Journal of Critical Methodological Studies

Abdul Karim Bangura


Introduction

In an earlier essay, Said and Bangura posited that ethnicity is a major political and human fault line in Africa. For centuries before European colonialism, ethnic allegiance set the boundaries of communal loyalties and the framework in which political dynamics transpired. While the nation-state is relatively new in Africa, being largely a European phenomenon, multi-national groupings have historically been common. Great empires such as Ghana, Mali and Songhay ruled subjugated peoples of different nationalities long before the arrival of the Europeans.1 In a later work, Bangura also asserted that a multiethnic African society, with its concomitant multiligualism, is arguably a richer society than a nation with only one dominant ethnic group. The multiplicity of life-styles and worldly views can make such a nation a more exciting and stimulating place to live.2
Over the years, however, many scholars have overstated the role of ethnicity in explaining Africa’s economic, political and social problems. One issue area in which this is prevalent is military coups d’etat. For example, Jenkins and Kposowa in their two similar studies were quite decided in asserting that of all the variables that have been shown in previous studies to influence military coups d’état in Africa, ethnicity has the strongest influence.3 These scholars’ finding contrasts sharply with those in Bangura’s meta-analysis of the causes of military coups d’état in Africa which showed that pluralism (encompassing ethnicity and cognate attributes) and institutional structure have positive, but small, effects on coups d’état; domestic environment for participation and social mobilization have small, albeit negative, effects on coups d’état; economic condition has a negative and medium effect on coups d’état; and foreign influence has a negative and small effect on coups d’état.4
This essay, which employs a meta-analytic methodology, applies statistical procedures to collections of empirical findings from individual quantitative studies that have examined the connection between ethnicity and military intervention in Africa through coups d’état for the purpose of integrating, synthesizing, and making sense of them. This helps in discovering underlying trends and principles developed from the accumulation and refinement of this body of studies. But before this is accomplished, a combined review of qualitative and quantitative studies on the issue is done first in order to allow for the description of interesting, worthwhile studies that are not included in the statistical analysis. Thus, the rest of the chapter is divided into various sections that (a) deal with the major differences and similarities between studies which are relevant for understanding the findings, and (b) present a statistical analysis of the pertinent quantitative studies that have dealt with ethnicity as a cause of military intervention in Africa through coups d’état.


Comparison of Studies

In this section, an attempt is made to delineate the differences and similarities between studies. This will help the reader to understand those relevant factors that may underlie the different findings in the studies.


Number, Types and Sources

A literature search on ethnicity and military intervention in Africa yielded 44 works. They include books, academic journal articles, monographs, and an unpublished study (an Orkand Corporation report). Each of these media, of course, calls for different ways of presenting findings. For example, whereas a book may be targeted at a wider audience and, thus, its author may soften its technical aspects, a journal article may not sacrifice technical aspects but may have to reduce details because of space limitation.
The studies were written by African, North American, and European researchers. They were also published in these continents. The cultural biases of these investigators may in some way underlie the measurements used and, consequently, the findings in the studies.


