1. Introduction and Summary
Democratic Peace page
The injuries that they themselves procure
Must be their schoolmasters.
---- Shakespeare, King Lear II.iv.
In Chapter 33 and Chapter 34 I have tried to establish the logic to assess the empirical implications of the conflict helix, and the empirical validity of my distinctions between state-societies and political systems. These chapters now enable me to so assess my seven propositions about societal conflict given in Chapter 32. First, however, a note on the empirical sources.
Since the early 1960's there has been an increasing number of quantitative analyses of societal level conflict. Many of these have tried to span the variety of social conflict behavior and the diversity of states. Various hypotheses, such as that relative deprivation causes conflict, have been proposed and tested. In testing the empirical worth of the conflict helix, I will utilize all these studies I have found, as well as my own.
Most of these studies are similar in mathematical logic (not epistemological) to that developed for the conflict helix at the societal level. That is, most use factor analysis and regression analysis.
For example, the Feierabends (1972, p. 144) define societal frustration as a ratio of want satisfaction to want formation. Want satisfaction is indexed by per capita GNP, caloric intake, physicians, telephones, newspapers, and radios. Want formation was indexed by literacy and urbanization. Leaving aside the validity of this index, my concern in screening all these studies was the underlying indicators and their relation to conflict behavior, in other words, the raw attributes and not the concepts they were meant to measure.
One more point. I do not always accept the methodological interpretation of a study's results. Many interpretations (such as a significance level) are a matter of intuition and perspective. In any case, I will include sufficient reference so that the reader himself may judge.
Now for the details.
Because of its extent, the evidence here and for the subsequent propositions will be presented tersely in numbered paragraphs rather than in subsections.
(1) In Cattell's (1949) common factor analysis of sixty-nine states for seventy-two varied attributes for data aggregated over 1837-1937, and 1927-1937, two general common conflict components were found. One (factor 1) comprised assassinations, riots and rebellions, and divorces to marriages. An anomaly was a high loading for area. A second (factor 6) involved homicides and assassinations negatively. Moreover, revolution was independent of all the components, with the highest loading being .35. Conclusion: Conflict manifestations are independent of historical structures of expectations (other components).
(2) Cattell, Bruel, and Hartman (1951) refined the previous study by eliminating twenty-nine states and redoing the common factor analysis. A distinct common conflict component that emerged was independent of the others and involved foreign clashes, riots and rebellions, war, divorces to marriages, and assassinations. Anomalies loading on this component were creativity, Nobel prizes, treaties, and area. Finally, revolution maintained its independence from the other components. Conclusion: Conflict manifestations are independent of historical structures of expectations (other components).
(3) Cattell and Gorsuch (1965) did a common factor analysis of fifty-one diverse attributes for fifty-two states on 1953-1958 data. The only domestic conflict attribute was "riots and rebellions." It had no high relationship with any component, and its highest loading was -.58 on a wealth component. Conclusion: Conflict manifestations are mainly
3independent of structures of expectations.
(4) Russett (1967) did a component analysis of fifty-four varied attributes for eighty-two states on data for the late 1950s. The only conflict attribute was "deaths from political violence," with its highest loading (-55) on a wealth component. Its communality was 42 percent across the five components re ported. Conclusion: Conflict manifestations are mainly independent of structures of expectations.
(5) I did (1972) a component analysis of 236 diverse attributes for eighty-two states on 1955 data. A clear independent conflict component emerged for all nine conflict attributes (assassinations, general strikes, major government crises, purges, riots, revolutions, anti-government demonstrations, and killed from social conflict) independent from all other components. Conclusion: Conflict manifestations are independent of structures of expectations.
(6) My (1979) component and common factor time series analyses of ninety-one diverse attributes for all states over the years 1950, 1955, 1960, 1963, and 1965 produced a clear independent conflict component, involving general strikes, riots, killed from social conflict, purges, government crises, and demonstrations. Conclusion: Conflict manifestations are independent of space-time structures of expectations.
(7) Adelman and Morris (1967) did several component analyses on forty attributes for seventy-four least-developed states on circa 1960 data; first on all attributes and states; then separately on African, Asian, and Latin American states; and separately for lowest developed, intermediate, and highest developed (among the seventy-four least developed). In general their two manifest conflict attributes -social tension and political stability -emerged as components independent of the economic and political components throughout these analyses. Conclusion: Conflict manifestations are independent of subsets of structures of expectations.
(8) Michael Hudson (1970) did a component analysis of sixty-three states for twenty-five attributes of violence, instability and political institutionalization on 1948-1965 data. One component (factor 2) of government instability was uncovered, and involved internal security forces, coups, major government change, regime change, and assassinations. Another of civil strife (factor 5) was found and comprised riots, demonstrations, armed attacks, and assassinations. These two components were independent of two others Hudson calls structural differentiation (factor 1) and structural durability (factor 6), as well as two others I would call liberal democracy (factor 3) and leftism (factor 4). Conclusion: Conflict manifestations are independent of political structures of expectations.
(9) Gregg and Banks' (1965) component analysis of sixty-eight attributes for 115 states
4found an independent conflict component comprising government instability, party system instability, military interventive, and domestic killed, and also (negatively) involving Western regional grouping, modem bureaucracy, political enculturation, westernization; and positively personalisimo, presidential-legislative-executive structure, Latin American areal grouping, and articulation by anomic groups. This component, therefore, represents violent conflict and instability that is higher among Latin American states and lower among Western developed. What is being tapped, here, is a difference in societies rather than specific common structures of expectations. This can be seen from the communality of demonstrations (.27) and domestic killed (.44), which are among the lowest in the analysis. This low communality and independent conflict component, then, imply the following. Conclusion: Conflict manifestations are independent of political structures of expectations.
(10) Levine (1973) did a separate component analysis of twenty-nine diverse attributes for all states for each of seven international eras determined through an overtime analysis (O-factor analysis). The seven were 1892, 1908, 1922, 1927, 1936, and 1952. Included in each analysis was a coups-attribute. Coups did not correlate with any of the major components for all seven analyses. For the last three analyses it related to a component defined by "selection of effective executive" and "legislative executive." In each case, however, coups had a moderate loading (.63, -.64) on the component and low communality overall. Conclusion: Conflict manifestations are independent of historical structures of expectations.
(11) Von der Meliden (1973) cross-tabulated the highest and lowest violent states and found no constant relationship to economic level (p. 109), to growth in economic standards and communication (p. 109), to level of communication (p. 109), to education (p. 98), and to size of government forces (p. 98). Insofar as these attributes reflect structures of expectations, then the Conclusion: Conflict manifestations are independent of structures of expectations.
