defensive actions as aggressive, and, ultimately, conflict may result where no party really desired it. Whether coup attempts are staged preemptively, or in reaction to real moves by the new leadership to restructure the military, they at best severely destabilize civil-military relations. At worst, they may lead to the collapse of constitutional politics, the ethnic fragmentation of the army, and even the outbreak of civil war.

In the sections that follow, I evaluate this theory against historical evidence. First, I use cross national quantitative data to analyze the effect of ethnicity when leadership changes hands via constitutional means. The theory predicts that a change in the ethnic identity of the leader is critically important in provoking military reactivity. Thus, if there is no significant relationship between a change in the ethnicity of the leader and subsequent coup attempts, then the theory is invalidated. While I would ideally also like to include an ethnic stacking variable, and thereby further submit the theory to cross-national testing, sufficient data is unfortunately not available.

Second, through comparative case study analysis, I will evaluate how well actual historical experiences align with the proposed mechanisms of the theory. Specifically, in the “theory-aligned” positive cases (Cameroon 1982 and Kenya 1978), where there is both a change in the ethnicity of the leader and a subsequent coup attempt, we should see that (a) the prior leader stacked the military with his co-ethnics, (b) action was taken against them by the new leader, or they feared that such action was imminent, and (c) the actual coup leaders were co-ethnics of the previous leader (and wished to re-install him or take power themselves). In “theory-aligned” negative cases (Botswana), we should find no evidence of ethnic stacking and/or no incentive for the military to react due to ethnic continuity between leaders. Finally, where there is a change in the ethnicity of the leader but no coup attempt (Kenya 2002), we should find that either the military was not stacked under the previous leader or that an intervening variable, such as a foreign power actively protecting the government, interfered with the ability of military officers to react.

Cross-National Evidence

From 1950 to 2005, there were 67 cases of power changing hands from one African leader to another by constitutional means. These cases are drawn from across all of Africa, including North Africa and the islands surrounding the continent (52 countries in total). At most, a single country contributed 5 transitions to the data set (Mauritius) with only a handful of countries represented more than twice (Burundi, Ghana, Liberia, Madagascar, Sierra Leone, and Tanzania each contributed 3 transitions).¹⁷Table 1 summarizes the reasons for these changes in leadership, broken down by electoral and non-electoral successions of power:

Table 1: Reasons for Constitutional Changes of Leadership, 1950-2005
	Number
Elections (n=48)
Planned Democratization or Regularly Scheduled	35
By Transitional Government after Coup	7
After Retirement, Resignation, Impeachment, or Death of Leader	5
By Transitional Government after Peace Treaty	1
Non-Electoral Successions (n=19)
After Death in Office (natural) of Leader	11
After Retirement (non-coerced) of Leader	4

A few preliminary notes on coding are in order. First, to qualify here as a case, the change in leadership must occur through competitive, national elections or by constitutional provisions given the natural death or non-militarily coerced retirement or resignation of the former leader. Hereditary monarchies, where the monarch still controls the government and the military, are not included since there is no possibility of an ethnic change in leadership occurring (examples include Morocco, Swaziland, and Ethiopia under Haile Selassie). Also not counted are transitions to

After Resignation (non-militarily coerced) of Leader	3
By Terms of Power Sharing Agreement	1
Total	67

Many of these power transfers are followed closely by violent military reactions. The most commonly tracked, and perhaps the most commonly experienced, of these reactions is the coup

an interim leader or committee that acts as the head of state during a constitutional process to appoint, elect, or otherwise determine the new leader. Where such a new leader does successfully come to power according to those constitutional means (i.e. the process is not violently interrupted), the case is counted and for coding purposes the “old leader” is taken to be the one who preceded the special, transitional arrangements. For example, where the violent deposition of a leader led directly to a transitional government that then quickly held elections (within a year or two), the ethnicity of the deposed leader and the ethnicity of the newly elected leader are used to determine whether a change in ethnic leadership took place. Finally, more recent transitions (post-2005) are not included in the data set as sufficient time has not elapsed to ascertain the military’s reaction. Cases were compiled from a variety of sources, most importantly the Archigos dataset on global political leaders, BBC Country Timelines, and Keesings Archives. 17 Many other countries have never experienced a constitutional change in leadership, including Burkina Faso, Chad, the Democratic Republic of the Congo, Equatorial Guinea, Eritrea, Ethiopia, Gambia, Guinea, Libya, Togo, Uganda, and Zimbabwe.

attempt. People pay attention when militaries storm capital cities. Thus, reliable, consistent, and accessible records exist for this type of event. Such is not the case for other possible military reactions, such as mutinies and small-scale armed forces infighting. Therefore, coup attempts are used here, as elsewhere, as the primary outcome variable of interest.^{^[1]}

Note, however, that other negative, and even equally violent, reactions from the military do occur and that these types of events may involve the same kinds of ethnic dynamics theorized here. For example, elections held in the Central African Republic in 1993 led to the transfer of executive power from Kolingba (of the southern Yacoma) to Patass´e (of the northern Gbaya). Patass´e rapidly implemented a general policy of providing patronage positions in his administration to fellow northerners. The expansion of this policy to the armed services, involving the blatant replacement of

Coup attempt data was compiled from the Archigos database on political leaders cross-referenced with Patrick McGowan’s data set on sub-Saharan African coups and then expanded to cover the years 1952-1955 (where applicable) and 2002-2009 with supplemental research conducted primarily via LexusNexus and Keesings Archives. A minimum of two reports citing evidence of actual military violence, or military occupation of a government building, or military occupation of a communication or transportation center, count as a coup attempt. Even in such cases, where there is strong evidence that there was no plan or ambition to seize national power, it is not counted as a coup attempt. Following McGowans coding procedures, only attempts where the military hung onto power for more than one week were considered successful.

southerners with northerners, sparked three army mutinies between 1996 and 1997.^{^[1]}Examples such as this suggest that the findings presented here, which are restricted to coup attempts, may actually understate the connection between ethnic changes in leadership and military reactivity.

Of the 67 constitutional power transfers, approximately 1/3 are followed by a coup attempt within 4 years (40% for electoral transitions).^{^[2]}For those countries experiencing a coup attempt within 5 years of a leadership transition, the average length of time between those events was 2.1 years with no coups occurring in the 5th year (see Figure 1 for distribution). Table 2 periodizes the data by decade, showing the patterns of constitutional changes of leadership and subsequent military coup attempts over time. From the table, we can clearly see that this is not just a cold war Figure 1: Time Elapsed between Power Transfer and Coup Attempt

phenomenon. As the third wave of democratization swept across Africa in the 1990s, the number of constitutional leadership transitions skyrocketed. Yet, so too did the number of coup attempts undermining those transitions. And while the record has improved in the 2000s, the problem has by no means disappeared.