Methodologies

The studies that dealt with the connection between ethnicity and military intervention in Africa can be divided into two methodological groups. One of these groups can be referred to as the Qualitative. This group comprises those researchers who emphasize words to generate descriptions of and explanations for the connection between ethnicity and coups d’état. The 31 (70%) qualitative investigations include studies by Baynham (1980),5 Bebler (1973),6 Bennett (1973),7 Bienen (1989),8 David (1985),9 Decalo (1976, 1986, 1990),10 Feit (1968, 1969),11 Higgot and Fuglestad (1975),12 Keil (1979),13 Laitin (1976),14 Laurean (2006),15 Lewis (1972),16 Lloyd (1972),17 Lofchie (1972),18 Luckham (1971),19 Luttwak (1968),20 Nelkin (1967),21 Noah (1970),22 Ocran (1977),23 Okolo (1981),24 Terray (1964),25 Tixier (1966),26 Uphoff and Ottemoeller (1970),27 Warren and Warren (1976),28 Welch, Jr. (1967, 1970, 1986),29 and Welch, Jr. and Smith (1974).30
The other group can be called the Quantitative. This group encompasses those researchers who conducted numerical representations and manipulations of observations in order to describe and explain the connection between ethnicity and military coups d’état in Africa. The 13 (30%) quantitative investigations include the studies by Jackman (1978),31 Jackman and Boyd (1979),32 Jenkins and Kposowa (1990),33 Johnson, Slater and McGowan (1984),34 Kposowa and Jenkins (1993),35 McGowan (1975),36 Morrison and Stevenson (1971, 1972),37 Orkand Corporation (1983),38 Thompson (1975),39 Wayman (1975),40 and Zimmermann (1979a, 1979b).41
Consequently, the results of the preceding studies might have differed due to their different methods of inquiry. Whereas the qualitative studies are basically enumerative, the quantitative studies are more causally oriented. Thus, although the qualitative studies are as important as the quantitative studies, the quantitative studies are methodologically more complex than the qualitative studies.
It may appear, however, as if the difference between the qualitative and quantitative groups is a somewhat artificial dichotomy: each group combines both approaches in its underlying assumptions. This is because the quantitative approach calls for a great deal of qualitative description prior to counting (in order to empirically ground each category) as well as after counting (statistical tendencies have to be interpreted as to what they reveal about causal relations). And the qualitative approach has an implicit notion that ‘more is better’: that is to say, the more instances of a phenomenon to be found, the more a researcher can trust his/her interpretation of an underlying pattern.
A study on the connection between ethnicity and military intervention in Africa which exemplifies the artificiality of the dichotomy between qualitative and quantitative approaches is that of Luckham (1971). He used a simple case study approach to examine plausible relationships between a set of independent variables (internal conflicts among members of the political elite, ethnic rivalry, civil uprising/disobedience, contagion, and self-interests/political motives of coup makers) and military coups d’état in Nigeria. In order to test for whether or not there were significant differences between ethnic compositions in Nigeria, however, Luckham employed the chi-square statistic.
Despite these underlying similarities, the qualitative and quantitative approaches used in the studies investigated here are different in some ways. In addition to some of the more obvious procedural differences (for example, quantitative studies categorize and count occurrences), the two types of approaches differ in their overall orientation toward inquiry: the qualitative focuses more on particularities and the quantitative focuses more on generalities.
Thus, the classification of works as qualitative and quantitative is important for the present chapter because procedures employed in meta-analysis are geared toward quantitative reviews and syntheses of the research literature that address the issues involved. Meta-analysis calls for more technical and statistical approaches as opposed to unscientific and impressionistic ones. Nevertheless, although qualitative studies are not included in the statistical analysis, they are discussed in various sections of the present chapter to help in comparing the underlying trends and principles developed in the large body of studies on military intervention in Africa.


Research Designs

A number of different research designs were employed by researchers in their examinations of the connection between ethnicity and military intervention in Africa. All of the investigators listed under the rubric of the qualitative approach naturally used the Case Study design. These researchers looked at the issue in considerable detail, typically using personal interviews, observation, and document analysis as data collection procedures. Investigators who used the Case Study design undoubtedly contribute to the cumulative knowledge on military intervention in Africa in the sense that they were able to mix the specific and the general, the peculiar and the typical, the descriptive and the thematic. Like all case studies, however, these studies are also atheoretical: they move in a theoretical vacuum. Their direct theoretical value is, thus, nil. But this does not mean that they are altogether useless.