(12) In his appendices Eisenstadt (1963) presents data on rebellion in the life of twenty-seven historical bureaucratic empires. In all there are fifty different empires and phases of empires for which rebellion is tabulated. I cross tabulated level of rebellion separately against Eisenstadt's data on autonomy of political goals, traditional legitimacy, differentiation of groups and institutions, free-loading power, centralization, institutionalization of power, scope of bureaucracy, autonomy of bureaucracy, group political participation, relative strength of ruler, and relative strength of aristocracy. No significant (p <.05) chi-square occurred. Insofar as these political attributes reflect political structures of expectations, then the Conclusion: Conflict manifestations are independent of historical political structures of expectations.
(13) Lodhi and Tilly (1973) analyzed collective violence and crimes in France after 1830. They found no relationship between urban growth (urbanization) and crimes against property, crimes against persons, and collective violence. Urbanization is a facet of the structure of expectations I call wealth. Conclusion: In France, conflict manifestations are independent of an urbanization structure of expectations.
The above provides overwhelmingly support for the proposition that conflict manifestations are random with regard to structures of expectations. No doubt I have missed some studies, but those covered should be the main sources of evidence available today. And they are largely unambiguous.
(1) Feierabend and Feierabend (1966) did a cross-national analysis of 1948-1962 domestic conflict for states with available data (usually sixty-seven states or more), and found those with the lowest change on socioeconomic variables were most stable; those with the greatest change, the least stable. Stability comprised a seven point scale measuring the intensity of civil strife, with civil war being the most intense. Conclusion: Social change produces conflict manifestations.
(2) The Feierabends (1972) did a stepwise regression analysis (their Study 6) of political instability. The rate of socioeconomic change had a relationship (t = 1.07; Beta = .19) to instability in the proper direction for 1948-1962 data, and a very high relationship (t = 6.42; Beta = .67) for 1955-1961 data. In the latter case, change accounted for 45 percent of the variance. Conclusion: Social change produces conflict manifestations.
One of the ingredients of change has been the extension of the state-the political system-into all regions of society. This extension of the superordinate class has created an expanding conflict front as one by one traditional structures of expectations and buttresses against the state were disrupted.
(3) Bwy (1972) found a strong curvilinear relationship between the ratio (percent) of defense expenditures to GNP and anomic violence (turmoil) for twenty Latin American states, 1958-1960. For these states, defense expenditures serve to buttress the political system against domestic enemies. Conclusion: Conflict manifestations are curvilinearly related to state power.
(4) Snyder and Tilly (1972) regressed participants in collective violence in France, 1886-1939, onto excess of arrests, national budget, man-days in jail, national elections, and time (t and t2). They found that the growth in the national budget and time significantly accounted for 50 percent in the variance of violence. In other words, the growth of the state was the major factor. When political attributes and measures of deprivation were analyzed together for 1830-1960 violence, only the political attributes were significant. Conclusion: Conflict manifestations are related to the growth of French governmental power.
(5) Tilly (1972) analyzed the riots and demonstrations (disturbances) in France, 1832-1958. He found that fluctuation follows changes in national conflict and do not correspond to the pace of urban growth. Political conflict becomes nationalized and then urbanized. Conclusion: Conflict manifestations are related to the growth of French governmental power.
(6) Levy (1969) analyzed 150 years of violent events in the United States. He found that violent actions against authority have increased, while violence for personal gain has decreased. Moreover, violence for political advantage has also increased and labor violence is decreasing after a peak decades ago. Group antagonisms and religious antagonisms have also increased. Considering that federal power has increased greatly during this period, Conclusion: Conflict manifestations are related to the growth of American federal power.
State-societies consist of diverse structures of expectations, but the state itself is a general structure subsuming and framing all the others. It is when this general structure itself is disrupted--when the society is changing type--that there is the most intense forms of violence: social revolution, civil war, and large-scale guerrilla war. Change in type, in the balance of societal power, produces the most violence. In the last 150 years, much of this change has been from authoritative to coercive or mixed societies in the process of economic development. Societies which were authoritative or coercive before development virtually never become exchange societies afterward. Those that were exchange societies before development almost always remain that way through the process.
(7) In their analyses described in Section 35.2, paragraph (1), the Feierabends divided states into modern, transitional, and traditional. The mean in stability rating for each level was 268, 472, and 420 respectively. The transitional period is the most unstable. Conclusion: Societies undergoing transition are most productive of conflict manifestations.
(8) Sorokin (1937-1941) analyzed data on the fluctuations of internal disturbance in the history of Greece, Rome, and Europe. For either quarter century or century periods he found no continuous trend, nor periodicity in their ups and downs. He did find that when cultures and social relationship are breaking down and undergoing transformation, internal disturbances (such as revolutions) increase. When these are crystallized, internal disturbances decrease. The "main and the indispensable condition for an eruption of internal disturbances is that the social system or the cultural system or both shall be unsettled" (Sorokin, 1957, p. 602, italics omitted). By transition Sorokin means from an ideational to a sensate supercultural system or vice versa; in my related terms this would mean from authoritative to exchange or coercive, or vice versa. Conclusion: Change in societal type is productive of intense conflict manifestations.
Aside from changes in societal-type, or those resulting from an increase in the power of the state, there are the shifts in specific structures of expectations constituting economic development or, in my terms, increase in wealth. In a poor, authoritative society, societal structures of expectations mainly involve communal groups that tic individuals together through their regional, religious, ethnic, racial, or linguistic attributes. The process of development breaks down these communal structures and transforms them into associational bases. Then, interest groups, businesses, unions, and the like become the bases of societal expectations. This transformation of expectations-the process of development produces considerable mass turmoil and unrest, conflict that is the means for reforming communal structures on an associational basis. We should expect, therefore, the highest conflict in the midrange of development, in the transitional societies. Some evidence on this has been provided in (7), above. In addition are the following.
(9) Tilly (1972) analyzed 1,393 conflict events for 3,250 groups in France over the period 1830-1960, and compared 1830-1860 to 1930-1960. He found a change in the pattern of conflict with industrialization. There was a decline in simple crowd and military participants; an increase in ideological crowd or activists and police participants. There was a decline in users of the same market, fields, etc. (communal groups) and a rise in the participation of those with an associational base. Conclusion: Development involves a change from communal to associational structures of expectations and a concomitant change in the source of conflict manifestations.
(10) The Feierabends and Nesvold (1969) analyzed around eighty-four states for 1948-1965 conflict data. They found that as development progresses and the political system changes there is more conflict. Once the system is modernized and developed, however, political stability comes about. Traditional oligarchies and developed European states have the least conflict. Conclusion: Development is curvilinearly related to conflict manifestations.
