Abstract
Military interventions continue to be prevalent in Africa. In the 21st century alone, 14 coups have been successfully staged. Whereas most studies of coup risk examine how militaries are organized or what structural conditions are associated with coups, we take a novel approach. We explore how coalition politics relate to coup risk. It has long been observed that regimes try to hold power by buying off urban consumers. We argue that focusing on urban consumers actually makes regimes more prone to military intervention. Instead, leaders who ally with established rural elites are more effective at thwarting coups. To test our hypothesis, we develop a unique data set of rural political strategies, coding regimes in 44 sub-Saharan countries from 1960 to 2000. Using a continuous-time Cox proportional hazards regression model, we find a robust correlation between policies supportive of rural elites and lower coup risk.
With the democratic “Third Wave” across Africa, it appeared that the study of coups, like Soviet studies, might soon be relegated to something of a distant age. However, recent events in Burkina Faso, Burundi, and Central African Republic suggest this may not be so. In fact, since the turn of the 21st century, 13 coups have been successfully staged in sub-Saharan Africa. 1 Although scholarship indicates competitive elections are likely to be restored after a coup (Marinov & Goemans, 2014), this does not help us anticipate which regimes are more likely to succumb to a coup or how some leaders manage to avert them.
For a long time, studies of coup-proofing focused primarily on how militaries were organized (Decalo, 1986; First, 1970; Horowitz, 1985; Kandeh, 2004) or on the structural conditions associated with coup risk (Bloom, Sachs, Collier, & Udry, 1998; Decalo, 1986; Londregan & Poole, 1990). But given the similarity of African military organizations and the shared vulnerabilities sub-Saharan states face, such structural approaches gave little means to predict which regimes would be more likely to fall prey to a successful military intervention (Belkin & Schofer, 2003). In recent years, the trend has been to analyze the utility of “counterbalancing,” the practice of dividing military command as a means to protect regimes from coups. Unfortunately, to date, the evidence for the efficacy of counterbalancing on coup-proofing is quite mixed. While several studies find a high correlation between counterbalancing and reduced coup risk (Belkin & Schofer, 2005; Powell, 2012), others do not (Albrecht, 2015; Böhmelt & Pilster, 2015).
Instead of focusing on the way regimes organize their military or what structural conditions accompany coups, we explore how coalition politics relate to coup-proofing. The contention of this article is that because of the weakness of African institutions and the continued predominance of clientelism in politics, how well a regime handles its military is related to its overall strategic vision. We make the counterintuitive claim that leaders who ally with established rural elites and develop multiregional bases of support face lower coup risk than those who cater to urban interests. This is because for a regime to survive, it has to withstand incessant urban unrest, curb ethno-regional competitors, and grow the economy so as to prevent the politicization of the military. These aims are best achieved by fostering rural support. Therefore, despite the obvious fact that urban unrest can destabilize regimes and lead to coups, we posit that military intervention is better averted by promoting local and regional elites.
To test if rural support increases the ability of sub-Saharan regimes to thwart coups, we develop a unique database of rural political strategies. We code regimes in 44 sub-Saharan countries on how effectively they promoted rural political incorporation. We then employ a continuous-time Cox (1972) proportional hazards regression model to test the effect of rural political incorporation on the likelihood of a regime suffering a successful coup. We find a strong and robust correlation between implementing policies supportive of rural elites and lower coup risk.
Rural Versus Urban Political Strategies
In recent years, scholars have begun to recognize the importance of rural politics for regime success in sub-Saharan Africa. New analyses underscore the influence of local rural elites on electoral outcomes (Baldwin, 2014; Koter, 2013). Political economists stress the centrality of agricultural policies and cross-ethnic elite coalitions to economic development (Booth, 2011; Booth & Golooba-Mutebi, 2012; Kelsall, 2013). There is also scholarship on the importance of creating ethnic coalitions for political stability (Arriola, 2009). Yet, none of these analyses link rural coalitions to coup-proofing. Thus far, the relationship between rural politics and coup risk has gone unexplored.
