Path: janda.org/c10 > Syllabus > Outline > Topics and Readings >
Modeling Continuous Variables > Multiple Regression
Research Applications

Wells: Multiple Regression

The Coup d'Etat in Theory and Practice:
Independent Black Africa in the l960s
Alan Wells, American Journal of Sociology, 79 (4), 871-887.
Abstract
Attempts have been made to explain coups d'etat by reference to macro social and economic variables, structural features of the military in general, and, more commonly, of military elites, or by reference to the individual psychological qualities of the coup participants. Such theories are often held to be mutually exclusive and competing. This paper examines these theories and then tests them against data for 35 independent countries in sub-Sahara Africa, a region for which it has been widely argued that only "micro" explanations of coup activity are possible. Each country is first assigned a score on a coup activity scale which provides an interval measure of the dependent variable. Colonial background, contagion, and temporal hypotheses do not explain differences in coup activity. The data are then subjected to multiple-regression analysis using sets of socioeconomic and military variables. Although no single independent variable is conspicuously related to coups, multivariate analysis does explain some variance in coup activity. An explanation of coup activity that integrates the three levels usually employed by social scientists is suggested. This takes into account societal conditions conducive to coups, the relative dominance of military institutions, and, for residual variance, the personal characteristics of military officers and external influences on them.

By coup d'etat I mean the forceful seizure of the machinery of state governments, a common phenomenon of history and of the contemporary world. The high frequency of such political action, however, has inspired surprisingly little theoretical work on the coup as a generic social act, or on the social structures in which it is likely to occur. Historicism, it seems, has been the dominant mode of analysis.

Early modern theoreticians usually concentrated on the mechanism of the coup rather than on social conditions. . . . Thus Janowitz (1968) claims that "militarism in the new nations of Africa and Asia is often reactive or unanticipated because of the weakness of civilian institutions and the breakdown of parliamentary forms of government" (p. 28). In a similar vein Bienen (1968b) claims that "when civilian governments are discredited, the military may be able to intervene without force because it does retain legitimacy untainted by civilian failures" (p. 38). Janowitz (1968), Bienen (1968b), LeVine (1967), Andreski (1968), and others have specified some of the social and economic factors that lead to these civilian failures. The result is a series of rather loose theories of the coup, employing variables of various levels of abstraction and immediacy to the coup itself. The purpose of this paper is to test the efficacy of such theories to the empirical realities of Africa in the past decade.

THE AFRICAN CASE

Just a few years ago, it was fashionable to view Africa with optimism and to discount the possibility of military intervention in politics. . . . The commonly held view now is that all of black-ruled sub-Sahara Africa is prone to military coups (Bell 1968, p. 272) and that coups are natural and endemic to the continent (Andreski 1968, pp 208-9). . . .

The reason for this switch in imagery stems from the African experience of the 1960s. The decade was marked by widespread violence, including the continuous unrest in the Congo and Sudan, the civil war in Nigeria, and the border wars fought by the Somalis (Matthews 1970). During the 1960s there were 27 successful coups in 13 countries in sub-Sahara Africa. Nine other countries experienced abortive coups d'etat, while only seven countries reported no coup activity during the decade . . . .

The belief in Africa's proneness to coups d'etat has led many commentators to overlook social factors in their explanations. Zolberg (1968), for example, relies on specific circumstantial and current factors. Without actually attempting an analysis, he claims that it is "impossible to specify variables which distinguish as a class countries where coups have occurred from others which have so far been spared" (p. 71). Janowitz (1964) concurs with this skepticism (p. 73). . . . Certainly unicausal explanations can be rejected by a light perusal of the African data, but even simple multivariate relationships would demand mental gymnastics going far beyond the running of mental checklists. The task here, then, is to systematically investigate how theories of the coup apply to the African continent; that is, to determine if social and economic variables can discriminate between those countries that have had coups or high coup activity and those that have not. Three common univariate explanations of coup activity are first investigated. The remaining analysis employs variables relating to socioeconomic conditions and military structures. Since no single variable can explain (differences within the sample of the 32 independent African countries, the analysis uses multiple-regression techniques to determine combined effects of these variables.

Three Common Explanations:
Colonial Background, Contagion, and Temporal Theories

The colonial background of African countries has often been argued to be a crucial factor in coup activity. This argument is usually based on the different political systems that have been inherited by the African countries from France, Britain, or Belgium (see, for example, Coleman and Brice .1962, pp. 364 if.). . . .