Investigators listed under the rubric of the quantitative approach attempted to infer causal relationships from a variety of Pooled Cross-Sectional Non-Experimental research designs. The primary difference in these designs lies in the way the statistical and temporal dimensions are employed. A total of six different designs can be discerned: (1) Simple Pooled Cross-Sectional, (2) Pooled Cross-Sectional with Percent Scores, (3) Pooled Cross-Sectional with Regression Analysis, (4) Pooled Cross-Sectional with Reciprocal (Path) Analysis, (5) Pooled Time-lagged Cross-Sectional with Regression and Discriminant Analysis, and (6) Pooled Cross-Sectional Design with a Time-lag between the Exogenous Variables and the Coup Events.
A Simple Pooled Cross-Sectional design combines measures on variables for a number of countries for different years into a single analysis using correlational statistics. This design was employed by Morrison and Stevenson (1972) and Wayman (1975). A major limitation with this design is that it cannot be used to measure change over time in any of the variables.
A Pooled Cross-Sectional with Percent Scores design combines measures on variables for a number of countries for different years into a single analysis by using percentages to standardize frequency distributions. This design was used by Thompson (1975). Like the simple pooled cross-sectional design, the pooled cross-sectional with percent scores design does not permit the measurement of change over time in any of the variables. In addition, the percent does not necessarily express what it claims to do. While nothing is wrong with the percent itself, conclusions drawn from using it can be unwarranted for one reason or another.
The Pooled Cross-Sectional with Regression Analysis design involves combining measures on variables for a number of countries for different years into a single analysis by using regression techniques to specify the nature of plausible relationships between the independent (and control) variables and the dependent variable(s). Investigators who used this design include Jackman (1978), Jackman and Boyd (1979), Johnson, Slater and McGowan (1984), Orkand Corporation (1983), and Zimmermann (1979a, 1979b). Like the previous designs, the pooled cross-sectional with regression analysis design does not allow a researcher to measure change over time in any of the variables. While it is relatively stronger than the previous designs, it has two shortcomings when used to predict data (as is the case in predicting the causes of military coups d’état across many countries in Africa): (1) There is the problem toward the mean. This means that in a randomly distributed measure of military coups d’état, predictions from extreme values can be expected to be less extreme. (2) Regression cannot legitimately be used to predict beyond the range of the data.
Using a Pooled Cross-Sectional with Reciprocal (Path) Analysis design entails combining measures on variables for a number of countries for different years into a single analysis by using reciprocal techniques to estimate the magnitude of linkages between variables and using these estimates to provide information about the underlying causal processes. This design was used by Morrison and Stevenson (1972). This design, like those discussed above, does not permit the measurement of change over time in any of the variables. Although this design has the advantage of allowing a researcher to measure both the direct and indirect effects that one variable has on another, which would identify the operative causal mechanisms and allow him or her to decompose the correlation between any two variables into a sum of simple and compound paths, it also has at least one major problem: inconsistencies arise and the choice of which solution to keep remains. This problem is often referred to as the “over-identification problem.”
A Pooled Time-lagged, Cross-Sectional with Regression and Discriminant Analysis design combines measures on variables for a number of countries for different years using regression techniques that relate current endogenous variables to past values of the exogenous and/or endogenous variables, and using discriminant techniques that form a linear function of the independent (and control) variables which maximizes the correct classification between countries which have had coups d’état and those which have not. This design was employed by McGowan (1975). The design, which also permits a researcher to measure change over time in the variables, is, nevertheless, marred by the same problems inherent in the pooled cross-sectional and time-series design. The major difference between the two designs is that while Ordinary Least Squares (OLS) and Generalized Least Squares (GLS) can be used to generate unbiased, relatively efficient, and consistent estimates in a pooled time-lagged, cross-sectional with regression and discriminant analysis design, only GLS can be used in a pooled cross-sectional and time-series design. This is due to the fact that the latter is a special case in which the error variances of concurrent cross-sections may be unequal whereas their covariances equal zero, and error variances of successive cross-sections are equal whereas the covariances are of the same structure and are also time dependent.
The Pooled Cross-Sectional Design with a Time-lag between the Exogenous Variables and Coup Events was employed by Jenkins and Kposowa (1990) and Kposowa and Jenkins (1993). This design helps a researcher to explain the structural likelihood of a phenomenon. While this approach produces consistent and asymptotically efficient estimates under conditions where OLS fails, it is more an art than a science. If a model fails to converge, the failure may hinge upon any number of parts of the iterative process.
In sum, the studies that have dealt with the connection between ethnicity and military intervention in Africa through coups d’état used a variety of research designs. The biases in the use of these different designs could have influenced the results of these studies.