Historically, sub-Saharan scholars and politicians alike held that in order to avoid coups, leaders had to support urban interests. The logic was that, because the largest military garrisons and most politically active groups are located in the capital city, African leaders can be easily imperiled by spikes in urban living costs. It was therefore presumed that political security would best be achieved by keeping urban food costs down and keeping urban youth employed. Throughout the 20th century, many African governments implemented “urban-biased” policies in the hopes of impeding destabilizing urban riots. In addition to focusing a disproportionate percentage of government investment on providing jobs to urban youth and industrial development, African regimes implemented pricing policies that favored urban consumers. As a result, rural farmers were forced to shoulder the financial burden of providing urban areas with low-cost food (Bates, 1981).
Recent history has further underscored the importance of mollifying urban constituencies for regime security. Since the “Arab Spring,” escalating urban riots in the Middle East and Africa has forced a series of long-standing autocrats out of power. Rural areas, on the other hand, appear to pose few hazards to the central government, especially in sub-Saharan Africa where historically the peasantry and rural elites have been disconnected from the state (Hyden, 1980). In fact, south of the Sahara, only a few short-lived peasant rebellions have occurred.
Therefore, the presumption has been that sub-Saharan regimes can disregard agricultural interests and rural majorities with impunity unless they face elections. Even today, although rural voters outnumber their urban counterparts, evidence suggests that regimes continue to skew public resources toward urban sectors (Byerlee, Diao, & Jackson, 2005; Ellis & Bahiigwa, 2003; Rola-Rubzen, Hardaker, & Dillon, 2001).
Certainly, there is plenty of evidence to support the idea that coup risk increases as urban protest intensifies (Londregan & Poole, 1990; O’Kane, 1987). However, what needs interrogating is not whether urban protest can lead to destabilizing conditions that make coups possible, but rather how best to manage this risk. The proposition forwarded here is that developing a rural support base may actually be a more efficacious strategy for reducing the threat posed by urban discontent than catering to urban constituents. With a rural base of support, regimes are less beholden to urban constituents. They are, therefore, less likely to try to appease urban groups and become subject to escalating urban demands that spiral out of control. As Huntington (1968) observed decades ago that, “If the government can build a bridge to the countryside, if it can mobilize support from the rural areas, it can contain and ride out the instabilities of the city”; otherwise, it will be “caught in an urban prison of instability and function at the whim of the city mob, the capital garrison, and the central university’s students” (p. 209).
Urban bias can also hasten economic decline. Squeezing producers to reduce food costs raises the likelihood farmers will choose “exit strategies,” such as cutting out export crops to grow food staples and selling-off products on the black market. Farmers’ exit strategies undercut the state’s productive base and capacity to generate revenue (Bates, 1981). Madagascar in the 1980s provides a typical example of how this has happened. The government favored urban populations by maintaining low prices for staple foods. Peasants reacted by concentrating on food crops and smuggling cash crops to neighboring islands where they received higher producer prices. As a result, “almost seventy-five percent of the 1982 peasant vanilla harvest was smuggled to the Comoro Islands” (Mukonoweshuro, 1990, p. 383). Such policies can have the unintended consequences of reducing urban food supply and raising urban food costs, thus bringing about the very outcome the policies were designed to avert.
Most of all, rural coalitions can be critical to containing ethnic mobilization. Generally, political entrepreneurs mobilize ethnic populations to secure political followings (Mamdani, 1996). The central government can dilute such ethno-regional appeals and weaken regional competitors by promoting local administrations and local elections that channel political competition away from the center (Chinsinga, 2005; Seely, 2001; Widner, 1993). In this way, “localized reform in favor of the peasantry—’empowerment’—consolidates the social base of the regime, and monopoly of power at the top, [and] neutralizes middle-class contenders for leadership from within civil society” (Mamdani, 1996, p. 214). A rural political strategy, thus, can provide a safety valve for political discontent without significantly limiting power at the top.
Rural Coalitions and Coup Politics: Hypothetical Mechanisms
It may be objected that even if a rural political strategy is central to maintaining political and economic stability, why would such a strategy have any bearing on the conduct of the military? In fact, there are several ways in which rural policies can directly impact the armed forces. We outline three central mechanisms.