Examination of the evidence, however, provides no support of this thesis. By the end of 1969 five out of 14 former British colonies had had successful coups, compared with six of 15 former French colonies. Two of the ex-Belgian states fell to military coups, but with a sample size of only three, this can hardly be taken to indicate that Belgian territories are peculiarly vulnerable. When scores on a coup activity index (used in the multiple- regression analyses which follows) are dichotomized, eight of the 14 former British territories rate high as against nine of 15 of the French. Thus the difference between the French and British traditions does not appear to be directly related to military intervention.

Contagion theories posit that the occurrence of a coup in one country stimulates those in other countries, especially neighboring ones. This idea is implicit in the work of Bell (1968), who talks of two waves of military intervention in the political affairs of Africa. The first, he notes, began in December 1962 and lasted for 14 months. The second began in November 1965 and continued to February 1966. . . .

The evidence in support of the contagion thesis cannot be denied. . . . The contagion thesis, however, is insufficient to explain all coup behavior. It cannot predict when and where an initial wave of coups will begin, nor which countries will follow the example of the first.

Temporal theories rest on the assumption that most newly independent countries are equally vulnerable to coups. Once independence is granted, it is only a matter of time before they have one. Janowitz (1964) has reported weak support for this thesis in a sample of new nations which excluded those dealt with here (p. 16). . . . Civilian rule, then, is only an interlude separating independence and military intervention. Its duration, according to Scott (1971), is about five years in Africa (p. 39).

by the experience of Africa, although the five-year claim is not very compelling. None of the five countries with less than six years of independence at the end of 1969 had experiThe general temporal thesis seems to be supportedenced coups. Among those that were independent for between six and nine years, two of the eight had coups. The remaining 22 countries had all been independent for 10 or more years at the end of the decade, and 11 had experienced a military takeover. The latter group, however, indicates that perhaps not all African countries are equally subject to military coups. The thesis is therefore generally valid. But as a single explanatory variable, time is not very discriminating: additional variables are clearly needed if a more exacting explanation of coup activity is sought.

A Multivariate Approach to Coup Activity

The literature on coups d'etat yields two main classes of explanatory variables in addition to those dealt with above. The first set relates to general socioeconomic conditions, the second to military structures. Selected variables in both classes, together with one extrasocietal variable, are utilized in the analysis that follows. The rationales for each of the variables and their indicators are described below.

Proneness to coups. --- This is the dependent variable in the analysis. It is the propensity of a country to fall victim to coups d'etat. This vulnerability, as Andreski has noted (1968, p. 209), seldom ends with the first military intervention: extensive--that is, revolutionary--changes seem necessary to forestall further coup activity. This "proneness" implies a continuous variable. Thus, countries that have experienced a coup are not necessarily equally prone to repeated activity. Similarly, countries with legally constituted governments are not equally immune to forceful intervention. The indicator used for Africa was a coup rate index shown in table 1. 

TABLE 1: Coup Activity in Independent Black Africa During the 1960s

Occurrences of Coup Activity
Coup Activity Index (10A+3B+C)
Years Independent
Rate Index ((1+a)b)

A
(Attempts)
B
(Aborted)
C
(Plots)
(a)
(b)

(c)

Burundi
2
1
2
25
8

3.25

Cameroon

0
0
10

0.1

Cent. Afr. Rep.
2

2
12
10

1.3

Chad

0
10

0.1

Congo Brazzaviile
3
3
2
41
10

4.2

Congo Kinshassa
2
4
5
37
10

3.8

Dahomey
5

50

5.1

Ethiopia

2
1
7
10

0.8

Gabon

1

3
10

0.4

Gambia

0
5

0.2

Ghana
1
1
1
14
10

1.5

Guinea

2
2
10

0.3

Ivory Coast

1
1
10

0.2

Kenya

1

3
7

0.57

Liberia

2
2
10

0.3

Malagasy Republic

0
10

0.1

Malawi

0
6

0.17

Mali
1

10
10

1.1

Mauritania

1

3
10

0.4

Niger

1
1
10

0.2

Nigeria
4

2
42
10

4.3

Rwanda

0
8

0.13

Senegal

1

3
10

0.4

Sierra Leone
2

1
21
9

2.44

Somali Republic
1
1

13
10

1.4

Sudan
2
0
2
22
10

2.3

Tanzania

1
2
5
9

0.67

Togo
2
2

26
10

2.7

Uganda

1
3
6
8

0.88

Upper Volta
1

10
10

1.1

Zambia

2
2
6

0.5
NOTE.Botswana, Equatorial Guinea, Lesotho, and Swaziland are excluded. Adjustment for war-induced changes in military government in the eastern region.