Statistical Estimators

Expectedly, only those investigators who employed quantitative methodology used statistical estimators for the variables they utilized. The exception here is Zimmermann (1979a, 1979b) who reported other investigators’ statistical estimators but presented none for his own causal model. While some researchers used single statistical estimators, others employed multiple estimators.
Partial Correlation Coefficient was used by Kposowa and Jenkins (1993), McGowan (1975), Morrison and Stevenson (1971, 1972), and Wayman (1975). Standard Regression or Beta Coefficient was employed by Jackman (1978), Jackman and Boyd (1979), Johnson, Slater and McGowan (1984), McGowan (1975), Morrison and Stevenson (1972), and Orkand Corporation (1983). Parameter Estimate or Ratio was used by Jackman (1978) and Jackman and Boyd (1979). Coefficient of Determination was employed by Jackman (1978), Jenkins and Kposowa (1990), Johnson, Slater and McGowan (1984), McGowan (1975), Orkand Corporation (1983), and Wayman (1975). Path Coefficient was used by Morrison and Stevenson (1971, 1972). F Ratio was used by Johnson, Slater and McGowan (1984) and Orkand Corporation (1983). Percentile was utilized by Thompson (1975). Maximum Likelihood Estimates were used by Jenkins and Kposowa (1990) and Kposowa and Jenkins (1993).
These studies, then, utilized a great variety of statistical estimators. Again, it is not farfetched to expect, therefore, that the findings of some of these studies differ because of these different estimators.


Time Periods

A variety of time periods were utilized in the studies. These time periods range from the 1940s to the 2000s. While some researchers employed single years in their analyses, others used multiple years. For the sake of brevity, these time periods and the attendant studies have been segmented into 13 categories.