As stated above, rural neglect often leads to economic hardship. Whereas the top military brass can generally insulate itself from economic recession, the junior and middle-ranking officers, who command troops or battalions that can be efficiently mobilized against a regime, are disproportionately affected by declining living standards and military cutbacks. As Kandeh (2004) explains, The general has more in common with a government minister, bureaucratic or wealthy businessman than with subaltern ranks who are materially closer to wage laborer’s and are culturally lumpen…Senior officers…routinely defend and protect the interests of the political class as whole…by subordinating the army to the predatory agenda of political incumbents. (p. 20)
Frequency of Coups Conducted by Subaltern Versus Senior Ranks.
Unfavorable agricultural policies and coup risk.
More broadly, political legitimacy in Africa depends on a regime’s “resonance with local cultures” (Clark, 2006, p. 133; Schatzberg, 2001). Local politics is central in Africa because clan and lineage groups are still geographically concentrated and chieftaincy remains a dominant sociopolitical institution across the region. Even people living in modernizing cities “retain ties with their rural place of origin and virtually every African village is affected by actions of migrants now resident in urban areas” (Gemmeke, 2008, p. 24). Urban migrants send home critical resources to their natal villages. In turn, villages provide urban dwellers a place of refuge during times of political or economic hardship. Villages are also spiritual and moral centers, where ancestral shrines are maintained. As a result, virtually all urban dwellers have strong ties to specific regions, localities, and chiefs.
For these reasons, supporting chieftaincy and ethno-regional contenders strengthens a regime’s legitimacy (Englebert, 2000). With greater legitimacy, the chance decreases that the military and the general population will deem extralegal measures as justified (Clark, 2006). Thus, on this account as well, the support of chieftaincy and ethno-regional elites can reduce the probability of a coup.
However, the most essential way in which rural support impacts the armed forces is in terms of ethnic politics. Because divisions within sub-Saharan militaries reflect broader societal divisions, “the military can become a hotbed for ethnic resentment and an instrument for the advancement of ethnic claims to power” (Horowitz, 1985, p. 443). In a region where class solidarities and crosscutting cleavages tend to be weak, military cliques, like political parties, are generally formed along ethnic and subethnic ties. The importance of ethnic ties is critical to both subaltern and senior-level ranks (Kandeh, 2004, p. 20). As a conspirator in the 1966 Ghanaian coup puts it, “the officer’s roots are in the countryside, and the outcome of this is that the officer becomes identified almost always with the ‘folk back home’” (Ocran, 1977, p. 64).
The persistence of ethnic ties in the military is in large part a vestige of colonial administrative policies when, commonly, the military rank and file were recruited from relatively disadvantaged regions. These legacies persist in African militaries today. Throughout West Africa, for example, the dominant pattern was for colonial administrations to recruit soldiers from the so-called martial tribes in the relatively impoverished Northern Sahel. The officer corps, in contrast, was recruited from the better-privileged, wealthier southern ethnic groups.
As a result, African security forces can easily be infected by local and ethno-regional discontent.
3
The case of Guinea-Bissau illustrates how the combination of these factors can be translated into a military coup. During the Sahelian drought of 1977–1979, Guinea-Bissau became dependent upon international food aid for the country’s staple food, rice. However, the government distributed the majority of the donated rice to the urban sector. With 80 per cent of the rice normally being grown by the Balanta, their discontent was directly transmitted to the soldiers, the vast majority of whom continued to retain direct ties with their rural homes, so that these economic and ethnic variables merged to further prepare the political stage for military intervention. (Forrest, 1987, p. 102)
Unfavorable rural political policies and coup risk.