Table 2.1: Periodization of Constitutional Changes of Leadesrhip
Decade		Number of		Number Followed		Percentage Followed
		Leadership Changes		by Coup Attempts		by Coup Attempts
1950-59		0		0		-
1960-69		5		3		60%
1970-79		7		3		43%
1980-89		8		2		25%
1990-99		25		9		36%
2000-05		22		4		18%
Total		67		21		31%
	Table 2.2: Periodization of Electoral Transitions Only
	Decade		Number of		Number Followed		Percentage Followed
			Leadership Changes		by Coup Attempts		by Coup Attempts
	1950-59		0		0		-
	1960-69		3		3		100%
	1970-79		2		2		100%
	1980-89		3		2		67%
	1990-99		22		8		36%
	2000-05		18		4		22%
	Total		48		19		40%

Given these overall trends, I now seek to evaluate the hypothesis that a change in the ethnic identity of the leader plays a significant role in provoking military reactivity. Coding ethnic change was, in practice, not as difficult as theory might suggest. While ethnic categories are certainly constructed, mutable, and porous at the edges, they also persist over long periods of time and leaders often readily identify, or are identified with one (or more) of them. Data was culled from a wide variety of primary, secondary, and tertiary sources including, but not limited to: documents from the British National Archives, Library of Congress country studies, the World Directory of Minorities and Indigenous Peoples, Minorites at Risk qualitative data, BBC articles and country timelines, the Encyclopedia of 20th Century African History, and scholarly articles on African history. The ethnicity of each leader (before and after the transition) was coded at the highest level of ethnic aggregation^{^[3]}—unless a compelling reason emerged to consider a sub-group. This was only done for 3 countries: clan was recognized in Somalia, the Ashanti were considered separate from other Akans in Ghana, and the Djerma were recognized as a politically relevant subgroup of the Songhai in Niger. Returning these cases back to the highest level of ethnic aggregation would only alter the final coding in 1 case, Somalia in 1967. If the ethnic identity of the leader prior to the transition was different from that of the leader assuming power, the ethnic change variable was coded as 1 (and 0 otherwise). By this coding method, over one half of constitutional leadership transitions in Africa, between 1950-2005, involved a change in the ethnicity of the country’s leader

(35 of 64, with 3 cases indeterminable).^{^[4]}

Table 3: Ethnic Changes in Leadership and Coup Attempts, 1950-2005
	Coup Attempt within 4 years?
Ethnic Change in Leadership?	Yes	No
Yes	19	16
No	2	27

Bivariate Logit: Coefficient= 2.775, SE= 0.808, significant at p ≤ 0.001 level

Predicted Probability of Coup Given Ethnic Change= 0.52

Predicted Probability of Coup Given No Ethnic Change= 0.107

Difference in Predicted Probabilities= 0.415

Missing Observations= 3 (Liberia 1997, Mozambique 2005, and Tanzania 2005; none with coup attempts)

Table 3 depicts the bivariate relationship between ethnic changes in leadership and coup attempts. The distribution of cases indicates that an ethnic change in leadership is practically a necessary condition for a coup attempt to occur after a peaceful, constitutional change in leadership. Where there is ethnic continuity in leadership, militaries very rarely attempt to seize power. Moreover, of the two cases that do defy the general pattern, one may not appropriately belong in this contextual setting: the Sudan in 1986. Not only was the Sudan once again embroiled in civil war at this time, but it is also one of the 7 cases in which elections were held by a transitional government immediately following a successful military coup. Indeed, of these 7 cases, fully 6 were followed by additional coup attempts within 4 years. A solid argument can be made that the 1986 Sudanese transition, as well as those like it, are caught up in cycles of violence with their own dynamics. The statistical models that follow will include an indicator variable for these cases in order to control for this possibility—although given that 5 of these 7 cases also involved ethnic changes in leadership, it is extremely difficult to parse out separate effects.

While the bivariate analysis of ethnic changes in leadership and coup attempts is informative, there are still relevant factors to control for and further relationships to explore. Due to the small size of the data set (64 total observations), I focus on a core of important variables: most notably, the ethnic change in leadership variable already discussed. To identify control variables, I return to the existing cross-national literatures on democratization and military coups. As noted earlier, these studies consistently find that such events are more likely to occur in poor countries as well as where there is a recent history of military intervention. The following logit models thus include a variable for relative prosperity, log GDPk, which is the natural log of GDP per capita in the year of the power transfer, as well as a variable for military intervention, prior.coups, which is the number of coup attempts in the previous ten years.

It is also reasonable to think that economic shocks, understood as sharp downturns in a country’s economic well-being, may also make a coup attempt significantly more likely. I constructed two indicator variables to capture such shocks. A general economic shock variable, econ.shock, was coded 1 if in any year during the 4 year period following the leadership transition the country experienced a negative growth rate of 1.0% or more (and 0 otherwise). A severe economic shock, severe.econ.shock, was coded 1 if in any year during the 4 year period following the leadership transition the country experienced a negative growth rate of 3.0% or more (and 0 otherwise).^{^[5]}A measure for general ethnic diversity was also included (1961 ELF scores),^{^[6]}ethnic.diversity, in order to replicate previous studies which, based on this variable, have dismissed ethnicity as unimportant in explaining cross-national variation in coups—which we would expect, given that country-wide indicators of diversity do not necessarily capture ethnic dynamics within military institutions. Finally, a version of each model seeks to evaluate whether the results change when an indicator variable, coup.election, is included for the 7 cases of violent political transition (coups quickly followed by elections).

The analysis is run on two slightly different versions of the data: the first model examines all constitutional changes in leadership (n=64) while the second model drops the 19 cases of nonelectoral successions of power (n=48). While certain unpredictable events can cause leadership to change hands without immediate elections in even the most advanced democracies (impeachment and the natural death of the president while in office, for example), I still think it important to analyze electoral successions on their own terms—especially since some of the non-electoral, yet still constitutional changes, in power contained in the data here did take place in arguably autocratic contexts.