In addition to successful coups, this takes into account both abortive coups and publicly revealed plots to intervene. Information on such activity was systematically gathered from the New York Times and cross-checked with other sources cited in this paper. A coup activity score was computed by assigning a value of 10 points to a successful coup 3 to an overt attempt which failed, and 1 point to an uncovered plot. To yield the index, the total score for each country was divided by the number of years of independence. This adjusts the low scores of newly independent countries which could not exhibit coup activity under colonial occupation, and in so doing controls for the temporal effects already discussed.

Socioeconomic Variables

It is widely recognized that to forestall military coups some degree of economic and social progress is necessary. . . . Nelkin (1967), perhaps, makes the strongest claim for the explanatory power of such variables when he states that "in every country the issues which best account for the ease of military access to power, relate to economic circumstances and their social consequences" (p. 131). The following are the specific variables of this type chosen for the multivariate analysis:

Population size--This variable may relate to coups in several ways. Countries with small populations may be easier to govern successfully than large ones, but their scale may facilitate intervention. In Africa, ethnic diversity tends to be highest in heavily populated countries, and this may magnify the problems of rulership and increase the possibility of coups d'etat. The indicator used for this variable was the estimated population for 1968 (U.N. 1968).

Population growth--High growth rates put pressure on resources within the country and lead to disadvantageous dependency ratios. Under these conditions widespread economic well-being is extremely difficult to achieve, and political instability and coups may result. The indicator used for this variable was the percentage annual rate of increase of the population for the mid 1960s (U.N. 1967).

Urbanization--It is often assumed that urbanizing societies are experiencing "modernization" and in the process building an immunity to coups. . . . The indicator used was the percentage of the population in towns larger than 100,000 people (Hance 1970, table 16).

Centrality--According to Luttwak (1969), states with a clear political center are the easiest to seize (p. 45). Measures of economic and transport concentration are included in the chosen indicator, a slight modification of an index developed by William A. Hance (1970, p. 234). Two points are scored if the capital city is the only major city in the country, and I if the capital is the most populous city, first port, or first river port; if it has the main airport, rail center, or road focus; and if it is the main manufacturing and administrative center.

Literacy--It may be hypothesized that an unlettered population is a breeding ground for military intervention. A poorly informed public is unlikely to oppose a coup. Conversely, a highly literate public is more likely to express its discontent with civilian rule and give encouragement to military plotters. The percentage of the population claimed to be literate, reported in the Percy and Stoneman (1968) was the chosen indicator.

Mass-media availability --One of the first actions of a successful coup d'etat is the seizure of broadcasting facilities. The more receiver sets in the country the easier it is for the new regime to consolidate its victory by rapid dissemination of news. This only holds true if the broadcasting facilities are relatively concentrated and accessible to seizure. If they are not taken they may be a powerful source of opposition. The indicator used here is the number of radio sets per 100 population (U.N. 1967).

Economic level--Economic well-being is usually claimed to be a factor in coup activity. Thus Luttwak (1969) has argued that "economic backwardness" is conductive to coups (pp. 18-25), while Miller (1970) adds that coups are "easier if the economy is small and fragile" (p. 16). By world standards the countries in this sample rate very low on most indices of wealth. In 1963 only Gabon Ghana, the Ivory Coast, and Senegal had GN1~ per capita of more than $200 U.S. In global perspective, then, all of the African countries are prime coup targets. The variable is included in the analysis to test its effect within a low wealth stratum. The measure used here was GNP per capita in 1963 (U.N. 1967, p. 582).

Economic growth--A growing economy is an important sign of effective government. If a high growth rate is maintained, public discontent and the possibility of violent intervention are likely to be low. The growth rate was calculated from GNP figures for 1963 and 1965 listed in the U.N. Yearbook of National Accounts Statistics (1969, vol. 2).

