1. Introduction and Summary
Democratic Peace page
An honorable man will not be bullied by a hypothesis
---- Bergen Evans, The Natural History of Nonsense, XIX
Chapter 33 presented my logic for empirically defining sociocultural space and determining the structures of expectations and the location of states in this space. The key element here is the space-time latent function, the common component delineated by observations--manifest attributes.
In this chapter, these empirical components will be delineated first. Then, two theoretical ideas will be assessed using these components. The first, that states vary along two primary political dimensions, producing a political triangle congruent with the different types of societies. The second, that at the state level three types of societies, exchange, totalitarian and authoritative, exist.
I have analyzed several hundred such attributes in line with Equation (33.4) of Chapter 33
These have been simultaneously analyzed (using the image model) for all states and the years 1950, 1955, 1960, 1963, and 1965 to uncover the fundamental space-time social components. I cannot present the design and details here,
The attributes of states manifest the following independent space-time components. First, there is the society's wealth, reflected in per person energy consumption, GNP, telephones, and calories consumed. Moreover, this space-time component also is the basis of variance among states in percent agricultural population, dwellings with water, illiteracy, and pupils per teacher. At its broadest, wealth (which some may conceptualize as modernization or economic development)
What do I mean by saying that wealth underlies ... ? Wealth can have two compatible meanings. First, some regions of potentialities, dispositions, and powers which forms the foundation for covarying manifestations have a certain common quality, that is, they all seem to manifest a wealth versus poverty continuum. Therefore, the space-time region or component is called wealth. Second, we may consider these commonly covarying manifestations to be the product of an unknown, underlying, and independent function of the complex multifold relationships among potentialities, dispositions, and powers. By virtue of a common quality (wealthy versus poor) of these manifestations we conceptualize this function as wealth.
Moreover, wealth also is an underlying structure of expectations. The general logic is discussed (and exemplified) in Section 33.3. No more need be added here than that the wealth of a state consists of a system of norms, rules, understandings, and expectations, a status quo distributing rewards and deprivations, and based on a particular balance of powers. Wealth and poverty are imbedded within a structure of powers.
As a space-time component of sociocultural space and a structure of expectations, then, how does wealth relate to the assumptions specified earlier? First, it is inferred via image factor analysis from manifestations (on the nature of factor analysis, see "Understanding Factor Analysis"). Second, it is defined relative to
Third, wealth was defined in the midst of all the covariances over space and time among hundreds of attributes.
To move on, another space-time component underlying societal manifestation is coercive power capability. This is manifested in the product of a society's energy production and people, and in its national income, defense expenditures, and population. These reflect the raw force a nation's political elite have at their disposal either to coerce other societies, to use force against them, or to protect itself against such power. That this interpretation is the proper one is indicated by the fact that this cluster of manifestations also includes threats made by the elite against other states, as well as protests and accusations (Rummel, 1979, Table 5.6). Thus, coercive power is a second, and independent, space-time component of sociocultural space--a second structure of expectations within states.
In wealth and coercive power we have found two status, space-time components (dimensions) of sociocultural space, as we should. They are status components because there is a consensus as to their desirable end. Wealth and power
These two components, then, delineate the status potentialities of sociocultural space for state-societies. While power is, however, pure status, wealth does double service. It is not only a status component, but it also forms a region of potential meanings, values and norms. Wealth is manifested through industrialization, science and technology, urbanization, education, and rapid mass communication, and therefore represents the more secular, sensate, technologically and scientifically oriented societies as opposed to the traditional and ideational.
A third space-time component, Catholic culture (Rummel, 1979, Table 5.9), is manifested in the proportion of Roman Catholics in the population, the percent divorces, the geographic location of the society, and the proportion of medical nonintergovernmental international organizations citizens join.