1. Single Years, 1960s. Research in this category include the studies of Bienen (1968), Feit (1968), Keil (1970), Laitin (1976), Lewis (1976), Lloyd (1972), Luckham (1971), Noah (1970) and Terray (1964).
2. Single Years, 1970s. Investigations in this group comprise those of Baynham (1980), Bennett (1975, 1980), Higgott and Fuglestad (1975), Lofchie (1972) and Warren and Warren (1976).
3. Single Years, 1980s. The one study here is by Okolo (1981).
4. Multiple Years, 1940s-1960s. The single study here is by Luttwak (1968).
5. Multiple Years, 1940s-1970s. Works in this group include those by Thompson (1975) and Zimmermann (1979a, 1979b).
6. Multiple Years, 1950s-1960s. Research in this category comprises those of Morrison and Stevenson (1971, 1972), Welch, Jr. (1967, 1970) and Welch, Jr. and Smith (1974).
7. Multiple Years, 1950s-1970s. The study here is by Morrison and Stevenson (1972).
8. Multiple Years, 1950s-1980s. The following studies can be placed here: Jenkins and Kposowa (1990) and Kposowa and Jenkins (1993).
9. Multiple Years, 1960s. The studies in this group comprise those of McGowan (1975), Nelkin (1967), Uphoff and Ottemoeller (1970) and Wayman (1975).
10. Multiple Years, 1960s-1970s. Research in this category include those of Bebler (1973), Bennett (1973), Bienen (1978), Decalo (1976, 1986), Jackman (1978), Jackman and Boyd (1979) and Ocran (1977).
11. Multiple Years, 1960s-1980s. This category entails the studies of David (1985), Decalo (1990), Johnson, Slater and McGowan (1984), Orkand Corporation (1983), Owusu (1986), and Welch (1986).
12. Multiple Years, 1970s-1980s. The only study here is that of Bienen (1989).
13. Multiple Years, 1950s-2000s. The one study here is by Laurean (2006).

Clearly evident from the preceding categories is that while all the studies identified earlier as quantitative used multiple years in their analyses, those identified as qualitative used either single years or multiple years in their analyses. The great variation in the time periods analyzed in the studies means that their results were conditional on certain characteristics of the African societies at various points in time.


Sample Size and Composition

The numbers of African countries examined in the studies vary from one to 34. Researchers who examined single African countries comprise the largest group in terms of sample size and composition. These researchers, according to the countries they investigated, include:

1. Ghana: Bennett (1973) and Ocran (1977)
2. Nigeria: Keil (1970), Luckham (1971), Noah (1970) and Uphoff and Ottemoeller (1970)
3. Uganda: Lofchie (1972)
4. Ethiopia: Warren and Warren (1976)
5. Somalia: Laitin (1976) and Lewis (1972)
6. Liberia: Okolo (1981)
7. Equatorial Guinea: Baynham (1980)
8. Niger: Higgott and Fuglestad (1975)

The rest of the studies examined multiple countries with various sample sizes and compositions. For the sake of brevity, these studies have been divided into five categories (note that the number of countries examined in each of these studies within these categories varies from two to 34):

1. West African States: the studies in this group include those of Bebler (1973), Feit (1968), Tixier (1966) and Welch, Jr. (1986).
2. Sub-Sahara African States: the investigations in this category include those by Decalo (1976, 1986), Jackman (1978), Jackman and Boyd (1979), Jenkins and Kposowa (1990), Johnson, Slater and McGowan (1984), Kposowa and Jenkins (1993), Laurean (2006), McGowan (1975), Morrison and Stevenson (1971, 1972), Nelkin (1967), Orkand Corporation (1983) and Terray (1964).
3. Cross-Continental (Africa-wide) States: research in this category includes the studies of Bienen (1989), Decalo (1990), Welch, Jr. (1967, 1970) and Welch, Jr. and Smith (1974).
4. Developing States Including Some in Africa: the only study here is that by David (1985).
5. World-wide Including Some African States: the works in this category comprise those of Lloyd (1972), Luttwak (1968), Thompson (1975), Wayman (1975) and Zimmermann (1979a, 1979b).

Clearly obvious from the preceding discussion is that all the studies that were identified earlier as quantitative examined multiple African states in their analyses. Those studies classified under the rubric of qualitative methodology used either single or multiple African states in their analyses. Great variation, thus, exists between studies in terms of sample size and composition. Any one study viewed in isolation presents problems of scope and generalizability.


Conceptualization of the Dependent and Independent Variables

There is a great consensus among the studies in terms of the way the dependent variable, coup d’état, is conceptualized within the general rubric of military intervention: that is, a bid for government power, normally by a sudden seizure of strategic points of power in a state by the removal of the ruler and his government. This means that it is likely to be—and most often is—an act of the military or a segment of it. As David noted, the military is behind virtually every coup attempt. By virtue of its superior organization, discipline, centralized command and monopoly of arms, the military is the most powerful institution in developing countries.42
The independent variable, ethnicity, is couched under the rubric of pluralism, which is conceptualized as extra-social dimensions or inherited memberships that foster national disintegration. The other cognate attributes include sectional, regional, linguistic, clan, tribal, racial, cultural and religious divisions.