Thus, although there is always a complex mixture of motivations behind any seizure of power, in ethnically divided societies, there is quite frequently an ethnic component (Harkness, 2016; Jenkins & Kposowa, 1992). The skewed composition of regimes and officer corps reinforces the apparent importance of ethnicity, facilitates the interactions of officers (including conspiratorial interactions) on an ethnic basis, and may also increase the tendency for politicians and officers to view others’ actions as ethnically motivated. (Horowitz, 1985, p. 471)
Even when the proximate cause of a coup comes from state actions that threaten key members of the military, these threats are often the last of a series attempts made by a regime to neutralize its regional opponents. Thus, this pattern can also be found in personally motivated coups, such as Idi Amin’s overthrow of Milton Obote. Although Decalo (1986) argues, the “more cogent reasons for the Amin coup would be purely personal ones” (pp. 111–112), here too ethno-regional politics was at the center of the events that transpired (Horowitz, 1985; Ingham, 1994; Nugent, 2004). “Obote’s attempt to ensure army loyalty by Langi and Acholi promotions [was] correctly identified [by Amin] as Obote’s attempt to maneuver against him and buildup of the countervailing military force” (Glentworth, 1973, p. 250). Within a year, Amin took charge of the state in a bloodless coup.
As the Amin coup demonstrates, it is often difficult to disentangle personal ambition and corporate military interests from ethno-political fears. More often than not, these concerns are imbricated with one another. Therefore, although not all coups are ethnically motivated, ethnic solidarity frequently plays a central role in coup plotting. This fact coupled with the broader importance of rural areas for political and economic stability means that rural political strategies can function as coup-proofing mechanisms.
In summary, for a regime to survive, it has to grow the economy, curb ethno-regional competitors, withstand incessant urban unrest, and prevent military politicization. These aims are arguably best achieved by fostering rural support. For all these reasons, promoting policies that support local elites and regional populations is critical to political security and regime survival.
New Tool for Analyzing Coup-Proofing Strategies
Accordingly, we propose a new method for analyzing coup-proofing strategies in sub-Saharan Africa. Two central ideal-typical approaches are outlined: a rural political strategy and an urban political strategy (Rabinowitz, 2013). As understood here, a rural political strategy involves a concerted effort to foster the cooperation of populations in diverse regions. This is achieved by allowing traditional chiefs to hold positions of prominence in their communities and by coopting established ethno-regional competitors. Traditional leaders are appointed to official positions in local administrations and/or on local councils and are allowed to oversee diverse local activities, like policing, courts, and development projects. With such policies, chiefs are able to keep positions of respect in their communities and have greater say or leverage over how resources are distributed locally. Established ethno-regional power brokers may also be allotted economically powerful positions in government ministries; in local, regional, and central government administrations; and/or in powerful parastatal organizations and enterprises. In best-case scenarios, regimes implement economic policies favorable to rural elites, such as rural infrastructural projects and pricing policies supportive of producers.
In contrast, an urban political strategy is a defensive posture focused on short-term gains. Leaders try to sustain power by forestalling urban discontent and suppressing ethno-regional competitors. Aside from bolstering the support of their own ethno-regional faction, the primary political and economic focus is on creating jobs for urban youth and keeping urban food costs low (Bates, 1981). 4 These policies end up disadvantaging rural areas in many ways. Regimes implement pricing policies that favor urban consumers over rural producers. Little is done to build rural infrastructure or support farming extension programs. Chiefs are given little control over local development or access to resources. Powerful ethno-regional competitors are excluded from the central government and administrative positions. In their stead, less-established regional representatives may be chosen to hold positions in the party, government, and parastatal organizations. In the end, the neglect of rural communities undermines the economy, undermines multiple regional interests, and destabilizes the state (see Figure 3). Above all, a rural strategy is effective because it is a concerted effort to maintain the quiescence in the periphery and expand the state’s productive capacity through the cooptation of regional and local elites.
Urban versus rural strategy.
However, there are many ways in which this can be achieved. The hypothesis forwarded here is that the greater the number of approaches taken in combination with one another, the more successful a regime will be at creating stability and surviving extra-constitutional removals. Hence, a rural political strategy is understood as an ideal type, along the lines that “democracy” is an ideal type. Just as it is generally held that the more democratic a regime is, the greater the possibility for political stability, economic growth, and so on, in a parallel manner, we are suggesting that the more successful a regime is at approaching an ideal-typical rural strategy, the greater the likelihood it will be able to dampen ethnic mobilization, prevent economic decline, minimize political alienation, and, therefore, reduce coup risk.