(11) Gurr (1969) found that groups participating in civil strife, 1961-1965, for 114 states broke down as follows. Communal and government groups participated most, and economic and political groups least, in the lowest developed states. However, communal groups were curvilinearly related to development, with a high concentration of communal conflict in developed European nations. Nonetheless, the increased participation of economic and political groups at higher levels of development bears out the Conclusion: Development involves a change from communal to associational structures of expectation and a concomitant change in the source of conflict manifestations.
Aside from the shift in conflict involved in development, communal heterogeneity itself is a source of conflict. Different regional, ethnic, racial, religious, linguistic groups demarcate cultural and often status cleavages in society. These divisions separate groups that are socioculturally distant and serve as a reservoir for opposing communal and interest conflict groups. Development may shift the base of conflict to associational interests, but communal identities can still be a source of unrest and disturbances, as the United States and Europe show.
However, the major structure of expectations governing society is more stable if there is both communal heterogeneity and development. This is because heterogeneity adds a different source of cross-pressures to those based on associational groups (structures of expectations). Therefore, there should be more turmoil for communally heterogeneous states, but less for those states which are heterogeneous and developed.
(12) Michael Hudson's (1970) regression analysis showed that for sixty-three states and 1948-1965 conflict data, the more ethnolinguistically heterogeneous and developed a state, the more government stability (fewer coups, assassinations, major government changes) with an R = .64. Moreover, the more the ethnolinguistic homogeneity and the less the development, the less the civil disorder (riots, demonstrations, armed attacks, assassinations). Just considering the peak conflict (crisis) states and years, Hudson found them concentrated in the homogeneous and developed states, or heterogeneous and undeveloped states. Conclusion: Communal heterogeneity and development stabilize a political structure of expectations.
(13) Von der Meliden (1973) cross tabulated violent states and found some positive relationship between ethnic and racial heterogeneity and violence. Conclusion: Communal heterogeneity produces violence.
Development or the increase in society's wealth means the diversification of institutions, associational. groups, in short, the division of labor. But institutional division is one part of it, the other is the increase in the goods available and in the opportunities for the satisfaction of diverse interests and for mobility. There are more specific structures of expectations, more overlapping and crosscutting structures, and more of an opportunity for disruption and reformation. Consequently, rapid development should lessen the intensity of conflict across society without an elimination of that unrest (or social friction) associated with changing specific structures of expectations. Rapid development, therefore, should decrease, but not eliminate, conflict.
(14) Hibbs (1973) analyzed conflict for 108 states for 1948-1967 data. One regression analysis showed that holding other attributes constant, the higher the rate of increase in energy consumption per capita (t = -1.99), the less the tendency to collectively protest (riots, anti-government demonstrations, and political strikes). In his final model developed after extensive path analysis, he included the rate of increase in energy consumption per capita (among other attributes) as a negative influence on collective protests. The rate, however, had no effect on internal war (deaths from political violence, armed attacks, and assassinations. Conclusion: Rapid development lessens manifest turmoil.
(15) Russett and colleagues (1964) related the deaths 1950-1962 per million population to the annual economic growth rate of fifty-three states. They found a curvilinear relationship, with deaths per million highest at the middle rate of change. In other words, for societies with the same population size, a slow rate of growth does not involve too great a disruption of established expectations, but as the rate of growth increases structures collapse and violence becomes a means for the rebalancing of power. After a point, however, the increasing growth rate reflects such a growing diversification that the tendency to violence dissipates along multiple divisions and interests. Conclusion: Holding population constant, rapidity of development after a threshold lessens violence.
(16) In his aforementioned analysis of Latin American states, Bwy found (1972) that GNP growth rate correlated -.63 with organized violence and -.33 with anomic (turmoil). Conclusion: Rapidity of development lessens manifest conflict.
(17) Parvin (1973) did a regression analysis of violent deaths per million for twenty-six states on national income per capita, income growth rate, income distribution (equality), mobility, communications, and urbanization. He found that for an unweighted regression (each state has equal weight), growth rate had the highest effect and in the negative direction (t = -2.82). Mobility was second highest, also in the negative direction (t = -2.10). For a weighted regression (each state weighted by population) growth rate maintained a significant negative influence (t = -6.50) while mobility had the greatest negative effect (t = -12.33). Conclusion: Rapid development lessens manifest conflict.
Change, however, is usually not synchronous.
(18) Using the Feierabend conflict data, Sofranko and Bealer (1972) assessed the degree to which developmental imbalance produced conflict. They determined imbalance by cross-sectional regression, where the residuals measured imbalance. When the direction of imbalance was taken into account, they found that balanced structural change tends to be more peaceful. The overdevelopment (greater change) of one sector tends towards conflict; when the political, educational, and economic structures are all overdeveloped conflict tends to be highest. The overdevelopment of the educational structure also leads to high conflict (but less than the overdevelopment of the three structures together). When only the political structure is underdeveloped, there is the least conflict. Since by political development, they meant increased centralization of political power and its intrusion into societies' activities, clearly conflict would be low so long as the aggrandizement of state power lagged behind changes in other institutions. For it is the assertion of such power, of social class dominance, that provokes the major conflict front in society. Conclusion: Unbalanced change in the structures of expectations produces high manifest conflict.
(19) The aforementioned Feierabends and Nesvold (1969) study found that as the rate of change in education increased across states, their internal strife and violence increased. The more education, the more dissatisfaction with established structures of expectations; the underlying balance between interests, capability, will is altered. Conclusion: Rapid change in education especially produces high manifest conflict.
I have shown that rapid economic growth rates reduce the societal effects of disrupting expectations. But the above findings indicate also a counterworking effect of rapid educational development. Therefore, manifest conflict should closely follow the balance between education and economic growth. But to treat this as a ratio of the two growth rates would be a mistake, for societies may have a structural imbalance not reflected in the growth rates. A highly developed but educationally retarded society with low economic and high educational growth rates would be expected to have little related conflict, since there is much economic and mobility opportunities for the newly educated. But the impact of simultaneous development and educational growth in a traditional society may be highly disruptive and conflictful. Therefore, we should deal with a ratio of structural levels, rather than of rates. The higher the educational development of a society compared to its economic development, the more likely will be large shifts in societal power and much attendant manifest conflict. The evidence follows.
(20) The Feierabends (1966) in their aforementioned analysis developed a measure of "frustration" which is a ratio of a state-society's per capita GNP, calories, physicians, telephones, newspapers, and radios to their literacy and urbanization. I do not accept this as a measure of systemic frustration,
10but it does roughly measure the ratio of economic development of a society 11to its educational development. 12The higher this ratio, the more developed education is compared to economic development. In fact, the Feierabends found that the higher this ratio, the greater the instability (chi square: p <.001). Conclusion: The more developed a society's educational structures compared to its economic, the more likely manifest societal conflict.