In addition to wealth, power, and Catholic culture, are two space-time components and structures of expectations of sociopolitical attributes. The first is totalitarianism (Rummel, 1979, Tables 5.2 and 5.3), which is manifested through communist regimes, lack of constitutionalism, a population mobilized by the political elite towards a future goal, complete censorship, and inability of groups to oppose the government. The second is authoritarianism (Rummel, 1979, Tables 5.4 and 5.5), which is determinate through one-party rule, vertical power distribution, restrictions on freedom of political opposition, noncompetitive electoral system, military participation in government, and censorship. Both components are correlated, of course, but discriminated at the authoritarian and totalitarian ends are such societies as Spain versus the Soviet Union, Saudi Arabia versus Yugoslavia, Afghanistan versus Albania, Haiti versus Cuba, Taiwan versus China, and South Korea versus North Korea.
Some other space-time components (Rummel, 1979, Tables 5.8 and 5.12) underlying covariance in societal attributes are density (people to area, railroad length to area, percent arable land, etc.), size (population, national income, area, etc.), and import dependency (imports to trade, exports to GNP, aid received). These components also form the basis of differences in sociocultural meanings and expectations, and describe the physical--objective--aspects of society, as well as define the subjective potentialities of sociocultural space and the state-society's structures of expectations.
Finally, one more space-time component, social conflict (Rummel, 1979, Table 5.11), delineates a cluster involving general strikes, riots, number killed, purges, government crises, and demonstrations. Analyses (Rummel, 1972, Table 9.15) show this cluster to also involve assassinations, revolutions, and guerilla war.
In summary, I have identified the following space-time components:
All except the last have been identified by others in cross-sectional analyses of states and should be considered highly reliable spatial components.
With some scientific confidence, therefore, we can accept the sociocultural space of states as delineated by the aforementioned space-time components. In addition, we can also assume that these components (with the conflict component excluded) comprise structures of expectations.
(1) The 1950-1965 space-time analysis (ninety-one attributes for all states): As mentioned above, two independent (but correlated) common components
totalitarianism and authoritarianism were uncovered. Of most interest is that liberal-democratic systems, ranging from the more socialist of the Swedish and Danish type to the more unregulated
(2) Ten political attributes for all states: I have done a component analysis of data for all states on the following ten political attributes: system style (whether or not a regime mobilizes the population to achieve some future directed goal), constitutional status, representative character, electoral system, freedom of group opposition, noncommunist regime, political leadership, horizontal power distribution, monarchical type, and military participation.
Two common components emerge:
Do these empirical components define the clusters of systems in political space that we would expect, given the theoretical political triangle of Figure 31.2? A grouping analysis (Rummel, 1970, p. 506)
Conclusion: The theoretical political triangle does reflect an empirical ordering of political systems for the ten political characteristics.
(3) Sixty-eight political attributes for all states: Gregg and Banks (1965) did a component analysis of sixty-eight political attributes for all states, including the ten analyzed in the previous section (data circa 1960), and found a totalitarian component, the largest in the data (34 percent of common variance), which opposed liberal democratic regimes to totalitarian. Another component was authoritarian,
Additional components uncovered aspects of political systems independent of the three regime types, such as whether they are developed democratic or undeveloped democratic, stable or unstable, two-party or multiparty, or modernizing authoritarian.
In another study, Banks and Gregg (1965) clustered nations on the same sixty-eight political variables, using Q-factor analysis. Five independent clusters (groups) appeared, the largest of which they called polyarchic and which corresponds to a libertarian system. As the most central members this cluster includes Norway, Luxembourg, Ireland, West Germany, Sweden, Australia, Netherlands, Iceland, Denmark, New Zealand, Austria, and Finland.
Another cluster they called centrist, involved all communist regimes. Clearly, this was the totalitarian cluster. A third cluster, which they labeled traditional, was an authoritarian type comprising Yemen, Nigeria, Laos, and Iran.
Finally, there were two mixed clusters. One, called elitist, appears to be a mix of authoritarian and totalitarian regimes in which elites are trying to force radical socioeconomic change on a traditional society. Its central members are Niger, Central African Republic, Dahomey, Gabon, Upper Volta, Ivory Coast, and Congo (Brazzaville).