Quantitative Study
To test if rural support is related to coup-proofing, we develop a unique data set of rural political incorporation. We code virtually all sub-Saharan African regimes from 1960 to 2000 on how effectively they promoted traditional elites and local autonomy. We then employ a continuous-time Cox (1972) proportional hazards regression model. The sample includes countries of all colonial backgrounds, all regions, as well as of different sizes and resource endowments. Regimes in power for less than 9 months, provisional/transitional governments, and regimes that came to power prior to 1955 were excluded. The final sample size is 135 regimes in 44 sub-Saharan countries (Figure 4). 5
Probability of coup by former colonial power.
The hypothesis tested is that the closer a regime comes to following an ideal-typical rural political strategy, the lower the probability the regime will suffer a forced removal. Our dependent variable is coup risk, measured as the likelihood of a successful forced removal. We define a “successful forced removal” as “the sudden, often violent overthrow of an existing government enforced by the military.” This definition includes any kind of extra-constitutional removal that is backed by the military. However, our definition excludes attempted coups and coup plots. Although several studies include these latter categories, we do not because we are looking at the efficacy of coup-proofing. We take a coup that is foiled, be it an attempted coup or a plotted coup, as an indication of effective coup-proofing.
For our independent variable, each regime is coded on seven measures from which a cumulative “Rural Political Score” (RPS) is created (see Online Appendix). A perfect RPS of 7 represents the ideal-typical rural political strategy. 6 The prediction is that as RPS scores increase, coup risk will decrease. Seven proxies make up the composite RPS index, based on three dimensions: political support base, levels of decentralization, and support of chieftaincy. Coding was based on primary and secondary sources, using an average of 10 sources and a minimum of 4 to ensure that information was triangulated. The mean RPS was 1.87 with a standard deviation of 1.77 (see Figure 5). The measures used are as follows.
Distribution of Rural Political Score.
Political support base
Original base: whether a regime had a predominately rural or urban base upon assuming power. The assumption is that leaders who begin with a rural base of support will generally be more supportive of traditional elites and traditional agriculture than those who begin with an urban base of support and are beholden to union and student activists.
Levels of decentralization
Administrative deconcentration: the state transfers administrative tasks from the central administration to regional “executive agencies,” giving local/regional state functionaries greater discretion over the allocation of resources and development.
Fiscal decentralization: the state designates funds to be used by local administrations at their own discretion.
Political decentralization: the state allows for freely elected councils to administer localities.
Support of chieftaincy
Village administration: the regime officially appoints chiefs to local and/or village administrative positions.
House of chiefs: the creation of a national body of chiefs, either as a House of Chiefs, their inclusion in an Upper House.
Legal recognition of chieftaincy: chieftaincy is explicitly recognized in the constitution.
For “levels of political concentration,” each factor is measured individually, rather than as a continuous variable. This choice was made to reflect the fact that each element (deconcentration, political decentralization, and fiscal decentralization) can be implemented separately or conjointly and each can have separate political impacts. 7
Because coding the strategies of regimes with asymmetrical time frames presents potential problems, each regime is coded on the basis of the policies it enacted during its first 10 years in office. Focusing on the first 10 years makes our coding more consistent and also militates against the possibility that we are measuring the effects of incumbency, rather than the efficacy of the policies themselves. The 10-year cutoff was chosen because it has been found that coup risk significantly decreases after a decade in power (Bienen & van de Walle, 1992; Gates, Hegre, Jones, & Strand, 2006; Sanhueza, 1999, p. 337). This finding is also consistent with our sample. The change in coup likelihood after 10 years strongly suggests that strategic choices made during these least certain years are the most critical to coup-proofing. 8
Against RPS, we test the effects of 11 competing explanatory factors on coup hazard. The first six are based on our codings. First, to see if colonial history was relevant to outcomes, we use a dummy variable for each colonizing power in our sample: Britain, France, Portugal, Belgium, and Other (Spain, South Africa, or U.S. slave resettlement). Because the end of the cold war and increased conditional aid after 1980 changed the political climate on the continent, we also use a dummy for what epoch a regime came to power. Regimes are coded on whether they came into power pre- or post-1980 to roughly coincide with the independence period and the structural adjustment period, respectively. In addition, although scholars rarely take into account the importance of foreign governments on coup politics, there is growing evidence of the active role the United States and French have taken in coup plots in Africa, particularly during the cold war era. 9 As a proxy for this type of neoimperialism, we code each regime on whether it was leftist when it came to power to see if communist-sympathizing regimes faced higher coup risk. Finally, coup risk has been frequently associated with previous coups. We therefore take into account whether prior to the regime coming to power, the country had experienced successful coups.