Model 1= all constitutional changes of leadership

Model 2= electoral changes of leadership only

b: coup.attempt_i= β₀+ β₁ethnic.change_i+ β₂logGDPk_i+ β₃ethnic.diversity_i+ β₄econ.shock_i

Table 4: Determinants of Military Coups after Constitutional Changes in Leadership

Logit, DV= Coup Attempt within 4 Years

	Model 1: all data		Model 2: elections only
	n= 64		n= 45
	a	b	a	b
Variable	Coefficient	Coefficient	Coefficient	Coefficient
	(SE)	(SE)	(SE)	(SE)
constant	0.568	0.392	-1.220	-0.436
	(4.085)	(4.017)	(4.825)	(4.586)
ethnic change	2.764**	3.376**	3.796**	3.889**
	(0.923)	(1.126)	(1.452)	(1.452)
prior coups	0.504*		0.838*
	(0.239)		(0.337)
ln GDPk	-0.494	-0.413	-0.192	-0.146
	(0.521)	(0.513)	(0.618)	(0.611)
ethnic diversity	-0.166	-1.446	-2.177	-3.512
	(1.264)	(1.280)	(1.884)	(2.086)
economic shock	-0.199	-0.172	-0.003	-0.102
	(0.764)	(0.799)	(0.887)	(0.881)
coup after		3.803*		4.176*
election		(1.595)		(1.854)

∗∗∗ = p ≤ 0.001, ∗∗ = p ≤ 0.01, ∗ = p ≤ 0.05

Missing Observations for all models= 3 (Liberia 1997, Mozambique 2005, and Tanzania 2005; none with coup attempts)

Note: When only severe economic shocks are considered, “prior coups” increases in significance (**) in model 1.b. No other meaningful changes occur.

The multivariate results confirm the findings of the bivariate analysis: the ethnic change variable is positive and statistically significant across all models and their variations. For the full data set, a movement from 0 to 1 in the coding of the ethnic change variable increases the predicted probability of a coup from under 10% to roughly 50%. When only analyzing the subset of electoral successions, the difference is even more striking: here, a movement from 0 to 1 in the coding of the ethnic change variable increases the predicted probability of a coup from under 10% to 66-69%. See Table 4 for the full results of the statistical analysis and Figure 2 for the predicted probabilities of all significant variables.

A prior history with coups also consistently and meaningfully raises the probability of a coup attempt after a constitutional change in leadership. A country with no past history of military interventionism has a 17-21% predicted probability of experiencing a coup attempt after a constitutional transfer of executive power. A country with the average number of coup attempts in the ten year period preceding the change in leadership (0.89) has a 23-32% predicted probability of another coup. Meanwhile, a country with the maximum amount of prior coups (6) has a remarkable 74-93% predicted probability of a post-transfer coup attempt. In each of these reported ranges, the first (lower) number was derived from the models based in the full data-set while the second (higher) number comes from the elections only data. Constitutional changes in leadership that occur directly as a result of military interventionism are, not surprisingly, less likely to stick than change induced by other means. The coup.election variable has a large, positive, and statistically significant effect on the likelihood of a future coup.

Interestingly, no other variables attain statistical significance—not even the economic indicators consistently correlated with coup attempts in other large-N studies. Neither GDP per capita levels in the year of the power transfer nor economic shocks in the period following it are statistically significant (although GDP/k is in the predicted direction). While only the general economic shock variable is reported in Table 4, each of the model variants was run with the severe shock variable in its place—without any notable change in the results. This seems a hopeful finding: poor countries are not necessarily more inhibited by their militiaries from engaging in electoral and other constitutional means of transferring power. Nor do economic downturns, even acute ones, pose insurmountable challenges to newly installed constitutional governments. Finally, the society-wide indicator of ethnic diversity (ELF score) did not predict military reactivity.^{^[7]}Notably, the corFigure 2: Predicted Probabilities of Significant Variables

relation between the ethnic change variable and countries’ ELF scores is only 0.17: more diverse countries do not necessarily have more ethnic changes in leadership than their less diverse peers.

Some would argue, however, that while the mere fact of diversity might not matter, ethnic cleavages may still be driving these results. Where there is a past history of hostility and violence between ethnic groups, leaders may in fact be forced into building military loyalty along ethnic lines. Indeed, pre-existing animosities certainly play an important role in undermining trust between a civilian government of one identity group and a military dominated by their historical rivals. If a high correlation existed between the severity of ethnic cleavages in a society and the use of ethnic recruitment strategies, then the findings here could be subsumed by this prior variable.

standard errors for the constant and several other variables became severely inflated and the model lost explanatory power. I think this resulted from the missing data within these measures—which non-randomly deleted cases from the data-set and significantly decreased the sample size. I thus did not attempt to run the elections-only data, with its already smaller set of cases, with these diversity measures.

Unfortunately, good measures of the relative severity of ethnic tensions and cleavages across societies are hard to come by—and their construction is beyond the scope of this project. Thus, we cannot directly test this alternative argument in the large-N framework.

Nevertheless, there are good reasons to think that conceding causal importance to such preexisting ethnic hostilities would not erase the independent effect of ethnic stacking given an ethnic change in leadership. First, even in highly ethnically politicized contexts leaders have chosen to build military loyalty on the basis of inclusion—i.e. the one does not automatically ensure the other. For example, both Nigeria and Ethiopia attempted to build ethnically diverse militaries in the post-independence period^{^[8]}not only in spite of ethnic and regional tensions but sometimes because of them.^{^[9]}Also, post-conflict states such as South Africa under Nelson Mandela have chosen to integrate former combatants from all sides into new, national armies.^{^[10]}Second, in those countries where ethnicity is not already a source of social conflict, leaders who choose ethnic loyalty as the basis of officer recruiting may effectively create such conflict. At the very least, governments that make an individual’s access to an important, prestigious, and materially rewarding realm of state activity dependent on their ethnic identity, give people a good reason to organize protest and dissent along ethnic lines. For these reasons, we should not assume a straightforward relationship between ethnic cleavages (no matter how severe) and military recruitment strategies.

Observational quantitative data has its limitations—while revealing important correlations between concepts of interest, it cannot itself establish causal effect. These results thus only confirm a suspicion that ethnicity matters to government stability as power changes hands between leaders. To understand why and how ethnic politics may drive military reactivity, we must turn to history and the qualitative evidence it provides.