(21) Using the same economic/education ratio and different conflict data, the Feierabends (1972) divided nations into levels of stability (manifest conflict) cross-tabulated against the degree of joint coercion and "frustration." The most stable are those with little coercion and a low economic/education ratio; the mid-stable were those that were coercive and at a mid-level on the economic/ education ratio; the most unstable were mid-level coercive and a high economic/ education ratio. Conclusion: The more developed a society's educational structures compared to its economic, the more likely manifest societal conflict.
(22) Terrell (1971) analyzed 1955-1960 factor scores for turmoil and internal war for seventy-five states (on factor scores and factor analysis, see "Understanding Factor Analysis"). Using the Feierabend economic/education ratio, he found an over-all low but positive relationship between the ratio and manifest conflict. However, at the higher values of the ratio the correlation was much closer (on correlations in general, see Understanding Correlation). Conclusion: The more developed a society's educational structures compared to its economic, the more likely manifest societal conflict.
(23) Zapf and Flora (1973) analyzed 1820-1960 data for ten states to determine if an imbalance of societal demands to capabilities predicted conflict. With demands operationalized as literary and voting participation,
13and capabilities as nonagricultural occupations and nonrural residence, they then hypothesized that political instability (revolutions plus ministerial change) would occur when demands exceeded capabilities and were not balanced by an income growth rate greater than ten percent. They were able to predict instability in forty-five out of fifty-one cases. Although their operationalization is oblique to an economic educational development ratio (and includes a measure -urbanization-in the capabilities side), there is sufficient overlap for the Conclusion: When not absorbed by a high rate of economic growth, the more developed educational structures are compared to economic, the more likely manifest conflict.
Overall, then, change does produce manifest conflict. Of course, as for all brief generalizations, qualifications are necessary. By change is meant sociocultural change-that in the meanings, values, norms, statuses, and class of society's members. Second, not all social change produces conflict, but only those changes which dissychronize a structure of expectations, making it susceptible to disruption or acting as triggers that disrupt such structures. At the individual level, these are changes in individual interests, capability (resources or coercive power), and will.
More specifically about state-societies, I argue, and the above evidence confirms, that violent conflicts swirl around the political elite (state), aggrandizing their power and extending coercion into all regions of society. This is the extension of the control of a dominant class. Manifest conflict occurs up to a point, however. A strong elite can repress overt collective conflict through terror. Second, the most conflictful periods in a society's history occurs during the transition to a different sociopolitical type. This constitutes a breakdown in the overall structure of sociocultural expectations and a reformulation of societal understandings, rules, and law-norms. It is a transformation of classes.
Third, the economic development of a society involves a change from mainly a communal basis of manifest conflict to an associational one (although communal elements may be mixed in, as in the United States). Development also provides diverse ways of satisfying interests and reforming expectations. Thus, development first upsets communal divisions but later eases rebalancing expectations, and is curvilinearly related to manifest conflict.
Fourth, at higher levels, where development is not of one communal group at the expense of another, communal divisions are additional crosscutting sources of interests and power and tend to reduce conflict. Thus, heterogeneous, developed states tend to be most peaceful.
Fifth, higher rates of development promote harmony, not conflict. The faster the development, the more the stability. Thus, while change in individual structures and balances will produce conflict, the over-all societal effect of rapid change in these socioeconomic structures is a larger peace. This is the conflict counterpart to the market. Individuals spontaneously pursuing their own interests produce an over-all benefit to all and a greater harmony.
And sixth, rapid growth in education tends towards manifest societal conflict, unless it is balanced by economic development. Development diffuses along multiple structures of expectations, opportunities, and crosscutting divisions the effects of the rapidly changing interests and capabilities produced by education. When such economic development does not keep pace, the probability of violence increases.
As a final observation, note that I have not mentioned the workhorses of change-conflict theorists: rapid technological change, population growth, resource depletion, and the like. This is because these objective changes must be mediated through our subjective perspective, our meanings, values, norms, statuses, and class. Technology, resources, and so on have social impact only in relation to the subjective sociocultural field. The keys for understanding this impact are the balance of powers, the structure of expectations, and what disrupts it. All else is of lesser importance for understanding conflict.
The form of power that shapes society, whether coercive, authoritative, or exchange, affects the type of conflict manifested. This will be shown in the next section. Regardless of the dominant power in society, however, all contemporary political systems must rely on at least some coercive power. All use sanctions to maintain civil and criminal laws. All coercively collect taxes. All maintain police and military forces, which in times of rioting or disorder will be used to disperse or control protestors. What distinguishes an exchange from a coercive society is not the presence of coercive instruments, but their extent and application. Coercive societies are totally regulated and mobilized by coercive power; exchange societies primarily use coercive power to maintain order and contain or manage disorder.
Regardless of the society, the application of force at moderate levels in a conflict will stimulate more conflict. The killing or jailing of opponents of protestors will create martyrs, a greater feeling of injustice, and a counter reaction. Such force may also signal a weak regime and the loss of legitimacy. It will communicate the issues and stimulate sympathy for the anti-government cause. Government interference in strikes on behalf of business or attempts to control or disperse antigovernment or anti-war demonstrators have repeatedly had this effect.
However, above a certain point, coercion and force become effective. If one is sure to be jailed for protesting or killed for organizing opposition, then the cost of overt opposition is too great for most. Repression, terror, and severe sanctions will suppress anti-government behavior. Therefore, we should find that such behavior occurs most in the middle range of coercion. This does not imply that full coercion eliminates manifest conflict. It still exists, and at a most intense level. But it is violence of the regime against the people that manifests conflict. Our data here, however, mainly bear on anti-regime or apolitical conflict.
(1) In Hibbs (1973) final path analysis models which he determined through numerous regression analyses, negative sanctions remained as one of the significant influences on manifest conflict. His equations show that both collective protest and internal war evoke governmental coercion, and that this in turn stimulates more mass violence. Indeed, collective protest met with repression tends towards internal war. However, while the short-term effects of repression may be higher violence, more repression tends to deter internal war in the long run. Hibbs defines repression as acts of negative sanctions, such as acts of censorship against mass media and the like, and restrictions on political activity. This measures more a difference between libertarian on the one hand, and authoritarian and totalitarian regimes on the other,
15and is far afield from police repression, jailings, torture, and so on. However, there is enough of a correlation between such repression and Hibb's measure to say, Conclusion: Government coercion against collective protest (turmoil) stimulates more such conflict, but the greater the coercion against internal war, the more it is suppressed in the long run.