The second mixed system, labeled personalist, is mixed libertarian and authoritarian, in which there is "a high degree of political personalism as the term is normally construed in the Latin American context" (Banks and Gregg, 1965, p. 4). The central members of this cluster are Guatemala, El Salvador, Panama, Peru, Honduras, Argentina, South Korea, and Syria.
Conclusion: The two Banks and Gregg analyses add additional validity to the political triangle.
(4) Forty-one socioeconomic and political attributes for seventy-four states: Adelman and Morris (1967) did a number of component analyses of attributes for seventy-four least developed states on early 1960's data. All communist states, as well as those considered developed, were omitted. Their sample is therefore skewed for our purposes, especially in omitting the totalitarianism component.
Nonetheless, for all their states and attributes they found an independent political system component, even when the analyses were done by region and by level of development. Moreover, for all their states they found two political components. One delineates the strength of democratic institutions, freedom of political opposition, and competitiveness; the other, the strength of traditional elite and opposition to development. One component is clearly libertarian; the other reflects an aspect of authoritarian regimes.
Conclusion: Libertarian and authoritarian dimensions are empirically distinguishable for different levels of wealth and different regions.
(5) Other studies: Totalitarianism and authoritarianism share some common features, such as disallowing political opposition or competitive politics. Therefore, as components both are correlated. Unless a range of political attributes are included and an oblique rotation done, only one independent totalitarian
authoritarian component may emerge. Such was the case with the different cross-national analyses of Russett (1967), Cattell (1949), Cattell, Breul, and Hartman (1951), Cattell and Gorsuch (1965), and Rummel (1972).
(6) Conclusion: Before discussing the strength of the conclusion, let me briefly review the logic. Attributes are manifestations of underlying common components, which form societies' structures of expectations. Theoretically, three distinct balances of power (structures of expectations) comprise the political systems of states. These balances are alternatives--vertices of a triangle in a political space spanned by two components. All this is theory.
If the theory is correct empirically, the covariation of data on diverse sociocultural attributes should cluster into two distinct political components, which as structures of expectations are independent of other, components (such as wealth or density) delineating social space. The best method consistent with the theory for determining this is common factor analysis and oblique rotation. Component analysis is an approximation, as is orthogonal rotation.
In all studies using common factor analysis and oblique rotation or a wide range of political variables (which even allow component analysis and orthogonal rotation to separate the two components), the two theoretical components emerge from the data: one is totalitarianism versus libertarianism, the other is authoritarianism versus libertarianism. They comprise the political triangle. In all other studies I have found, a separate political component emerges, which opposes totalitarian and authoritarian systems to libertarian ones.
The term "emerges from" has tremendous power to it. Consider. Data on many diverse sociocultural attributes for all or most nations are thrown together. Nothing in the method or data necessarily forces out of these data an independent political component. Were attribute labels arbitrarily assigned to random numbers, the likelihood of separate political components emerging as consistently across studies would be near the probability that monkeys typing at random could duplicate this page.
Thus, the theoretical political components have high empirical validity. The pure libertarian, authoritarian, totalitarian systems have empirical counterparts. The political triangle exists.
(1) Clusters in 1950-1965 space-time. A Q-factor analysis (component analysis) was done on eighty-seven attributes and all states for each of the years 1950, 1955, 1960, 1963, and 1965.
Table 34.1 orders the clusters found for each year according to the type of society they manifest. Note that the exchange society is a consistently a separate, independent cluster among states. These are societies in which there are many associational interests, much freedom of individual and group activity, and overlapping group membership. Cross pressuring of status and class memberships are high, and social interaction is solidary and contractual.
Second, we find a large, coercive cluster independent from the exchange on the eighty-seven attributes. These are totalitarian-communist societies; little need be said of their coercive nature. This coercive type emerges consistently, despite the fact that the number of political variables specifically indexing communist regimes out of the eighty-seven is no more than eight.
Third, the authoritative society clearly emerges as a cluster in 1950 and 1955. In 1960, 1963, and 1965 the cluster breaks up due to the sharp increase in the number of states of varied and mixed authoritative characteristics, and internal changes in the states themselves.