5 Compartive Historical Evidence

Since the cross-national results satisfactorily support the ethnic change theory, according to Lieberman’s mixed methods design we can turn to a model-testing small-N analysis to assess its robustness.^{^[11]}The goal is to use contextual evidence to ascertain whether the theory works in the manner specified—to trace the intermediate steps of the model and see whether they reasonably explain the behavior of historical actors.^{^[12]}It thus makes sense to deliberately choose “on the line” cases—cases that, according to the cross-national evidence, should be explained by the theory under investigation.³¹I thus primarily select country-cases that either experienced both an ethnic change in leadership and a coup attempt (the yes-yes box in Table 3) or country-cases that experienced neither (the no-no box). Table 5 presents the selected cases distributed across the original bivariate table:

Table 5: Case Study Selection

	Coup Attempt within 4 years?
Ethnic Change in Leadership?	Yes	No
	Cameroon 1982
Yes	Kenya 1978	Kenya 2002
No		Botswana 1980, 1988, (2008)

The selected cases are purposefully drawn from African countries broadly considered stable— states without a history of coups or other forms of severe instability. This distinction serves to clearly separate this context of peaceful transitions of power from contexts in which other dynamics of violence may “over-determine” the outcome. Also, for each country included, I examine every constitutional power transfer, thereby encapsulating the experience of the country over time. This leads to the inclusion of one “off the line” case: Kenya, which in 2002 witnessed a peaceful ethnic change in leadership, via elections, without provoking military reactivity. While deviating slightly from Lieberman’s suggestions for mixed-methods design, I think that including this case is useful both as an additional test of the theory and as a way to develop further explanations for how countries have avoided military interventionism despite an ethnic change in leadership. If the model predicts a military reaction, then we should find evidence of some intervening variable that blocked the likely coup attempt from taking place.

5.1 On the Line Cases: Ethnic Change, Coup Attempt

If a change in the ethnicity of executive leadership indeed provokes military reactivity, then what specific processes should we expect to bear out in the “theory-aligned” positive cases? First, we should find direct evidence that the pre-change leader indeed stacked the military with his coethnics, creating an entrenched patronage system that would be endangered by a change in the ethnic affiliation of the chief executive. Second, we should see that the new leader either took direct action against this identity-based military patronage system or that the officer corps feared such action was imminent. Finally, we should find that the actual coup-leaders were co-ethnics of the previous leader and that their aim was to restore that leader to power or seize it for themselves.

5.1.1 Cameroon (1982)

Available evidence suggests that the 1982 transition in Cameroon fits the ethnic stacking theory. The first President of Cameroon, Ahmadou Ahidjo, a Muslim Peuhl from the North of the country, did indeed stack the military with his own co-ethnics. After Ahidjo’s retirement in 1982, the new President, Paul Biya of the southern Bulu people, began to make restructuring moves against the powerful and privileged Presidential Guard, comprised of Peuhl and allied northern groups. In reaction, elements of the Presidential Guard attempted but failed to overthrow the new government and reinstate Ahidjo. President Biya then moved forward and succeeded with ethnic stacking policies of his own, which have stabilized Cameroonian civil-military relations, at least until the next transition.

Ethnic Stacking Under Ahidjo

At the time of decolonization, there were very few native Cameroonian officers in the colonial military, the first significant class of which graduated from an officers training school in Yaound`e that year: 121/175 officers in 1961 were still metropolitan Frenchmen.^{^[13]}Likewise, the Gendarmes, an elite force of rural military policemen generally considered of a higher calibre than the military itself, were also still officered by the French (only 39 African officers for a force larger than the army in 1961).^{^[14]}Thus the first Cameroonian administration had control over the structuring of the early native officer corps, including recruitment into it.

We also know that the new native officers, and the cohorts to follow them, were predominantly comprised of northerners, specifically the allied Fulani and Peuhl groups. A resistance movement against French rule began in Cameroon around 1955. The armed wing of this movement, the Arm`ee de Liberation Nationale Kamerun (ALNK), was dominated by members of the southern Bamil´ek´e and Bassa ethnic groups.^{^[15]}Both the ALNK and its political wing, the UPC, adopted a Marxist ideology and insurgents received military training and support from various communist governments, including both China and the USSR. The ALNK/UPC were thus perceived by both the British and the French as an extreme security threat to the region and the UPC was banned.^{^[16]}The insurgency movement and its political sympathizers (and southerners more generally) were then excluded from government, even prior to independence. Decolonization thus entailed the handing over of power to a northern-dominated and western-friendly government led by Ahmadou Ahidjo, who continued to fight against the ALNK with British and French aid until the last of the rebels were captured in 1970.^{^[17]}This government, from the time it attained significant powers of self-government in 1958 until Ahidjo retired in 1982, favored the northern Fulani and Peuhl groups, recruiting them extensively into both the civil service and the security forces.^{^[18]}

After Ahidjo: Biya and the Military

In 1982, Ahidjo retired from office and was succeeded according to constitutional procedures by Paul Biya, the current Prime Minister and a christian southerner from the ethnic Bulu group. In January of 1984, Biya secured his occupancy of the President by winning his first full term in a general election. Shortly thereafter, Biya began to act against the former President and to attempt a restructuring of the military in order to diminish the power of Ahidjo’s followers. In February of that same year, Ahidjo was convicted and sentenced to death in abstentia by the Yaound`e military court for subversion and conspiracy to carry out revolution.^{^[19]}Then, on April 5th, President Biya announced his decision to transfer certain northerners from the elite Republican Guard to other military units. The Republican Guard was, at the time, a force of 1,000 soldiers outside the normal military command structure, charged with the protection and security of the president, and still dominated by northerners.^{^[20]}

The very next day (on April 6, 1984), a coup attempt was mounted against Biya by approximately one half of the Republican Guard. The rebels took control of the radio station, attacked the presidential palace in Yaound`e with artillery, seized the airport, and severed communication links with the outside world. It took nearly four days of intense fighting for loyal troops to put down the rebellion. It was claimed, at the time, both by the Minister of State for the Armed Forces, Mr. Tsoungui, and by the Army Chief of Staff, General Semengue, that all of the rebels were northern muslims. Subsequently, areas of the north were subjected to a six-month “military clampdown,” involving roadblocks and security checks. The Republican Guard was also immediately disbanded, its loyal members being placed under the command of the Chief of the National Gendarmerie.^{^[21]}

Using the failed coup attempt to his advantage, Biya then moved forward with discriminatory hiring and promotion policies of his own, both within the civilian government apparatus and in the military. Over the course of his 25 year reign, southerners have come to dominate both politics and the military.⁴¹In particular, members of Biya’s own southern Bulu group, as well as members of the closely related Beti group, disproportionately hold key positions in the military.^{^[22]}Stability was thus re-achieved through renewed ethnic matching policies. Biya, however, is not a young man and the near future will bring another change of leadership in Cameroon. If a northerner returns to power, or perhaps even if a rival southern group claims the Presidency, then this next transition could be just as dangerous as the first; bringing with it the potential for widespread destabilization.