(2) The Feierabends (1973) scored eighty-four nations on their level of coerciveness for each year, 1945-1966. Their scale covers basic freedoms and civil rights as well as freedom to oppose the regime. As such it clearly distinguishes between libertarian regimes on the one hand and totalitarian and authoritarian on the other, and only inferentially reflects repression or coercion in my sense. The Feierabends found a curvilinear relationship between political instability and coercion, with the highest instability in the middle range. In another analysis, using an actual count of coercive acts (which is closer to our concern with coercion), they found that high levels of governmental coercion deter manifest conflict and with one-month lag. Conclusion: Coercion is curvilinearly related to manifest conflict.
(3) On somewhat different conflict data, the Feierabends (1972) found a curvilinear correlation between political instability and coercion (eta = .72, versus a linear correlation of .41). This curvilinearity also appears in their cross tabulation between coercion and violence given in the Feierabends and Nesvold study (1969). Conclusion: Coercion is curvilinearly related to manifest conflict.
(4) Gurr (1972) measured coercion as the military and police forces available to a government. For 114 states and their conflict data, 1961-1964, he found a curvilinear correlation between these data and coercion.
16Clearly, there is an inferential leap to a conclusion about coercion as repression. Nonetheless, the correlation between such repression and available coercive forces suggests the Conclusion: Coercion is curvilinearly related to manifest conflict.
Conflict is manifested through the balance of powers comprising society and the use of coercion. Another form of conflict is the class struggle. In fact, societal conflict is mainly class-conflict-a contest between the ins and outs, the supporters and the opponents of the status quo. It is a conflict over authoritative power. When the main balance of powers in society involves governmental power, as in all modern societies, then class-conflict is the major cleavage. For such conflict is over the general balance of powers, over who gets what. Thus, we should find that class conflict in my terms is a predominant feature of conflict across societies. Some studies mentioned in the previous section already bear positively on this [Section 35.2, paragraphs (4)-(6) and (9)]. In addition, are the following.
(5) Eskola (1970) factor analyzed the perceptions of economic, political and social contrasts by 740 persons from twenty-seven Finnish municipalities. He found three major factors: opposition between ruled and rulers; party-political contrasts (e.g., right versus left); opposition of new and old (urban versus rural, youth versus aged). The greatest in magnitude were contrasts of rich versus poor, right versus left, communists versus others, and workers versus bourgeoisie. Moreover, supporters of leftist parties perceived the greatest opposition between rulers and ruled. Conclusion: Finns perceive class division as a major political and social contrast.
(6) Taft and Ross (1969) found that labor violence in the United States accounted for a large amount of U.S. conflict. Just from January 1, 1902, to September 30, 1904, 198 were killed and 1,966 injured in labor violence. A gross underestimate of killed in all labor disputes is 700, with over 160 interventions of state and Federal troops. Conclusion: Class manifest conflict has been a significant part of U.S. history.
(7) Gurr (1969) charted the participation of different groups in conflicts (p. 584-85). Working-class and middle-class groups participated in the majority of turmoil; working class groups participated in all internal wars. Conclusion: The subordinate class of workers participates in most manifest conflict.
Such is the positive evidence for the class basis of conflict, where class is understood as a division between those who obey or command. So far little empirical research has been done on the question, and the one study directly applied to this conception of class has been wide of the mark. This is Lopreato's (1967) test of Dahrendorf's theory of class conflict, which regarding class at least, is close to my own.
Lopreato analyzed a sample of 780 public opinion interviews of Italians, 211 from the command class and 569 from those who obey. He found that only a small percentage (15 and 9 percent, respectively) were aware of a two-class division, and among those only 10 and 8 percent, respectively, came close to seeing this dichotomy in Dahrendorf-like class terms. But the actual subsample involved is too small (three and four Italians, respectively!) to consider reliable. Finally, among the total sample of 780, 29 percent of the command class and 45 percent of the obey saw class relationships (however they perceived classes) as hostile. Lopreato concluded from this that Dahrendorf's theory has little validity. Conflicts follow the articulation of heterogeneous interests, although at least in Italy, social conflict "is a real and bold property of social relations" (Lopreato, 1967, p. 292).
I question the validity of Lopreato's analysis for two reasons. First, class is a latent aspect of coercive organizations, and its manifestations should be particularized to such organizations. Such has been the case in the manifest development of dichotomous union versus management-owner classes in Western industrial organizations. Where such coercive organizations exist in a pluralistic society, where class divisions crisscross society, we should not expect a particularized class view within a quasi-coercive organization to be generalized to the state-society. The place to look for such a generalization is the society-turned-antifield, the coercive state. In totalitarian states, the command-obey dichotomy is a cleavage superimposed on all major organization and the state itself. It is there that the we-they, ins-outs opposition should be most perceived. In sum, Lopreato erred in choice of level and society. At the state level he should have analyzed data from a coercive society; and for the pluralistic society he did analyze, he should have dealt specifically at the coercive organizational level.
This one study aside, we find considerable evidence for the proposition that power shapes conflict. We find that as the state grows in power, collective social conflict becomes more concerned with state power. We find, as will be shown in the next section, that conflict takes different forms depending on the type of society that is defined by its power basis. We find that one type of power--coercion--is curvilinearly related to the level of conflict. And we find some support for the class basis of contemporary social conflict. In conclusion, then, conflict at the societal level is a phenomenon of power. That which shapes the balance of powers--the interests, capabilities, and wills of participants--molds the form that conflict will take. Indeed, to understand power, its balances, and its balancing, is to comprehend social conflict.
The "therefore" is based on this logic. Conflict manifests the rebalancing of power following a breakdown of structures of expectations and the reformulation of new expectations. In exchange societies, the constant hubbub of balancing which crisscrosses society creates a relatively continuous turmoil, although its intensity and scope will fluctuate. However, these societies lack a rigid cleavage and pervasive conflict front. Therefore, conflict remains largely nonviolent. In authoritative societies, cleavages can occur easily along communal lines or around contenders for authority. There may be turmoil, but also revolutions, civil war, and guerrilla war. Finally, in coercive societies, extreme terror and force repress collective manifest conflict. Conflict, if any, usually is within the elite or by the elite against the populace. Therefore coups, elite instability, and purges are common among coercive societies. Conflict across societies, then, should be manifest in a space of three components, each of which represents the underlying power-basis and is a continuum along which society varies.
The operational question is one of determining whether such components exist. Here, the logic for determining underlying components as probability densities in the manifest covariance of social attributes can be applied, as was done in Section 35.1.