Military Variables

Since the military must take action to bring about a coup d'etat, it is often argued that certain structural factors of the military may be conducive to coups. According to Miller (1970), the military may intervene if it feels threatened "as an institution" by civilian rulers (p. 16). Similarly Bienen (1968b) has argued that interest group demands may be a motivating factor, as they were in the East African mutinies in the sixties (p. 38). . . .

Several commentators on civil-military relations in less developed countries attributed coups to the modern characteristics of military organizations. Unlike civilian institutions in such countries, the military are usually national in scope. For these reasons, Gutteridge (1965) claims that they may be "the only group with the know-how to run an advanced society" (p. 10). . . .The variables chosen to represent these structural factors are outlined below.

Military participation ratio--This has been defined by Andreski (1968) as "the proportion of militarily utilized individuals in the total population" (p. 33). He claims that it is directly related to the balance of power between civilians and the military, so it should be a factor in coup activity. The indicator used here was the percentage of men of military age on active service in the armed forces (Booth 1970, p. 27).

Military size--Despite the protestations of commentators denying its importance (Scott 1971, p. 39), the absolute size of the military may be a crucial factor in its potential for effecting a military coup. . . . The total number of men serving in the regular armed forces (Booth 1970, p. 27) was employed as a measure of the variable.

Size of the police Force.--Gutteridge (1965, p. 65) and Janowitz (1964, pp. 37-38) have suggested that the police may be a factor in coups d'etat. . . . The indicator was the number of men in the police and paramilitary (e.g., the Gendarmerie in former French Africa) services. The figures used, like those for the military, are estimates for 1970, the only recent year for which full African data were available (Booth 1970, p. 27). .

The defense budget--Defense expenditure gives an estimate of the scale of military operations and influence in a country. The measure used was the estimated defense budget in thousands of U.S. dollars for 1968 (Grundy 1970, p. 26). . . .

Defense expenditues and the economy--This variable concerns the priority of military over civilian spending. Two indicators were employed: the percentage of the total government budget and the percentage of the estimated GNP in 1968 spent on defense (Booth 1970, p. 26).

External Factors

It is often argued that external effects can influence military coups. The general ideological position of the United States, for example, seems tolerant of military intervention. Thus according to Bienen (1968a), "Many American government officials, military leaders and academicians believe that the military in the developing areas is the best counterforce against both internal and external disruption . . . something of a guarantee against the sort of chaos which would ultimately make development, hard as it is now, almost impossible" (p. xiii). . . .

On the other hand, it has been argued by Lutwak (1969) that "the target state must be substantially independent and the influence of foreign powers in its internal political life must be relatively limited" (p. 32). . . .

Such outside influences are of course extremely difficult to measure quantitatively across the sample of countries dealt with here. The chosen indicator relates only to United States aid (and hence perhaps influence) in the African countries. The data used were World Bank-IDA figures for the cumulative total of loans and credits to the African countries as of June 1968 in millions of dollars. . . .

CORRELATION ANALYSIS AND FINDINGS

The sample consisted of 31 black African countries. Four countries-- Botswana, Equatorial Guinea, Lesotho, and Swaziland--were excluded because of incomplete data. All are anomalous for the region and might also have been excluded on the basis of their recent independence and continued domination by white-ruled countries. First, the analysis investigates zero-order correlations of the two sets of variables and coup activity; then it examines the multivariate relationship of the dependent and independent variables. The covariation of socioeconomic variables and coup activity is presented in table 2.

TABLE 2: Correlation Matrix of Socioeconomic Variables and Coup Activity Index (9)

2
3
4
5
6
7
8
9

1

Population
-0.03
-0.18
-0.06
0.11
-0.14
-0.24
-0.12
-0.06

2

Population growth

0.18
0.16
0.07
-0.29
-0.26
-0.07
-0.06

3

Urbanization

0.43
0.02
0.06
0.38
0.06
-0.06

4

Centrality

-0.20
0.10
0.17
0.10
0.03

5

Literacy

0.16
-0.06
-0.23
0.22

6

Mass media

0.64
0.01
-0.24

7

Economic level

0.09
-0.02

8

Economic growth

0.01

The socioeconomic variables, with the exception of mass-media availability and economic level, are not strongly interrelated. Literacy and mass media emerge as the strongest predictors of coup activity when there are no controls for other independent variables.