5.1.2 Kenya 1978

The evidence from Kenya also supports the ethnic stacking theory. Additionally, this case highlights the potential importance of foreign military power as an intervening variable. Both Presidents Kenyatta and Moi stacked the militaries with their co-ethnics. And both Presidents Moi (following Kenyatta) and Kibaki (following Moi) made moves against their predecessor’s coethnics within the military. Yet, the officer corps only violently reacted against Moi’s restructuring efforts—and not against Kibaki’s. The parallels between these two cases of transition are overwhelming and yet they nonetheless had different outcomes. One likely explanation points to the role of foreign military powers: there was a strong relationship between the American and Kenyan militaries at the time of Kibaki’s assumption of power while Moi was more or less on his own.

Ethnic Stacking Under Kenyatta

Post-independence political power in Kenya fell into the hands of Jomo Kenyatta (a Kikuyu) and the Kikuyu and Luo dominated KANU party—who inherited an ethnically fractured society with a military dominated by politically marginalized ethnic groups, particularly the Kamba and Kalenjin.^{^[23]}That the military establishment was dominated by a hostile ethnic group^{^[24]}and its allies was, undoubtedly, disturbing to the new government. Kenyatta thus sought, quite immediately, to alter this situation; a choice that very well could have been disastrous.

Kenyatta was, however, very clever in his restructuring of the military and took full advantage of the British protection afforded to him.^{^[25]}First, he refrained from immediately purging other groups from the army. Instead, he built up the paramilitary General Service Unit, which operated under the police command structure, as a counterwright to the regular army while at the same time turning it into an all-Kikuyu force (all non-Kikuyu officers were effectively purged by 1966).^{^[26]}Kenyatta then stacked the newly established air force with Luo and Kikuyu soldiers and officers, territorially dispersed the still relatively small army, and made army rapid deployments dependent on the air force.^{^[27]}Only after these steps had been taken, and beginning in earnest in the early 1970s, did Kenyatta start stacking the army officer corps with Kikuyus, making sure co-ethnics were in key command positions.^{^[28]}Thus, by the mid-1970s, Kenyatta had turned the Kenyan military from an institution with barely any Kikuyu representation to one that was thoroughly Kikuyu dominated.

After Kenyatta: Moi, the Military, and the Coup Attempt

In 1978, Kenyatta passed away while still occupying the presidency. He was replaced according to procedures outlined in the constitution by his vice-president Daniel arap Moi (a Kalenjin). Almost immediately, Moi began dislodging Kikuyu officers from both the police and the military on charges of corruption.^{^[29]}Yet, despite his efforts, Moi’s overall transformation of the military establishment remained slow and limited and he was unable to sack many of Kenyatta’s men from top posts.⁵⁰Thus, while certainly threatened by Moi, the Kikuyu officers in the military remained in positions from which they could orchestrate a response even years after his succession to power.

Four years after Moi took office, ethnically-based factions in the military finally moved against him. A military coup attempt was made on August 1, 1982 and was spear-headed by Luo and Kikuyu junior officers and NCOs of the Air Force and the General Service Unit. The rebels seized the Nairobi Airport, the Voice of Kenya radio station, the telecommunications station, and the post-office. Within a matter of hours, they were put down by loyal sections of the Army and the

General Service Unit. Moi was unharmed in the attempt.^{^[30]}

5.2 On the Line Cases: No Ethnic Change, No Coup Attempt

What evidence should we see in “on the line” cases that have neither ethnic changes in leadership nor coup attempts? The first possibility is that we have continuity—that the previous leader did indeed stack the military with his own co-ethnics, but that the transition of power itself was irrelevant—since the new and old leaders were of the same ethnic group and thus no threat was posed to the existing system of military patronage. The second possibility is that an ethnic patronage system was never created within the military. If the military is a diverse organization insulated from ethnic tampering, with merit-based hiring and promotion procedures, then the ethnicity of the executive leader (and changes between leaders) would be irrelevant.

5.2.1 Botswana 1980 and 1988

The case of Botswana fits both scenarios outlined above. From its inception, the Botswana Defense Force has been a multi-ethnic, non-discrimminatory institution. Changes in the ethnicity of the leader thus would not pose an imminent threat to the officer corps, and should not, all else being equal, provoke a coup attempt. Second, the Tswana tribes have consistently held executive power in Botswana—there has been no ethnic change in leadership. Thus we would doubly predict

stability.

Unlike throughout most of post-colonial Africa, Botswana’s leaders were not swayed by the nationalistic imperative of having one’s own military. Instead, they decided to spare themselves the cost of forming an army and focused their limited budgetary resources on social and economic development programs.^{^[31]}By the mid-1970s, however, general regional destabilization led to a reconsideration of this policy. The gravest security threat emanated from Ian Smith’s apartheid regime in neighboring Northern Rhodesia, and its frequent violations of Botswana’s borders while in pursuit of Zimbabwean freedom fighters.^{^[32]}Increasing revenue streams from diamond exports also contributed to this policy reversal by mitigating financial restraints. Thus, in 1977, the Botswana Defense Force was formed. It was initially organized as a single battalion and the majority of its soldiers were drawn from the existing Police Mobile Unit.⁵⁴The first class of officers were trained at Sandhurst (in Britain) and consisted of 17 men.^{^[33]}They graduated in time to fill out the military command structure upon creation of the army (i.e. no Europeans or other foreigners were seconded to the officer corps). This first class of officers included the President’s son, who received the second in command position in the military hierarchy. Aside from this one possible exception, no other attempts have ever been made in Botswana to manipulate the army leadership on the basis of kin, tribe, ethnicity, or any other identity category.^{^[34]}

Three Peaceful Transitions: Khama to Masire, Masire to Mogae, and Mogae to Khama

The Tswana, an ethnic group comprised of eight culturally similar tribes that speak the same language, likely form a numerical majority in Botswana. No census with ethnic categories has been conducted since colonial times, however, so we cannot be sure. Regardless, the Tswana have monopolized political power for centuries. The British not only ruled indirectly through the Tswana tribal elite, but treated the Tswana as the only identity group that existed in the territory: governance was based on the Tswana, who also dominated the civil service, and other tribal groups were simply assimilated into this structure.^{^[35]}

Seretse Khama, chief of the largest of the eight Tswana tribes, led the Botswana Democratic Party (BDP) to a landslide victory in the 1965 pre-independence parliamentary elections. Indeed, the BDP won 28 out of 31 seats, many with overwhelming margins.^{^[36]}Seretse Khama thus became the nation’s first leader: Prime Minister in the final days of British overlordship and then President upon independence (the title of the office changed with independence).