We also include controls from a number of different sources. To see if coup likelihood is associated with how repressive a regime is, we use a measure of relative political openness from the Polity IV data set. Polity IV’s “polity2” index ranks regimes from −10 to 10 on how politically open is. The index takes into account “Competitiveness of Executive Recruitment,” “Openness of Executive Recruitment,” “Constraints on Chief Executive,” “Regulation of Participation,” and “Competitiveness of Participation.” 10
Ethnic and religious divisions are commonly hypothesized to be associated with political destabilization. To test RPS against these factors, we use Reynal-Querol and Montalvo (2005) coding of ethnic fragmentation and religious polarization. 11 We also test economic factors using African Development Indicators (ADIs). 12 As a partial test of the urban-bias thesis, we use rate of urbanization to see if urban population growth potentially contributes to greater destabilization and coup likelihood. A country’s level of economic prosperity has long been associated with coup risk. Therefore, we include the natural log of gross domestic product (GDP) per capita, to see how it correlates with the increased hazard of forced removal. Lastly, having valuable resources has been found to have a destabilizing effect on developing countries. To test this hypothesis, we include annual data on oil or mineral exports as the percentage of GDP.
The model
Within our model, regimes become at risk of coups on the day they are founded. They remain at risk until they (1) experience a coup; (2) end by means other than a coup, such as an election; and (3) are right-truncated by the end of the data collection window (December 31, 2015). The hazard of regime i in country j experiencing a coup is modeled as:
where h
0(t) is the baseline hazard for time t, RPS
ij
is the Rural Political Score for regime i in country j,
Using the Cox model enables one to model the effect of each variable independently. Therefore, we have been able to measure the ratio of coup hazards in two regimes that are in all respects equal but differ by one unit of RPS. This is given by:
Thus, e α is the relative change in the hazard of a coup associated with a one-unit increase in RPS. The same argument is applied to all variables in the model, substituting the appropriate coefficient in place of α. If a variable’s hazard ratio is greater than one, the risk of a coup is positively correlated with the variable; if it is less than one, the risk of a coup decreases as the variable increases. All estimates presented below are expressed as hazard ratios rather than raw coefficients.
Equation 1 assumes the baseline hazard is the same for all countries. As a robustness check on our primary models, we also estimate a model allowing for country-specific baseline hazards:
In this model, h 0j (t) is the baseline hazard for country j at time t. Note that the country-level variables, Xj , drop out of this model; they are implicitly incorporated in the country-specific baseline hazards, as are any other variables that are constant over time within a given country. Finally, given that regimes are nested within countries, all models presented below are estimated with robust standard errors that allow for the possibility that coup hazards are correlated across regimes within countries.
Additionally, we use the Kaplan–Meier nonparametric estimate of the survivor function. The survivor function documents the proportion of regimes that have not experienced a coup by time t, years since formation before a coup occurs. The Kaplan–Meier nonparametric estimate accounts for censoring, that is, the possibility that some regimes end without a coup, by a democratic transition, death in office, or by reaching the end of the data collection window before a coup occurs.
Results
The Kaplan–Meier results are striking. Figure 6 displays the Kaplan–Meier survivor function by three categories of RPS: (i) low (RPS < 2), (ii) middle (RPS 2 < 4), and (iii) high RPS (RPS 4). At every point in time, those regimes with above average RPS survive at a higher rate and for a longer duration than those within the average or below average RPS. Figure 7 displays the same results for the hazard function. In all the estimates, the higher the RPS, the greater the likelihood a regime survives the hazard of a coup (see Figures 6 and 7).
Survivor function by category of Rural Political Score.
Hazard function by category of Rural Political Score.