Measurement of Variables

As can be expected, only those studies classified earlier under the rubric of quantitative methodology measured the variables they used. The measurement of coups d’état is generally couched within the prism of military involvement. It was measured by its users in a number of different ways.
Jackman (1978) and Jackman and Boyd (1979) used a weighing technique: [successful coups d’état X 5] + [unsuccessful coups d’état X 3] + [plots X 1]. (The X means points—thus X 5 means five points.) Zimmermann (1979a, 1979b) employed frequency counts; he counted both successful and unsuccessful coups d’état. Thompson’s (1975) index for coup d’état proneness involved counting the number of years of coups d'état. Military Involvement was measured by Wayman (1975) by scoring one point each for one of three separate levels of military involvement: (a) high—active military officers in government, (b) moderate—active military officers govern behind the scenes, or (c) low—active military officers do not govern at all.
Total Military Involvement was measured by Jenkins and Kposowa (1990), Johnson, Slater and McGowan (1984), Kposowa and Jenkins (1993) and Orkand Corporation (1983) by using a weighing method: [successful coups d'état X 5] + [attempted coups d'état X 3] + [reported plots X 1]. This measurement is, therefore, akin to the one used by Jackman (1978) and Jackman and Boyd (1979) for coup d’état.
Elite Instability was measured by McGowan (1975) and Morrison and Stevenson (1971, 1972) as follows: [coups d’état X 5] + [attempted coups d’état X 3] + [plots X 1]. Thus, it is this latter measurement of elite instability that is similar to Jackman’s (1978) and Jackman and Boyd’s (1979) measurement of coup d’état and Johnson, Slater and McGowan’s (1984) and Orkand Corporation’s (1983) measurement of total military involvement.
Communal Instability was measured by McGowan (1975) by developing a weighted index in the following manner: [civil wars X 5] + [rebellions X 4] + [irredentist events X 3] + [ethnic violence X 1]. Morrison and Stevenson (1971, 1972) used a somewhat similar measurement, but their weighing scales were slightly different (that is, from those of McGowan): [civil wars X 4] + [rebellions X 3] + [irredentist events X 2] + [inter-ethnic violence X 1].
Thus, it appears at this juncture that the studies by Jackman (1978), Jackman and Boyd (1979), Johnson, Slater and McGowan (1984), Kposowa and Jenkins (1993), McGowan (1975), Orkand Corporation (1983) and Morrison and Stevenson (1972) are the most likely candidates for the statistical analysis of the present chapter. The studies identified here entail similar measurements for military intervention in Africa through coups d’état.
Pluralism was measured by Jackman (1978) and Jackman and Boyd (1979) as the percentage of population in the largest ethnic group. Johnson, Slater and McGowan (1984) and Orkand Corporation (1983) used a political pluralism measurement for 1965 which is based on the relative degree of political democracy, number of political parties, political party dominance, the degree of fractionalization or party cleavages in a country’s lower house of the legislature, and the level of political violence. McGowan (1975) measured pluralism in terms of ethnicity by counting the number of ethnic languages in 1967. Morrison and Stevenson (1971, 1972) employed a cultural pluralism index which involved delineating a factor loading for 15 indicators related to national cultural variance in ecological adaptation, in authority/stratification, and in social organization. Thompson (1975) used a cultural homogeneity index involving a number of measurements in terms of race, language, and religion. Wayman (1975) measured pluralism in terms of the degree of cleavage between modern and traditional sectors by using the percentage of the population discriminated against plus the percentage of population with potential separatist tendencies. Finally, Jenkins and Kposowa (1990) and Kposowa and Jenkins (1993) measured pluralism by developing a single ethnic antagonisms variable based on the population measures for community stratification, political authority, social stratification, and number of ethnic groups.