(1) In my (1965) study, I factor analyzed (component analysis) 1962-1963 data on demonstrations, major government crises, purges, riots, bombings, small scale terrorism, small-scale guerrilla war, plots, coups, and number killed in 105 states; and also separately analyzed 1963-1964 data. The results of both analyses were compared to twelve other factor and component analyses involving at least some domestic conflict data. Here I will present only the results of the comparisons.
18These showed turmoil-largely riots and demonstrations-to be a consistent component of cross-state domestic conflict. A second component which I called revolution and which largely consisted of purges, coups, and plots, appeared in some of the analyses. A third component also appeared in some of the analyses. I called it subversion, which was largely made up of guerrilla activities and terrorism. Revolution and subversion appeared to be distinct, but their separate natures were combined as a single dimension in some studies. Conclusion: At least two, and possibly three, components of domestic conflict exist, one of which is turmoil.
(2) Bwy (1972) did a component analysis of nine kinds of conflict manifestations for twenty Latin American states and found an "anomic violence" (turmoil) component (riots, anti-government demonstrations, and general strikes), an "organized violence" (revolution plus subversion) component (guerrilla war, government crises, purges, revolutions, and killed), and a "lack of conflict"
19component (assassinations highly negative, and general strikes and revolutions moderately positive). Conclusion: Three components of conflict exist for Latin America, one of which is turmoil.
(3) Hibbs (1973) did component and common factor analyses of five conflict attributes for 1948-1957 and for 1958-1967. The five were riots, antigovernment demonstrations, political strikes, assassinations, armed attacks, and deaths from political violence. He consistently uncovered a "collective protest" component (riots, anti-government demonstrations, and political strikes) and an "internal war" (deaths, armed attacks, and assassinations) one. He did not include coups, revolutions, purges, and subversions, which in number and type should have helped to define a third component.
20Conclusion: At least two components of conflict exist, one of which is turmoil.
(4) The Feierabends (1966) did a component analysis of thirty attributes for all states on 1948-1962 data. Not all were conflict manifestations in my terms, but of those that were they found the following components: "turmoil" (microstrikes, general strikes, macro-strikes, micro-demonstrations, macrodemonstrations, macro-riots, severe macro-riots, mass arrests of insignificant persons, and terrorism plus sabotage); "revolution" (arrests of significant persons, martial law, coups, and revolts); "purge" (vacation of office, arrests of insignificant persons, executions of significant and insignificant persons); "riot" (micro-riots and severe macro-riots, and mass arrests of insignificant persons); "demonstration" (micro-demonstrations, arrests of insignificant persons); "imprisonment" (crises within a nongovernmental organization, imprisonment of insignificant persons); "civil war" (only civil war loaded above an absolute .50), and "guerrilla warfare" (assassination, guerrilla warfare). The largest components were"turmoil," "revolt," and "purge." Clearly, when many different kinds of conflict manifestations are included, the effect is to bring out the mixed varieties (as there are mixed societies and political systems). Nonetheless, taking into account the consistent independence of turmoil, revolutionary and subversion type dimensions, and the nature of the largest components leads to the Conclusion: Three components of conflict exist, one of which is turmoil.
(5) Banks (1971, 1972) did component analyses of eight conflict attributes for all states and for each of the years 1922, 1929, 1936, 1949, 1956, and 1963. "In every case, a three-factor structure emerged, the components of which could roughly be interpreted as delineating Rummel's turmoil, revolutionary and subversive dimensions" (p. 7). He then did a time series component analysis of 1922, 1929, and 1936, and another of 1949, 1956, and 1963. Three components appeared for each overtime series, but the conflict loadings changed appreciably between the periods. A rerun over the six years, however, brought out the three components found in the separate cross-state analyses. Banks concluded that a "basic distinction between 'organized' and 'unorganized' domestic conflict behavior is empirically tenable and remains valid longitudinally, at least for the contemporary era subsequent to World War I. A case can be made for a further distinction between 'revolutionary' and 'subversive' aspects of organized conflict behavior, but these dimensions do not remain equally stable throughout the same time-span." (p. 10). Conclusion: Three components of conflict exist, one of which is turmoil.
(6) Morrison and Stevenson (1971) did component analyses of twenty-five conflict attributes for thirty-two African states for each of these periods: first six years after independence, 1960-1964, 1965-1969, independence to 1969. They found five components in each analysis, and their loadings indicated that turmoil is an independent component, and that two additional ones are "elite instability" (plots, coups, attempts, elite instability) and "communal instability" (mutiny, civil war, rebellion, irredentism, ethnic violence, communal instability, and revolt). Conclusion: Three components of conflict exist for African states, one of which is turmoil.
(7) Wilkenfeld (1969) did a component analysis of seventy-four states for nine conflict attributes. He found independent turmoil and "internal" war components, the latter consisting of guerrilla war, government crises, revolution, purges, killed. Conclusion: Two components of conflict exist, one of which is a merging of subversion and revolution dimensions and the other of which is turmoil.
(8) Firestone (1969) did a time series component analysis (super-P) of nine conflict attributes for seventy-four states over nine different periods, 1955-1966. He found three components: turmoil, domestic planned violence, and number killed. Conclusion: Three components of conflict exist, one of which is turmoil.
The above constitute the major empirical evidence for the existence of three components. The consistency with which a turmoil component appears in different studies on different data for different years, and states, and regions is remarkable. Moreover, although there is some ambiguity about their nature, and sometimes they merge into one or spread across more than two components, the tendency is for two additional components to appear, one apparently reflected in elite instability and the other in communal, and authority based conflicts. The evidence confirms the theoretical proposition that conflict manifestation comprise a three-component space.
There is, however, one additional question. I have shown a three-component space to represent phenomena, but I have not shown that each component reflects the appropriate society. Remember that power shapes conflict. The conflict space therefore mirrors the different power bases: exchange, coercive, authoritative. The only appropriate evidence for this is the following.
(9) In Banks' (1972) aforementioned longitudinal component analyses, he did a Q-factor analysis of all states on the eight conflict attributes for 1922-1963 data. Four major groups emerged, the largest of which was developed Western and which included as highest loaders (in decreasing order) New Zealand, Canada, United States, Australia, United Kingdom, Sweden, Panama, France, Italy, and Belgium. The second group consisted of mainly authoritative states, such as (in decreasing order) Honduras, Paraguay, Saudi Arabia, Afghanistan, Costa Rica, and Albania. The third group was a subset of Western "crisis" states, involving Luxembourg, Netherlands, Finland, Norway, and Chile. Finally, a group of pure coercive states emerged, comprising U.S.S.R., Hungary, Yugoslavia, Rumania, and Bulgaria. In effect, we find states clustering in conflict space as we would divide them into societal and political system types. Conclusion: Three conflict components reflect the three societal and political system types.