Importantly, the Kaplan–Meier survivor estimate validates the use of a rural political strategy as an ideal type. The model illustrates that what is predictive of survival is the higher combination of the components of the rural political strategy, rather than any one factor in isolation.
The effects of RPS on regime duration are equally striking. Figure 8 presents a scatter plot of the correlation between RPS scores and the length of time a regime is in office. In the scatterplot, it is indicated whether a regime ended in a coup through a normal “exit,” such as elections or natural death in office, or whether a regime was still in power when the data collection period ended (right censored).
Effect of Rural Political Score on regime duration (with indication of type of exit).
The pattern is quite distinct. The vast majority of coups occurred within the first 10 years of power and to leaders with RPS of less than 3. Regimes with RPS below 1 have a particularly high rate of coups. In contrast, the vast majority of leaders who did not experience a coup and/or remained in power beyond 10 years had relatively high RPS.
In addition to the Kaplan–Meier survivor estimates, we run eight Cox (1972) proportional hazard models, using “coup” as the hazard measured. While the Cox model is inherently continuous, a number of our variables vary annually. One observation is included for each year or part thereof for each regime, for a total of 1,835 regime years. RPS is significantly correlated with lowered risk of coup across all eight models. Table 2 presents the first four hazard models. Model 1 is the basic model. It shows the effect of RPS on coup hazard only using former colonial power as a control. The hazard ratio is 0.724, suggesting the annual coup hazard declines by 28% for every one-point increase in RPS. The effect is highly significant but does not take into account covariates that may be correlated with RPS.
Determinants of Coups—Cox Proportional Hazards Regression (Models 1–4).
Note. Exponentiated coefficients; t statistics are in parentheses. Standard errors clustered at country level. RPS = Rural Political Score.
*p < .05. **p < .01. ***p < .001.
Model 2 adds regime-specific controls: epoch, ideological orientation, and previous coups. In this second model, as expected, previous coups are highly correlated with coup risk. Each previous coup increases the annual hazard rate for a subsequent coup by 49%. Coming to power during the structural adjustment period (as compared to the independence period) is also highly significant, reducing the likelihood of coup by 64% as compared with regimes that assumed power before 1980. Even with these controls, the effect of RPS is little changed.
Model 3 includes a Polity IV measure of relative political openness. Due to missing data in the Polity IV data set, three regimes in two countries are lost and the total number of regime-year observations drops from 1,835 to 1,764. In the model, the level of democratic openness has no significant effect. On the other hand, the effect of RPS is unaltered. RPS keeps its t score and effect size. Surprisingly, the former colonial power and the ideological orientation of a regime has no significant effect in any of these models.
Model 4 includes the ethnic and religious composition of the country. Again, because of missing data for these variables, another 10 regimes in four countries drop out and the count of regime-year observations drop even further to 1,617. A one percentage point increase in ethnic fragmentation decreases coup hazard by a little more than 1%. However, religious polarization appears to be uncorrelated with coup risk. In this model, the three variables that were formerly significant remain so. The estimated effect of RPS actually increases, with a decline of 30% of coup risk for every one-point increase in RPS. Previous coups and coming to power in the structural adjustment period also remain highly correlated with coup risk. In addition, two formerly insignificant factors become significant in this model. Unexpectedly, regimes that have higher levels of participation and openness are more likely to have a coup, increasing the hazard of a coup by 6.6% with every one-point increase on a scale of democratic openness ranging from −10 to +10. Regimes in countries that were not colonized by the central imperial powers also have an increased risk of coup.
Table 3 presents four additional models: two models that include time-varying ADIs and two country fixed effects models. Model 5 adds GDP per capita and increased urbanization as controls based on World Bank data. With these two additional controls, two more countries drop out, reducing the number to 36 and the number of regime-year observations drops to 1,414. A 10% increase in GDP per capita decreases coup risk by 12%. 13
Determinants of Coups—Cox Proportional Hazards Regression With Development Indicators and Country Fixed Effects (Models 5–8).
Note. Exponentiated coefficients; t statistics are in parentheses. Standard errors clustered at the country level. Pop = population; RPS = Rural Political Score; GDP = gross domestic product.
*p < .05. **p < .01. ***p < .001.