Results

Since researchers who used the qualitative method intermeshed the conceptualization of their variables with the results of their investigations, the suggested relationships between independent and dependent variables in those works are consequently confirmed. Thus, it suffices to refer the reader to the discussion on the conceptualization of the independent and dependent variables above. Researchers who used the quantitative method, on the other hand, began by delineating a number of variables from the available literature and then performed statistical tests to determine which variables are related. Thus, the discussion that follows is exclusively on the results from those quantitative studies.
Since the quantitative studies were done by individuals and smaller groups of researchers with different interests and foci, grouping their results is a very difficult task. What follows, therefore, is a discussion of how best to make sense of those results without losing their substance.
Jackman (1978) and Jackman and Boyd (1979) found social mobilization and presence of a dominant ethnic group and multipartyism to be politically destabilizing. According to them, electoral turnout in the last election before independence was stabilizing. Taken together, these investigators found that a dominant ethnic group and multipartyism explained over 80% of the variance of coups d’état in Sub-Sahara Africa in the period covering 1960-1975.
Jenkins and Kposowa (1990) and Kposowa and Jenkins (1993), as stated in the introduction of this chapter, were quite decided in asserting that of all the variables that have been shown in previous studies to influence military coups d’état, ethnicity has the strongest influence. They added that ethnic competition and economic dependence seem to create a structural context for military intervention and related instabilities in Africa.
Johnson, Slater and McGowan (1984) and Orkand Corporation (1983) found that African states with relatively dynamic economies that were not mobilized before independence and had maintained or restored some degree of political participation and pluralism while keeping their military forces small and non-political had been the most stable in terms of military coups d’état and associated forms of military intervention in politics. The alternative is true for countries that had the opposite set of characteristics.
McGowan’s inquiry (1975) yielded four categories of findings. First, African states with large and influential military forces and extensive ethnic pluralism and social mobilization had extensive communal instability. African states in which government policy was achieving some degree of success in promoting national integration and economic development, as well as flows of foreign aid to them, had a low likelihood of communal instability. Second, African states that had fragmented party systems and increasing degrees of social mobilization were most likely to experience elite instability. Those states that were making progress in the areas of national integration, economic growth, and foreign assistance flows were less likely to have military coups d’état and other related elite instability phenomena. Third, five factors were statistically related to coups d'état. These included social mobilization, political party disunity, interest group size, government economic success, and external support. Finally, countries that had coups d’état through the end of 1969 exhibited greater levels of social mobilization, larger urban interest groups, worse economic performance, less political party unity, and less metropole foreign aid per capita than did the non-coup countries.
Morrison and Stevenson in their 1971 study found that cultural pluralism increased the likelihood of conflict between members of communal groups and the probability of both elite and communal instability. In addition, their results revealed that greater degrees of variation in the ethnic background of elites increased the likelihood of elite instability; modernization decreased the likelihood of political instability. In their 1972 study, they found that the relationship between the incidence of elite and communal instability was weak while those between turmoil and elite or communal instability were quite strong.
Thompson in his 1975 study revealed that coups d’état were the outcome of the military's corporate grievances and not so-corporate motivations. He also found that military coups d’état which reflected greater organizational cohesion were more likely to be successful.
Wayman (1975) found political institutionalization to be the most important factor for military involvement; the second most important factors were conflicts over modernization; the third most important factors were effects of external war. He also discovered that effects of civil disorder were weakly related to military involvement and that the importance of brute force had no measurable effect.
In sum, the most agreed upon explanations for coups d’état outcomes include economic condition, institutional structure, military’s corporate or personal interest, domestic environment for participation, pluralism, social mobilization, and foreign influence. Since the focus of this chapter is on the connection between ethnicity and coups d’état, the following section deals only with these two variables.


Statistical Analysis

The analysis that follows is twofold. First, since the selected studies employed correlation coefficients or other statistics that can be easily converted into correlation coefficients, the correlational relationship between the independent variable, ethnicity/pluralism, and the dependent variable, coup d’état/military intervention, is determined first by calculating their raw (unadjusted) statistical results. Second, weighted statistical results are computed to empirically examine the impact of sample size (number of countries included in the analysis of each study).
A simple method for synthesizing correlational results essentially involves obtaining the average of the correlation between two variables. This is typically done by averaging the raw Pearson correlation coefficients (r) using formula one.

r’= Σr [ 1 ]
n

where r is Pearson correlation from each study, and n is the number of correlation coefficients combined. The r’ (mean r) is then used and reported as the effect size indicator.43 To determine the impact of sample size, formula two is utilized.

(ws)r’ = Σ(ws)r [ 2 ]
n

where (ws) is weighted sample size, r is Pearson correlation from each study, and n is the number of correlation coefficients combined.
The issues surrounding the interpretation of what constitutes either a small, medium, or large effect for correlational analysis is beyond the scope of the present chapter. In the absence of a single agreed-upon standard, Cohen’s guidelines for small (r = .10), medium (r = .30), and large (r = .50) effect sizes are utilized here because they have been widely employed in meta-analytical studies.44
The selected studies that used pluralism/ethnicity to explain military intervention/coup d’état outcomes and their attendant statistics are as follows: (a) Jackman (1978)—r = .27; (b) Jackman and Boyd (1979)—r = .33; (c) Johnson, Slater and McGowan (1984)—r = -.76; (d) Kposowa and Jenkins (1993)—r = .36; (e) Orkand Corporation (1983)—r = -.76; (f) Morrison and Stevenson (1971)—r = .27; and (g) Morrison and Stevenson (1972)—r = .52. Using formula one, these statistics are calculated as follows:

r’ = .27 + .33 -.76 + .36 -.76 +.27 + .52 = .03
7

An r’ of .03 suggests that pluralism has a positive, but very small, effect on coup d’état outcomes. Weighing for sample size, formula two is employed as follows:

(sw)r’=
(.29)(.27)+(29)(.33)+(35)(-.76)+(33)(.36)+(35)(-.76)+(32)(.27)+(32)(.52) = .0008
(225)7

The weighted r’ here is .0008. This means that pluralism has an almost non-existent effect on coup d’état outcomes. Sample size, however, appears to explain the very small effect delineated above. In sum, the connection between ethnicity and military coups d’état in Africa seems to have been overstated.