Table 3 shows the correlation of military variables and military intervention. United States aid was included as a variable here because, as the table indicates, it is strongly related to three of the military variables. It appears that large standing armies are being underwritten by U.S. aid; only the relation between defense budget and the size of the military (as would surely be expected) yields a higher correlation. Zero-order correlations between the dependent and independent variables are higher than those for the socioeconomic variables due to their closer interrelationship and, perhaps, immediacy to military intervention.
TABLE 3: Correlation Matrix of Military Variables and Coup Activity Index (8)

2
3
4
5
6
7
8

1

Military participation ratio (MRP)
0.22
0.27
0.36
0.28
0.39
0.37
0.25

2

Size of military

0.63
0.85
0.43
0.07
0.76
0.36

3

Size of police force

0.76
0.20
0.080
0.80
0.24

4

Defense budget

0.42
0.20
0.88
0.37

5

Military $ / government $

0.50
0.27
0.36

6

Defense expenditure / GNP

0.12
0.32

7

U.S. Aid

0.32
The multivariate relationships between coups and the above sets of variables were investigated in the stepwise regression analyses summarized in table 4. The first column shows the effects of socioeconomic variables alone, the second of military variables, and the third of all of the variables on coup activity. The ß coefficients are interpreted as unstandardized partial correlations and indicate the contribution of each variable to the total explained variance in the dependent variable

TABLE 4: ß Coefficients and Multiple Correlation in Three Analyses of Coup Activity

Analysis I
Analysis II
Analysis III

Population
-0.01

0.07

Population growth
-0.15

0.21

Urbanization
-0.08

0.04

Centrality
0.06

0.21

Literacy
0.02

0.05

Mass media
-0.32

-0.52

Economic level
0.01

-0.52

Economic growth

-0.02

Military participation ratio (MPR)

0.16
2.06

Size of military

0.01
0.02

Size of police

-0.01
0.07

Defense budget

*
-0.05

Defense $ / government $

0.03
0.08

Defense budget / GNP

0.26
0.23

U.S. aid

0.01
-0.02

Multiple correlation R
0.41
0.48
0.75
*Variable excluded for insufficient F-level.

The socioeconomic variables alone yielded a multiple R of .41 and accounted for a little more than 16% of coup variance. Mass media and population growth emerge as the strongest (negative) predictors of military intervention. The military variables in the second analysis explain a little more variation, about 23%, with MPR and defense budget/GNP the strongest predictors. Centrality and the predominant variables in the first two analyses emerge the strongest predictors of coups when all independent variables are included. Together they explain 56% of the variation in coup activity. Even for Africa, therefore, such variables should not be discounted out of hand.

CONCLUSIONS

On a simplistic level, a coup occurs because a group of officers decides to stage one. They may be reacting to a long- or short-term crisis, or perhaps none at all. They may have plotted extensively or, as Miller (1970) argues for Africa, they may fall into situations that are "totally unstructured a series of happenstances when decisions by a man pushing his own interests at a certain point in time may be critical" (p. 17). Such explanations are appealing, but the analysis presented above indicates that more general variables also have a bearing on coup activity. . . .

Whether or not a section of the military elite can isolate itself sufficiently to plot a coup depends in part on internal conditions (such as cohort rivalry) in the officer corps. But their motivation to intervene and the opposition that they can anticipate stems from less immediate factors. First the plotters may be subject to foreign influences: contagion effects were noted for the African sample, and foreign moral and perhaps tactical support may encourage them. The elite are also influenced by the overall well-being of the defense establishment (which may, as in Africa, be closely associated with foreign aid), and the l)power it affords them. Their action may be predicated on the failure of civilian politicians, which is often the outcome of socioeconomic deterioration Socioeconomic factors may also influence the military elite directly or through their impact on military structures (heavy demands of the latter might also contribute to socioeconomic malaise). . . . 

The inclusion of socioeconomic and military variables in the explanatory model is justified on theoretical and empirical grounds. Further use of the model, it is hoped, will show that it usefully portrays coups as complex, but not unexplainable, social phenomena.

 
REFERENCES

Andreski, Stanislav. 1968. Military Organization and Society. 2d ed. Berkeley: University of California Press.

Bell, M. J. V. 1968. "The Military in the New States of Africa." In Armed Forces and Society: Sociological Essays, edited by Jacque van Doom. The Hague: Mouton.