In summary, the existence of three components of manifest conflict corresponding in nature to what we would expect from the three types of societies has been substantiated, and Banks' grouping analysis verifies that this conflict space does divide state-societies into these three types. Conflict is shaped by power in three directions, along three dimensions.
(1) Adelman and Morris (1973) did a discriminant analysis of political participation in seventy-four least developed states, 1957-1962. Their measure of political participation reflects whether a more or less exchange type society exists. The best gauge among a wide range of socioeconomic attributes was mobility (F = 17.6). Alternatives were political stability
21(F = 3.4) and lack of social tension (F = 2.9). Moreover, mid-range developing states, and the highest developed of the developing states, have their levels of political participation separately discriminated by political stability, among other attributes. Conclusion: Comparatively, exchange societies manifest the least intense conflict.
(2) Gurr (1969) charted the objectives among conflict groups for 114 states, 1963-1968. The objective of seizing political power varies by economic development; it is low for developed states and least for polyarchic (libertarian) systems. Assuming that the aim to seize political power leads to the most violence, we have the Conclusion: Violence attendant upon trying to seize political power is the lowest for exchange systems.
(3) Sorokin (1957) in his quantitative time series analysis of internal disturbances in the history of Greece, Rome, and Europe found that: "At its height, therefore, the 'capitalistic regime' . . . was the most orderly of social systems and gave the greatest assurance of internal and external peace and of Sensate liberty and freedom for individuals" (p. 594, italics omitted). Conclusion: Exchange societies have the lowest aggregate manifest conflict.
(4) Using a scale of democracy (developed by Russell Fitzgibbon) for 1955 and twenty Latin American states, Bwy (1972) found it correlated -.50 with 1950-1960 organized manifest conflict. He also found the 1950 to 1955 change towards democracy along this scale correlated -.71 with organized conflict. Moreover, the change in Cutright's scale of democratic development towards democracy 1950 to 1955 correlated -.44 with anomic conflict (turmoil) 1955-1951 and -.61 with organized conflict. Conclusion: The more a society is exchange (libertarian), the lower its manifest conflict.
The final question is whether this low conflict in exchange societies is really along a turmoil component.. I know of no studies bearing on this except the following.
(5) Banks (1972) provides a table of the states with the highest cumulative conflict totals separately for assassinations, general strikes, guerrilla wars, major government crises, purges, riots, revolutions, and anti-government demonstrations. When these totals are matched against the four groups he determined by Q-factor analysis [Section 35.4, paragraph (9)], the exchange-libertarian type group is mainly a nonrevolutionary group with riots and demonstrations, i.e. turmoil. Conclusion: Exchange societies mainly manifest turmoil.
In Conclusion, then, exchange societies are more peaceful than authoritative and coercive ones. Moreover, when they manifest conflict it is usually protests, strikes, riots, and demonstrations. The reason for this has been given previously. Exchange societies bleed off conflict in divergent directions, preventing that fatal congruence of cleavages and oppositions that leads to intense struggles over societal powers and consequent extreme violence.
(1) In Banks' (1972) aforementioned Q-factor analysis, one of the state-groupings he found in the conflict space was of authoritative societies. He named this a revolutionary group, for many members had high accumulative totals for revolution. Secondly, Banks regressed the loadings for fifty-one nations on this group onto national government revenue and expenditure per capita, energy consumption per capita, secondary school enrollment per capita, university enrollment per capita, percent literate, physicians per capita, and gross domestic product per capita. The multiple R was .874 (considerably high for fifty-one states), with all negative correlations between these independent attributes and the group loadings. Assuming that authoritarian societies are generally low on these attributes, these results clearly show that the more traditional-authoritative societies are the more revolutionary. Conclusion: Authoritative societies tend to be revolutionary states.
(2) Huntington (1968, p. 408) displayed a chart of coups and coup attempts by type of political system. The noncommunist states with no effective party i.e., authoritative states, had more coup or attempted coups than any other states. Conclusion: Authoritative societies tend to be most coup prone.
(3) One of the components Gregg and Banks (1965) found for the Cross-Polity Survey data was bipolar, opposing personalismo political system, domestic killed in conflict, military interventive, and Latin American to system stability, Western, modern bureaucracy, political enculturation, and Westernization.
One end identifies traditional-authoritarian systems in contrast to Western and shows the manifest intensity (killed) of conflict to be one of its characteristics. Conclusion: Authoritative societies have the most intense manifest (antigovernment) conflict.
In sum, the evidence presented above in conjunction with that for the existence of the three components of conflict (Section 35.4) support the proposition that authoritative societies manifest a particular form of communal/ traditional conflict. When the legitimacy of traditional rulers--the bond holding the balance of powers together--is eroded or destroyed, violence is the means for establishing a new legitimacy.
(1) Conquest (1968) carefully considered casualty figures for Stalin's rule, 1930-1950, and arrived at an estimate of twenty million dead, "which is almost certainly too low and might require an increase of 50 percent or so" (p. 533). For comparison, the casualty figures of the First World War were approximately 10 million. It is remarkable that this war, the concern of so many humanists, historians, and social scientists, should have a casualty list that is half of that of one coercive society for twenty years. Conclusion: Coercive societies manifest violent elite repression.
(2) There is no need to reference the approximate 6 million Jews killed in cold blood by Hitler and his regime. Conclusion: Coercive societies manifest violent elite repression.
(3) Figures on the Chinese death toll under Mao Tse-tung have yet to gain widespread credibility. The U.S.S.R. in 1969 put the figure killed at 26.3 million from 1949 to 1965. Peking radio gave one monthly (!) figure killed of 2.4 million. A Minister of Finance allegedly stated in a publication that over 2 million were killed in a three-year period. A French diplomat estimated executions at between I and 3 million in over a one-year period. Taiwan's government estimated those killed, 1949-1969 at 40 million, excluding the Cultural Revolution.
22And a report to the American Senate Judiciary Committee put the number of deaths between 32 and 62 million. 23No doubt the cost of repression in China has been high, most likely exceeding the death toll in the Soviet Union. We do not require a consensus on the exact figures for the Conclusion: Coercive societies manifest violent elite repression.
Elite repression against those ruled minimizes overt conflict against the regime. It is one of the ironies of quantitative cross-national analysis, therefore, that communist regimes emerge from the data as relatively peaceful, and that the dummy variable, communist or not, predicts to less rather than more violence. Repression has its effect.
(4) Hibbs' (1973, p. 171) path analyses show that the existence of a communist regime, among other attributes of states, will predict negatively (t = - 5.65) to collective protests. and negatively (t = - 4.42) to internal war. Conclusion: Coercive societies repress conflict manifestations.