Conclusion

The available works that have dealt with the connection between ethnicity and military intervention in Africa through coups d’état have followed a tradition of reviewing previous studies on the subject by dividing them into what are frequently referred to as the two major schools of thought: the “systemic” and the “praetorian.” The former is said to hold the view that the military intervenes in politics because of systemic ills in a country and, thus, seeks to protect the political and institutional structures of society. The latter is said to purport the notion that the military intervenes in politics because it seeks to preserve and/or extend its corporate interests. While this approach has certain merits, it does not, however, tell readers about the methods, techniques, and type of data various researchers used to arrive at their conclusions. It also does not account for those studies that employ both “systemic” and “praetorian” variables in their analyses.
The present essay, which employs a meta-analytic method, views the research activity on the connection between ethnicity and military intervention in Africa as the accumulation and refinement of information and knowledge. This approach is a small step toward establishing guidelines for reliable and valid reviews, integrations, and syntheses for future research on military intervention in Africa. Given the enormous amount of data that must be gathered, processed, and synthesized from many disciplines, it is true that the exercise is a difficult one. But it behooves social scientists to make the effort, if the traditional approach of reviewing literature in an unscientific, impressionistic fashion is to be abandoned.


Endnotes

1. A. A. Said and A. K. Bangura. “Ethnicity and Peaceful Relations.” Peace Review.
Vol. 3, No. 4, winter 1991-92, pp.24.
2. A. K. Bangura. Multilingualism and Diglossia in Sierra Leone. (Lawrenceville, VA: Brunswick Publishing Corporation, 1991), p.28.
3. J. C. Jenkins and A. J. Kposowa. “Explaining Military Coups d’État: Black Africa,
957-1984.” American Sociological Review. Vol. 55, No. 6, December 1990, pp.861-
875; A. J. Kposowa and J. C. Jenkins. “The Structural Sources of Military Coups
in Postcolonial Africa, 1957-1984.” The American Journal of Sociology. Vol. 99, No.1,
July 1993, pp.126-163.
4. A. K. Bangura. “Explaining and Predicting the Causes of Military Coups d’État in
Africa: A Meta-Analysis.” In A. K. Bangura, ed. Research Methodology and African
Studies. Vol. 1. (Lanham, MD: University Press of America), 1994.
5. S. Baynham. “Equatorial Guinea: The Terror and the Coup!” The World Today.
February 1980.
6. A. Bebler. Military Rule in Africa: Dahomey, Ghana, Sierra Leone and Mali. (New York: Praeger), 1972.
7. V. P. Bennett. “The Motivation for Military Intervention: The Case of Ghana.”
Western Political Quarterly. December 1973.
8. H. Bienen. Armed Forces Conflict and Change in Africa. (Boulder: Westview Press),
1989.
9. S. R. David. Defending Third World Regimes from Coups d’État. (Lanham, MD:
University Press of America), 1985.
10. S. Decalo. Coups and Army Rule in Africa: Studies in Military Style. (New Haven:
Yale University Press), 1976.; S. Decalo. “Military Rule in Africa: Etiology and Morphology.” In S. Baynham, ed. Military Power and Politics in Africa. (London: Croom Helm), 1986; S. Decalo. Coups and Army Rule in Africa: Motivations and Constraints. (New Haven: Yale University Press), 1990.
11. E. Feit. “Military Coups and Political Development: Some Lessons from Ghana
and Nigeria.” World Politics. Vol. 20, No. 2, January 1968, pp.179-93; E. Feit.
“The Rule of the ‘Iron Surgeons’: Military Government in Spain and Ghana.”
Comparative Politics. Vol. 20, No. 2, April 1969.
12. R. A. Higgot and F. Fuglestad. “The 1964 Coup d'Etat in Niger: Towards an
Explanation.” The Journal of Modern African Studies. Vol. 13, 1975.
13. J. Keil. “The Price of Nigerian Victory.” Africa Today. Vol. 17, 1970.
14. D. D. Laitin. “The Political Economy of Military Rule in Somalia.” The Journal of
Modern African Studies. Vol. 14, 1976.
15. A. M. Laurean. “The Logic of the Coup.’ Peace and Conflict Monitor (University for Peace). March 15, 2006. Available at
http://www.monitor.upeace.org/archive.cfm?id_article=352 Retrieved on
August 23, 2006.
16. I. M. Lewis. “The Politics of the 1969 Somali Coup.” The Journal of Modern African Studies. Vol. 10, 1972.
17. P. C. Lloyd. Class, Crises, and Coups. (London: MacGibbon and Kee), 1971.
18. M. Lofchie. “The Uganda Coup—Class Action by the Military.” The Journal of
Modern African Studies. May 1972.
19. R. Luckham. The Nigerian Military. (Cambridge: Cambridge University Press),
1971.
20. E. Luttwak. Coup d'État: A Practical Handbook. (London: Allen Lane), 1968.
21. D. Nelkin. “The Economic and Social Setting of Military Takeovers in Africa.”
Journal of Asian and African Studies. Vol. 2, 1967.
22. M. E. Noah. “The Nigerian Civil War and the Gullitles.” Africa Today (Special
Book Review Issue). Vol. 17, No. 2, 1970.
23. A. K. Ocran. Politics of the Sword: A Personal Memoir on Military Involvement in Ghana and Problems of Military Government. (London: R. Collins), 1977.
24. J. E. Okolo. “Liberia: The Military Coup and Its Aftermath.” The World Today.
April 1981.
25. E. Terray. “Les Revolutions Congolese et Dahomeens de 1963: Essai
d’Interpretations.” Revue Française de Science Politique. Vol. 14, 1964.
26. G. Tixier. “Les Coupes d'Etat Militaires en Afrique de l’Ouest.” Revue du Droit
Public et de la Science Politique. November-December 1966.
27. N. T. Uphoff and Rev. H. Ottemoeller. “After the Nigerian Civil War: With
Malice Toward Whom”? Africa Today (Special Book Review Issue). Vol. 17, No.2,
1970.
28. H. Warren and A. Warren. “The US Role in the Eritrean Conflict.” Africa Today.
Vol. 23, No. 2, 1976.
29. C. E. Welch, Jr. “Soldier and State in Africa.” The Journal of Modern African Studies. Vol. 5, 1967; C. E. Welch, Jr., ed. Soldier and State in Africa: A Comparative Analysis of Military Intervention and Political Change. (Evanston: Northwestern University Press), 1970; C. E. Welch, Jr. “Military Disengagement from Politics?: Incentives and Obstacles in Political Change.” In S. Baynham, ed. Military Power and Politics in Black Africa. (London: Croom Helm), 1986.
30. C. E. Welch, Jr. and A. K. Smith. Military Role and Rule. (North Seituate, MA:
Duxbury Press), 1974.
31. R. W. Jackman. “The Predictability of Coups d’État: A Model with African
Data.” American Political Science Review. Vol. 72, 1978.
32. R. W. Jackman and W. D. Boyd. “Multiple Sources in the Collection of Data on
Political Conflict.” American Journal of Political Science. Vol. 23, 1979.
33. Jenkins and Kposowa, 1990.
34. T. H. Johnson, R. O. Slater and P. McGowan. “Explaining African Military
Coups d’État, 1960-1982.” American Political Science Review. Vol. 78, 1984.
35. Kposowa and Jenkins, 1993.
36. P. McGowan. “Predicting Political Instability in Tropical Africa.” In M. K.
O’Leary and W. D, Coplin, eds. Quantitative Techniques in Foreign Policy Analysis and
Forecasting. (New York: Praeger), 1975.
37. D. G. Morrison and H. M. Stevenson. “Political Instability in Independent Black
Africa: More Dimensions of Conflict Behavior within Nations. Journal of Conflict
Resolution. Vol. 15, 1971; D. G. Morrison and H. M. Stevenson. “Cultural
Pluralism, Modernization, and Conflict: An Empirical Analysis of Sources of
Political Instability in African Nations.” Canadian Journal of Political Science. Vol. 5, 1972.
38. Orkand Corporation. Analysis of the Causes of Coups d’État in Sub-Saharan Africa,
1960-1982. (Silver Spring, MD: Orkand Corporation), 1983.
39. W. R. Thompson. “Regime Vulnerability and the Military Coup.” Comparative
Politics. Vol. 7, 1975.
40. F. W. Wayman. Military Involvement in Politics: A Causal Model. Sage Papers in
International Studies Series, No. 02-035. Beverly Hills: Sage Publications.
41. E. Zimmerman. “Toward a Causal Model of Military Coups d’État.” Armed Forces
and Society. Vol. 5, 1979a; E. Zimmerman. “Explaining Military Coups d’État:
Towards the Development of a Complex Causal Model.” Quality and Quantity. Vol.
13, 1979b.
42. David, 1985, p.3.
43. F. M. Wolf. Meta-Analysis. (Thousand Oaks: Sage Publications), 1986.
44. J. Cohen. Statistical Power Analysis for the Behavioral Science. (Hillsdale: Erlbaum), 1977.