Bienen, Henry. 1968a. The Military Intervenes: Case Studies in Political Development. New York: Russell Sage Foundation.

Bienen, Henry. 1968b. "Public Order and the Military in Africa: Mutinies in Kenya, Uganda, and Tanganyika." In The Military Intervenes: Case Studies in Political Development, edited by Henry Bienen. New York: Russell Sage Foundation.

Booth, Richard. 1970. "The Armed Forces of African States, 1970." Adelphi Papers, no. 67 (May).

Coleman, James S., and Belmont Brice, Jr. 1962. "The Role of the Military in Sub- Sahara Africa." In The Role of the Military in Underdeveloped Countries, edited by John Johnson. Princeton, N.J.: Princeton University Press.

Foltz, William 3. 1966. "Building the Newest Nations." In Nation Building, edited by Karl W. Deutsch and William 3. Foltz. New York: Atherton.

Goodspeed, Donald J. 1962. The Conspirators: A Study of the Coup d'Etat. New York: Viking.

Grundy, Kenneth W. 1968. Conflicting Images of the Military in Africa. Nairobi: East Africa Publishing House.

Gutteridge, William. 1965. Military Institutions and Power in the New Stales. New York: Praeger.

Hance, William A. 1970. Population Migration a,and Urbanization ion in Africa. New York: Columbia University Press.

Janowitz, Morris, 1964. The Military in the Political Development of New Nations. Chicago: University of Chicago Press.

Janowitz, Morris, 1968. "Armed Forces and Society: A World Perspective." In Armed Forces and Society: Sociological Essays, edited by Jacque van Doom. The Hague: Mouton.

Johnson, John 3., ed. 1962. The Role of the Military in Underdeveloped Countries Princeton, N.J.: Princeton University Press.

Lee, J. M. 1969. African Armies and Civil Order. New York: Praeger, for the Institute for Strategic Studies.

LeVine, Donald N. 1968. "The Military in Ethiopian Politics: Capabilities and Constraints." In The Military Intervenes: Case Studies an Political Development, edited by Henry Bienen. New York: Russell Sage Foundation.

LeVine, Victor T. 1967. "Independent Africa in Trouble." Transaction 4 no. 8 (July- August): 53-62.

Luttwak, Edward. 1969. Coup d'Etat: A Practical Handbook. Greenwich, Conn.: Fawcett.

Malaparte. Curzio. 1932, Coup d'Etat: The Technique of Revolution. Translated by Sylvia Saunders. New York: Dutton.

Matthews, Robert 0. 1970. "Domestic and Inter-State Conflict in Africa." International Journal 25 (3): 459-85.

Mazrui, Ali A. 1969. "Militarism and Political Militancy in Tanzania." In The Military Profession and Military Regime, edited by Jacque van Doom. The Hague.Mouton.

Miller, Norman N. 1970. "Military Coup in Uganda." American Universities Field Staff. East Africa Series, vol. 10, no. 3.

Nelkin, Dorothy. 1967. "The Economic and Social Setting of Military Takeovers in Africa." Journal of Asian and African Studies 2, nos. 3-4 (July-October): 230-44.

Percy, E. A., and E. A. Stoneman, eels. 1968. handbook of New Nations. New York: Crowell.

Rivkin, Arnold. 1969. Nation-building in Africa. New Brunswick, N.J.: Rutgers University Press.

Scott, Roger. 1971. "The Politics of New States." In The Politics of New States, edited by Roger Scott. New York: Harper Torchbooks.

U.N. 1967. Statistical Yearbook. New York: International Publications Service.

U.N. 1968. Demographic Yearbook. New York: International Publications Service.

U.N. 1969. Yearbook of National I Accounts Statistics. New York: International Publications Service.

Van Dorn, Jacque. 1968. Armed Forces and Society: Sociological Essays. The Hague:Mouton.

1969. "Political Change and the Control of the Military." In The Military Profession and Military Regime. The Hague: Mouton.

Zolberg, Aristide R. 1968. "Military Intervention in the New States of Tropical Africa:Elements of Comparative Analysis." In The Military lntervenes: Case Studies in Political Development edited by Henry Bienen. New York: Russell Sage Foundation.

1969. "Military Role and Political Development in Tropical Africa." In The Military Profession and Military Regime, edited by Jacque van Doom. The Hague: Mouton,