However, urban population growth is not significant. Despite the increased number of controls and decreased number of observations, RPS actually increases both in significance and in effect. RPS is most highly significant in Model 5, in which the hazard ratio drops to .494 suggesting that a one-point change in RPS decreases the coup hazard rate by 51%. Being in a country that has already experienced a coup and ethnic fragmentation remain significant in Model 5. Religious polarization is significant in this model as well, with an effect similar to ethnic fragmentation. Countries that were colonized by Belgium have a lower probability of a successful coup.
The addition of data on oil and mineral revenues in Model 6 drops the number of observations even lower in Model 6, to 1,097, because data on oil and mineral revenue are not available before 1975. Neither of these natural resource variables is statistically significant, and all the variables that were significant in Model 5 remain significant in Model 6, although the size of the effect of per capita GDP is reduced. This may be because African countries that have valuable resources also have relatively high per capita incomes. The change in hazard ratio, therefore, might indicate that the relative prosperity of the population is less important to coup probability than a regime having enough revenue to allocate a sufficient share to its military. However, it is also possible that the difference has to do with the smaller sample included in Model 6 due to missing data on oil and mineral revenue before 1975. Democratic openness returns to significance in Model 6.
To further test the robustness of the RPS score as a predictor of coups, two fixed effects models are run to control for unobserved country-specific correlates. Note that coefficients in this model are only identified based on within-country differences between regimes. This is represented in the hazard model as a proportional shift in the baseline hazard for a given country j by a constant factor for any given time t. Model 7 has almost all the observations but excludes World Bank data; Model 8 includes the World Bank data but has fewer observations.
RPS is significant in both fixed effects models; indeed, it is the only factor that remains significant once country-level fixed effects are included. These findings suggest there is little probability that any unmeasured country-specific factors threaten the validity of the finding about the effectiveness of a rural political strategy in reducing the risk of coups (at least not factors that are relatively constant over time within countries).
In summary, supporting and incorporating local elites appear to substantially reduce the risk of coups. The effect is robust with respect to a variety of model specifications and covariate patterns. All the models consistently and strongly support the thesis of this study.
Conclusion
This study argues that the political incorporation of rural areas in weakly institutionalized states is central to coup-proofing strategies. By analyzing how the center institutionalizes its connection to the periphery, we have sought to show that coalition politics and broader political strategies are critical to coup risk. All the statistical models strongly support our hypothesis that coup-proofing is related to policies directed at the political incorporation of rural areas and ethno-regional elites.
Three decades ago, Horowitz (1985) speculated that the most successful coup-proofing measure would be “the use of political ingenuity to prevent ethnic conflict from reaching the point at which it spills across civil military boundaries” (p. 557). A rural political strategy is just that. A rural strategy serves as a coup-proofing strategy, first and foremost, because incorporating established local and ethno-regional representatives reduces the probability that the regime will antagonize groups in the security forces and provoke them into action. Very simply, when a regime coopts rather than fears competing elites, chances diminish that key groups in the military will become targeted and will, in turn, target the regime.
A regime’s capacity to use rural strategies to stay in power cannot extend indefinitely. It is likely that African presidents who have been in office for decades, such as Museveni, Mugabe, Biya, and Kagame, will be forced out of office as Compaoré recently was in Burkina Faso. However, their ability to lower their risks and stay in power is what is of interest. How did these leaders avoid the continuous threat of military intervention for decades? For this, we argue, one must look to how well they established alliances with representatives of local and regional interests early on in their rule.
Although focused on sub-Saharan Africa, these observations have implications for understanding coup politics in other regions of the world as well. In highly factionalized societies with weakly institutionalized states, where local leaders often control land and resources, symbolically represent their “people,” perform sacred rites and duties, or sanction in-group behavior through traditional courts and hearings, the incorporation of traditional elites is critical to regime survival. The parallels may be strongest in Middle Eastern and South Asian countries where colonialism made deeper inroads and the divisions among ethnic groups followed similar patterns. Analyzing the political strategies regimes adopt vis-à-vis their rural populations may therefore be the key to predicting which regimes will succumb to military interventions and which will be able to survive coup threats in a number of different regions.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplementary material is available online with this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