Additional evidence on the conflict repressing effect of coercion is given in Section 35.3. A certain type of manifest conflict does occur, however, among the elite; its visible traces are executions and purges. Elite instability and factionalization is the chronic problem of totalitarian regimes. Does elite instability manifest itself in the data?
(5) In Banks' (1972) aforementioned Q-factor analysis of states on conflict data, 1922-1963, most communist regimes exclusively grouped among themselves in this manifest conflict spaces. They all shared a conflict profile of low conflict and many purges. Conclusion: Coercive societies manifest mainly elite instability.
The empirical evidence is sufficient to support the proposition that coercive societies manifest elite repression /purges.
These propositions are substantiated beyond the evidence presented in this chapter when they are imbedded within the taxonomy of societies and political systems tested in the Chapter 34. Indeed, empirical analyses following the logic of the conflict process equation establish the empirical validity of my distinctions among exchange, authoritative, and coercive societies; also among libertarian, authoritarian, and totalitarian regimes. And empirically the political systems are congruent with the societies, as assumed theoretically.
Three forms of power, three balances, and structures of expectations. Therefore, three societies. Three congruent political systems. And three components of conflict.
"Tir[esias] ... By the fates which spun thy thread! Cho[rusl. Which are three."
---- Dryden and Lee, Oedipus III.i
* Scanned from Chapter 35 in R.J. Rummel, The Conflict Helix, 1976. For full reference to the book and the list of its contents in hypertext, click book. Typographical errors have been corrected, clarifications added, and style updated.
1. The logic concerning underlying components and structures of expectations can be developed for regression analysis, as I will do for a future volume on international relations (Vol. 4: War, Power, Peace). In brief, the least squares line represents a structure of expectations, and the residuals reflect the degree of incongruence of structure and underlying balance of powers.
2. A paragraph is often given to a study involving extended and massive data analyses, and requiring numerous volumes for presentation of the results. Therefore, the brevity of my summary is no index to the depth and scope of the analyses described. All those who are cited here will appreciate this.
3. A loading of .58 on an orthogonal dimension would mean 66 percent of the variance between component and conflict behavior is statistically independent.
4. The number of states is not given in Gregg and Banks (1965). Since their Q-factor analysis was of the same data involved in Banks and Gregg (1965), which was on 115 states, I assume this is the correct number.
5. See Tilly (1973). See also Cattell's (1953) time series factor analysis of the United Kingdom, 1837-1937, and Gibb's (1956) similar analysis of Australia, 1906-1946. Both found a growth factor, partially involving strength of police.
6. I am talking about the scope of conflict and not its over-all intensity. Levy (1969, p. 92) found that the present period has much lower violence (killed) while injuries are much higher.
7. See the excellent empirical analysis of Pride (1970), which shows what can be done when scientists free themselves from the dictation of standard methods. I am assuming that his political system measures are congruent with the type of society, as shown in Section 34.4 of Chapter 34.
8. Exceptions to this will occur when development only translates into more power and wealth for one communal group at the expense of the others. As in Northern Ireland, this then deepens the communal cleavage and lays the basis for civil war.
9. For evidence that different societal structures may change at different rates or follow different tracks, see Banks (1974) and Zapf and Flora (1973).
10. My reasons derive from an analysis of the frustration-aggression hypothesis in the next volume (Vol. 3: Conflict In Perspective).
11. The numerator includes attributes highly intercorrelated with a wealth or development dimension. See Rummel (1972, 1979).
12. About literacy there is no question. Urbanization roughly measures the practical and institutional education developed through the greater opportunities, communication, contacts, and over-all diverse social experiences an urban area provides.
For empirical evidence, see Banks' (1974, Table 9) time series factor analysis of twelve socioeconomic attributes, which results in urbanization having the highest loading on the school enrollment factor. See also Gibb's (1956) similar time series analysis of Australia which shows the number of nonrural employed to be loaded on factors with government expenditure on education and number of university students; and Cattell and Adelson's (1951) similar analysis of the United States, which showed population in urban areas and percent population attending school loaded together.
13. There is some ambiguity as to whether they use levels or rates. From the context, I assume levels. Voting participation I also assume measures educational levels, which has been a consistent finding of voting studies. See Berelson and Steiner (1964, p. 423).
14. Some libertarians would eliminate even such limited government coercion. Here, I am concerned with existing state-societies, all of which maintain and apply some coercive means.
15. His definition leaves unanswered whether there has to be an act of censorship, say, or whether a continuing policy of no freedom of the press counts. If not, then a mildly authoritarian regime which occasionally takes action against a usually free newspaper is more repressive than a totalitarian regime which allows no press freedom and is thus without a free press to take action against.
16. See also Gurr (1969).
17. A common error is to assume that because a set of variables form a single dimension when analyzed along with other variables, the set when analyzed by itself should form one component. Thus, it may appear that the one domestic conflict component found in the midst of diverse state-attributes provides evidence against the existence of these conflict components. However, the covariance within a set of variables can be submerged in the larger matrix. It is similar to viewing mountains from 40,000 meters, where the separate mountains merge into one range which stands in contrast to the plains and oceans. Our concern, however, is with the 1,000-meter view of the mountain range alone, such that the individual mountains can be seen and mapped.
18. The studies were Rummel (1963, 1966), Tanter (1964, 1966), Cattell (1949), Cattell, Bruel, and Hartman (1951), and Chadwick (1963). Some involved more than one factor analysis.
19. Bwy clearly misinterpreted the meaning of negative loadings.
20. Hibbs did not indicate his factor cutoff, so that I cannot assess whether he has a third "meaningful" component he did not extract or rotate.
21. For Adelman and Morris, instability is characterized in part by significant domestic violence. Social tension is measured by considerable violence and social tension. Clearly, instability and social tension overlap.
21a. [Note added for web site edition, July 3, 1998] The following statistics on genocide and mass murder that I gathered for this section simply shocked me. I had never realized that this much killing in cold blood had taken place. I had taken up the study of war because of the massive killing involved, but these statistics showed that even more people were murdered by governments. I could not at this time undertake a separate research project on this killing, but when ten years later I finished my research and writing on collective violence and war, I finally was able to more systematically collect data on government murder. I was then able to determine in fact how many people probably were so killed, and if the conclusions about freedom and other forms of violence also held up for genocide and mass murder (which later I would call democide). This web site displays the results.
22. For all these estimates, see McWhirter and McWhirter (1974, p. 388).
23. Titled "The Human Cost of Communism in China," it was prepared for the Senate Committee on the Judiciary, 92d Congress, lst Session, 1971 (U.S. Government Printing Office Stock No. 5270-1138), by Richard L. Walker, director of the Institute of International Studies at the University of South Carolina.