Since that first election, the BDP and the Tswana have continued to dominate electoral politics and have exclusively held executive power. In 1980, Seretse Khama passed away while still in office and Quett Masire (a Tswana), the vice-president, was confirmed as the new president by a vote of the National Assembly. Then, in 1998, Masire retired and the vice-president, Festus Mogae (also a Tswana), was promoted—again through constitutional procedures. Finally, and although this transition is not included in the larger data set because of its recency, in 2008, Mogae retired and Seretse Khama Ian Khama (first son of the former president, Seretse Khama) assumed the presidency. None of these transitions elicited any sort of violent reaction from the military. Indeed, Botswana has never experienced a coup attempt.

Even though the Tswana dominate civilian politics, their leaders have consistently refrained from politically interfering with military recruitment and have not intentionally stacked the officer corps with co-ethnics. This inspires some hope that were someone of a different ethnic background to rise to presidential power, the military would continue to remain aloof from politics—for the Tswana in the military would have no reason to fear the loss of a privileged position they have never held.

5.3 Off the Line Case: Ethnic Change, No Coup Attempt

In “off the line” negative cases, where there is a change in the ethnicity of the leader but no coup attempt, we should find one of two scenarios. The first possibility, as outlined in the previous section, is that the military is already a diverse organization with non-ethnically based recruitment and promotion procedures (presumably merit-based). Here, ethnic changes in leadership do not challenge ethnic patronage networks because those networks do not, in fact, exist. The second possibility is that all of the same dynamics exist as in the positive “on the line” cases (the yes-yes box in Table 3), but that some intervening variable interrupts the process of staging a coup attempt. The case of Kenya in 2002—as well as the previously discussed post-independence restructuring of the Kenyan military under Kenyatta—suggest that the active protection of a foreign military power can serve such an intervening role by protecting a newly elected government from military reactivity as it restructures or dismantles military ethnic patronage networks.

5.3.1 Kenya 2002

After the 1982 coup attempt, and reasonably worried about military loyalty, Daniel arap Moi decided to continue and even accelerate his “Kalenjinization” of the armed forces.^{^[37]}The Air Force was immediately disbanded and then re-built from the ground up under the direction of a loyal army officer.^{^[38]}Moi then gradually reduced Kikuyu dominance of the army officer corps by replacing those who retired with a Kalenjin officer—or when a qualified Kalenjin was unavailable with another non-Kikuyu. By the mid 1990s, both the army and the GSU were once again thoroughly Kalenjin.^{^[39]}

After Moi: Kibake and the Military

In 2002, after having been barred from seeking another term, Moi peacefully stepped down from power and did not contest the elections of that year. The opposition candidate, Mwai Kibaki (a Kikuyu) won in a landslide victory. After about a month in office, Kibaki conducted a massive sweep of the civil service, military, and police—forcibly retiring or replacing those considered loyal to Moi. These purges included both police chiefs and members of the military top brass,^{^[40]}who were then replaced with Kikuyu’s.^{^[41]}Much of the middle ranks as well as the rank-and-file, however, escaped the purges and remained ethnically diverse. The government seemed to fear that touching the middle ranks and below would cause massive upheaval.^{^[42]}Nonetheless, Moi’s actions against the non-Kikuyu top brass could easily be seen as threatening by other out-group officers, who remained in place and potentially capable of staging a coup.

Why No Coup Attempt? The Potentially Stabilizing Role of British and American Forces

And yet, no coup attempt took place. On the surface, there is nothing to suggest that this transition should have been any less destabilizing to civil-military relations than the last. Both power transfers involved a change in the ethnicity of the leader, across the same major societal cleavage. Both new leaders also immediately made attempts to replace military and police officers with co-ethnics. There is even some reason to think that military intervention would be more likely under the current regime than under Moi’s: the military has been deployed domestically to curb major ethnic violence between the Kikuyu and Kalenjin after the disputed 2007 elections.^{^[43]}Yet, the Kikuyu officers attempted to overthrow Moi but the Kalenjin officers have left Kibaki alone. The question is, why?

One possible explanation is the return of a strong foreign military presence to Kenya. The British military had mostly departed, both in terms of stationed home units as well as seconded officers, by the early 1970s. Indeed, they only maintained a small training team in Kenya from the mid-1970s on.^{^[44]}During the transition from Kenyatta to Moi, there was thus no outside military presence to interfere with a coup attempt. In the last 10-15 years, however, Kenya has once again become a major military partner to a Western power—this time to the United States. Military cooperation began in earnest in the 1990s, when Kenya assisted the U.S. with its brief intervention in Somalia. The relationship deepened after the 1998 Nairobi Embassy bombing and was then further cemented after 9/11, when Kenya became a key partner in U.S. anti-terrorist campaigns. Kenya shares military bases, communications networks, and intelligence with the U.S. military and is also a major recipient of U.S. military aid, receiving a total of nearly $80 million from 1998-2004. Moreover, the U.S. government, including the Joint Chiefs, consider Kenya a ‘critical’ ally—and Washington is deeply concerned by the country’s possible destabilization as a result of ethnic conflict.^{^[45]}It is conceivable that this relationship between the U.S. and Kenyan militaries, combined with the aid flows involved, is having a mitigating effect on the strategic calculations of

Kalenjin officers, who might otherwise have decided to intervene in the political sphere.

6 Foreign Military Protection: Some Evidence

If foreign military protection does play an important role in preventing military reactivity, then we should see some evidence of this cross-nationally. While the extent of missing data prevent a full statistical analysis, we can still test the hypothesis by critically examining a subset of cases where the data is relatively complete—where there was an ethnic change in leadership but no coup attempt (the “yes-no” box in Table 3). Here, we predict that the absence of military violence should have resulted from either (a) the concurrent absence of prior ethnic stacking and/or (b) the intervening presence of a foreign military. Thus, when we place all of the cases with an ethnic change but no coup attempt into a separate 2x2 table—broken down according to the ethnic stacking and foreign protection variables—we should find empty the box corresponding to prior ethnic stacking and no foreign military protection.

For the most part, the data breaks down in the way the theory predicts (see Table 6). Alas, one country proves an exception: both of Malawi’s constitutional transfers of power (in 1994 and 2004) show evidence of prior ethnic stacking without any foreign troop presence. Why then no violent reaction from the military? Does this case undermine the theory as currently specified or is there an explanation for its exceptionalism still consistent with the arguments advanced in this article? I will argue the latter—that in Malawi regional identity (north v. south) has pre-eminent importance; that military stacking has occurred along regional rather than merely ethnic lines (southern); and that transfers of power have not yet crossed boundaries of regional identity (all leaders have come from the south). Thus, if we were to recode Malawi’s constitutional changes in leadership according to regional identity, there would be no change in “ethnicity” and hence no prediction of a coup attempt.

Table 6: Further Analysis of Cases Experiencing an Ethnic Change in Leadership but No Coup Attempt

	Foreign Military Protection?
Prior Ethnic Stacking?	Yes	No
Yes	Gabon 1967 (FRA) Kenya 2002 (USA)	Malawi 1994, 2004
	Burundi 2003 (UN)	Mauritius 2003, 2005
No	Liberia 2005 (UN)	Rwanda 2000*
	Senegal 2000 (FRA)	Tanzania 1985, 1995
	South Africa 1994 (UKG)	Zambia 2002
*The Rwandan military prior to the 2000 transition was in a sense ethnically stacked. It was dominated, however, by Tutsis—the co-ethnics of the post- rather than the pre-transition leader. Thus the transfer of power itself creates no incentives for military reactivity.
Missing Observations= 2 (Ghana 2000 and Nigeria 1998)

Malawi and Regional Identity

In Malawi, arguably, ethnicity matters far less than one’s region of origin: economic disparities, government biases, and voting behavior fall along regional lines while tribal chiefs were never integral to the colonial state and intermarriage across ethnic lines is commonplace. Historically, regional differences have played a more prominent and important role in determining people’s access to goods, services, and employment opportunities than ethnic differences, especially across the northsouth divide. The north of the country suffered from poor soil quality and economic neglect by the colonial state—leading to the mass migration of northerners to the south.^{^[46]}Additionally, during colonial times and unlike throughout much of Africa, tribal chiefs had no real standing in the governance of the colony—meaning that ethnicity did not serve as an intermediary to the state in as sharp of terms as it did elsewhere.^{^[47]}Post-independence leader Hastings Banda continued to uphold the importance of regional identity by instituting discriminatory policies against the north and preferentially allocating development projects to the Central and Southern Regions.^{^[48]}Voting in the 1994 elections continued to fall along regional lines with the victorious United Democratic Front drawing its base of support from the South.^{^[49]}Finally, intermarriage across ethnic lines is so common that chiefly lineages are scrambled across groups.^{^[50]}These facts suggest that regional identity may be of primary importance in Malawi—and if so, then this transition should be considered a continuity rather than a change in “ethnic” leadership.

Moreover, the “ethnic stacking” that has occurred within Malawi’s military institutions also reflects the importance of regional identity—recruitment practices have favored southerners, excluded northerners, and relied on “ethnic neutrals”. After independence, Hastings Banda selectively recruited southerners into the armed forces, with the Chewa (his co-ethnics) and Lomwe groups particularly over-represented in both the police and the army (including the officer corps).^{^[51]}At the same time, Yao and muslim recruits were discouraged but not barred from service and few northerners were permitted to attain high rank.^{^[52]}Interestingly, Banda placed Lomwe, rather than Chewa, in key senior positions of the armed forces. He purportedly did this because they were (and are) perceived as “neutrals” due to their “immigrant status.” The Lomwe began migrating to Nyasaland during the colonial period, roughly around the turn of the 20th century, to escape from the brutalities of Portuguese rule in Mozambique. They were not militant and did not displace existing groups, remaining largely landless. Instead, the Lomwe settled on European plantation estates, taking low-wage jobs nobody else wanted, and quickly became the labor backbone of the plantation sector.^{^[53]}Banda thus created a military stacked with southerners, but with an officer corps particularly dominated by ethnic “neutrals.”

We would not expect a military so structured to react violently against a constitutional transition in executive power unless that change crossed regional boundaries—which it never has. Malawi has experienced two constitutional transfers of power, both of which involved a change in the ethnic (but not regional) identity of the leader. In 1993, after becoming seriously ill, Banda decided to retire from office. Elections were held in 1994 and Bakili Muluzi (a southerner and muslim Yao) of the United Democratic Front (UDF) won the presidency. Power thus passed between southerners, but across ethnic groups. Muluzi continued Banda’s policies of disproportionately promoting southerners, and of discriminating against northerners, in officer and command appointments. In 2004, President Muluzi completed his second, and final, constitutionally permitted term in office. Elections were held and once again the UDF won, this time bringing Bingu wa Mutharika (a southerner and ethnic Nyanja) to executive power. Once again, power passed between southerners and across ethnic groups.

There were no coup attempts nor any other violent reactions by the military after either of these transitions in executive leadership. It is not unreasonable to speculate that the prior stacking according to regional identity and the use of ethnic neutrals in the officer corps contributed to this non-reactivity. If the politically relevant groups in the sphere of civil contestation concur on a policy of stacking ethnic neutrals in the military, then that group has nothing to fear from a change in leadership. And it appears, at least while southerners govern, that the privileged Lomwe position in the military hierarchy is secure.

Returning to examination of table 6 and the question of foreign military protection: if we can legitimately re-classify the Malawi cases in terms of regional identity—in which case they move out consideration here as they properly belong in the “no-no box” in Table 3, then the “yes-no” box here is essentially empty. This evidence supports the theory in general and also the notion that foreign military protection may be playing an important role in fostering stability where such stability might otherwise be incredibly difficult to achieve.

7 Conclusion

The combined weight of the quantitative and qualitative evidence presented here lends significant support to the claim that understanding ethnic politics really is critical to understanding civil-military relations in Africa. Ethnic changes in leadership, even when they occur by peaceful and constitutional means, often lead to violent military reactions—which can be directly traced to the policies of past leaders who have stacked military institutions with their own co-ethnics in an effort to secure military loyalty. These systems of ethnic recruiting created vast, identity-based patronage networks within African military institutions—networks inherently threatened by a change in the ethnic-identity of the chief executive and willing to act violently in defense of their power and prestige.

Yet, all is not lost. We can and should strive toward the normative ideal of constitutional politics where no single ethnic group dominates the state. While this article underscores the difficulty and danger involved in restructuring military institutions—in dismantling ethnic patronage networks and building merit-based recruitment and promotion systems not tied to a particular leader’s identity—it also suggests a way that the international community could play a pivotal role in helping countries make this transition. The experience of Kenya suggests that the influence of foreign militaries, whether through territorial proximity or aid flows, may exert a mitigating effect on the strategic calculations of domestic military actors, giving them a reason to stay out of politics. Foreign troops can protect and shield struggling democracies while they restructure their militaries and extricate themselves from inherited systems of military ethnic patronage.

Appendix: Constitutional Changes in Leadership

Country	Year of	Reason for	Ethnic	Coup
	Leadership	Change	Change?	Attempt
	Change			in 4 yrs?
Algeria	1978	Natural Death	No	No
Algeria	1999	Elections	No	No
Angola	1979	Natural Death	No	No
Benin	1991	Elections	Yes	Yes
Botswana	1980	Natural Death	No	No
Botswana	1998	Retirement	No	No
Burundi	1993	Elections	Yes	Yes
Burundi	2003	Power Sharing Agreement	Yes	No
Burundi	2005	Elections	No	No
Cameroon	1982	Retirement then Elections	Yes	Yes
Cape Verde	2001	Elections	No	No
Central African Republic	1993	Elections	Yes	Yes
Comoros	1998	Natural Death	Yes	Yes
Congo-Brazzaville	1992	Elections	Yes	Yes
C˜ote d’Ivoire	1993	Natural Death	No	No
Djibouti	1999	Elections	No	No
Egypt	1970	Natural Death	No	No
Gabon	1967	Natural Death	Yes	No
Ghana	1969	Resignation then Elections	Yes	Yes
Ghana	1979	Coup then Elections	Yes	Yes
Ghana	2000	Elections	Yes	No
Guinea Bissau	2000	Coup then Elections	Yes	Yes
Guinea Bissau	2005	Coup then Elections	Yes	Yes
Kenya	1978	Natural Death	Yes	Yes
Kenya	2002	Elections	Yes	No
Lesotho	1998	Elections	No	No
Liberia	1971	Natural Death	No	No
Liberia	1997	Peace Treaty then Elections	No	No
Liberia	2005	Elections	Yes	No
Madagascar	1993	Elections	No	No
Madagascar	1996	Impeachment then Elections	No	No
Madagascar	2002	Elections	Yes	Yes
Malawi	1994	Retirement then Elections	Yes	No
Malawi	2004	Elections	Yes	No
Mali	1992	Coup then Elections	No	No

Country	Year of	Reason for	Ethnic	Coup
	Leadership	Change	Change?	Attempt
	Change			in 4 yrs?
Mali	2002	Elections	No	No
Mauritius	1982	Elections	No	No
Mauritius	1995	Elections	No	No
Mauritius	2000	Elections	No	No
Mauritius	2003	Elections	Yes	No
Mauritius	2005	Elections	Yes	No
Mozambique	1986	Natural Death	No	No
Mozambique	2005	Elections	?	No
Namibia	2004	Elections	No	No
Niger	1993	Elections	Yes	Yes
Niger	1999	Coup then Elections	Yes	Yes
Nigeria	1979	Elections	Yes	Yes
Nigeria	1998	Natural Death then Elections	Yes	No
Rwanda	2000	Resignation	Yes	No
S˜ao Tom´e and Pr´ıncipe	1991	Elections	No	Yes
S˜ao Tom´e and Pr´ıncipe	2001	Elections	Yes	Yes
Senegal	1981	Retirement	No	No
Senegal	2000	Elections	Yes	No
Seychelles	2004	Resignation	No	No
Sierra Leone	1964	Natural Death	No	No
Sierra Leone	1967	Elections	Yes	Yes
Sierra Leone	1996	Coup then Elections	Yes	Yes
Somalia	1967	Elections	Yes	Yes
South Africa	1994	Elections	Yes	No
South Africa	1999	Elections	No	No
Sudan	1986	Coup then Elections	No	Yes
Tanzania	1985	Retirement	Yes	No
Tanzania	1995	Elections	Yes	No
Tanzania	2005	Elections	?	No
Tunisia	1987	Retirement	No	No
Zambia	1991	Elections	No	No
Zambia	2002	Elections	Yes	No

Primary Sources: Notation

British National Archives (Kew)

CO= Colonial Office

WO= War Office

Keesings= Keesing’s World News Archive

Available: http://www.keesings.com

Newspapers= full citation in text

U.S. National Archives (College Park, MD)

RG= Record Group

[1] World Directory of Minorities and Indigenous Peoples- Central African Republic: Overview.

[2] Four years is chosen as a reasonable time frame in which to measure a military reaction for two principle reasons: first, it takes some time to plan and stage a coup attempt and strategic actors may indeed wait a while to let the new administration settle in and see how disagreeable their policies will actually be. Second, expanding the period of measurement any further would increase the likelihood of confusing coups conducted for completely independent reasons with the effects of the leadership transition itself.

[3] In no case was regional identity used as the basis of coding, even though region plays an important role in the politics of many African countries. Excluding this identity category may again lead to an understating of results.

[4] The ethnic identity for at least one of the two leaders involved in the transition could not be ascertained for Liberia in 1997, Mozambique in 2005, and Tanzania in 2005.

[5] If during the 4 year period of observation there was a coup attempt or another constitutional change in leadership, the period was truncated so as not to introduce the potential for reverse causality or other threats to inference. Coup attempts themselves have been known to have disastrous effects on economic growth, especially when they descend into more widespread violence, and thus I attempted, wherever possible, to cease observing growth rates in the year prior to a coup attempt.

[6] Sao Tome and Principe’s score was estimated from Cape Verde’s score as the 2 island nations were uninhabited prior to Portuguese colonization and had the same immigration patterns and racial structure imposed on them by the Portuguese.

[7] When other diversity indicators were substituted into Model 1.a, in place of the ELF score—such as Alesina’s or Fearon’s ethnic fractionalization scores or Posner’s politically relevant ethnic groups score (PREG)—they also failed to attain statistical significance and did not meaningfully alter the significance or direction of other variables (there were some fluctuations in magnitude). When these other diversity indicators were incorporated into Model 1.b, the

[8] Adekson, 1976, p.254; Keegan, 1983, p.175-180.

[9] Immediately prior to decolonization, British authorities pushed for the Nigerian military to be national. They believed that responsibility for defense and for the armed forces should be exclusively a federal matter as any alternative arrangement would carry with it the possibility of private armies and civil war. (British National Archives, CO 968/478, Document 2, p.8-9.)

[10] See Cawthra, 2003.

ADVERT

27 September 2017

Angola:Joao Lourenco is the new president

Former defense minister takes over from Jose Eduardo dos Santos, who ruled the oil-rich nation for 38 years.

Lourenco takes over from dos Santos, left, who ruled Angola for almost four decade.

25 September 2017

Military Responses to Constitutional Changes of Leadership in Africa