Conclusion: The results of this clustering analysis show that distinct exchange, coercive, and authoritarian societies exist as major clusters of states in sociocultural space.
(2) On the components of 236 attributes. In my analysis of 236 attributes, I determined the clusters of nations in the space defined by the largest five components, as shown in Rummel (1972, Table 12.4). One grouping is of authoritative societies, involving Afghanistan, Ethiopia, Yemen, Iran, Outer Mongolia (1955 data), Saudi Arabia, and Albania. Another centrally involves the United States, Canada, Australia, and Sweden (also the Union of South Africa at the fringe), and is a grouping of exchange societies. The other clusters are mixed, and represent the dominance of one component or another, such as Catholicism, size, or density. Conclusion: These clusters are ambiguous about the existence of a coercive society, while for 1955 the authoritative society is empirically distinct; the exchange cluster is also present, but not with the distinctiveness of the previous analyses.
(3) Regions of sociocultural homogeneity. Russett (1967) also grouped states on their distances in the space of four components (having omitted a size component). A strong exchange cluster was found, which he called Western Community. It involves virtually all states in which exchange interaction dominates (see his Table 2.2). A second independent cluster of all coercive- communist societies was Eastern Europe.
No clear authoritative cluster emerged, although a group Russett called Afro-Asia came close. Centrally it comprises (for around 1960 data) Tunisia, Iraq, Iran, India, Malaya, Turkey, Morocco, and Thailand. There is also one mixed Latin American cluster, and another called Semi-Developed Latins.
Conclusion: Russett's study strongly affirms the independent existence of the exchange and coercive societies.
(4) Pre-World War II clusters: Cattell (1950) determined the clusters of sixty-nine states on seventy-two attributes, some of which were historical data (wars, revolutions, treaties, for example) for 1837-1937. Most data, however, were aggregated for 1927-1937. Using a coefficient he developed which measures both profile and magnitude similarity, Cattell visually determined the clusters of nations in the sociocultural space defined by these attributes. It was an intuitive Q-factor analysis.
Cattell found a number of clusters.
There are also two authoritative clusters, a "Mohammedan" cluster of Afghanistan, Iraq, Turkey, Arabia, and Egypt; and an East Baltic one of Finland, Estonia, Lithuania, Poland, and Latvia.
Finally, there are two exchange clusters, one of Scandinavia (Denmark, Sweden, Norway) and the other of New Zealand, Australia, Netherlands, Belgium, and Canada.
A totalitarian cluster does not emerge, since Stalin's Soviet Union was unique during this period. Indeed, the U.S.S.R. stands by itself in this space.
Cattell's clusters are "messy," in that his method does not assure independent clusters of states, as do those previously surveyed. Moreover, the grouping is for a period of time when virtually all states comprised exchange, mixed, or authoritative societies. Nonetheless, we do find in his social space of seventy-two attributes distinct exchange and authoritative types, which appear pure, in that there is no case where one type is included with the other.
Conclusion: Exchange and authoritative societies exist historically.
(5) Conclusion. Across these various studies we find emerging clusters of states which manifest my pure types. And these are usually the largest clusters.
Again the power of the term "emerge" should be understood, especially as it refers to various collections of attributes for different time periods. The attributes were selected and data collected without presupposition about types of societies. Moreover, the methods applied for discerning clusters in the sociocultural space are not defined to force out any particular kind of cluster. Therefore, to consistently find exchange, authoritarian, and coercive (post-World War II) societies attests to the empirical validity of these types.
In sum, we can define empirically three different structures of expectations constituting state-societies. These structures manifest common attributes which distinguish all states sharing them. Indeed these structures are sufficiently powerful to dominate over subordinate structures of expectations (such as size or density), and define unique clusters of states in the space of their diverse social attributes.
In all studies, the defining component or set of attributes of coercive societies is a totalitarian political system. Moreover, all authoritative societies found in the above studies manifest an authoritarian political system. Finally, the exchange cluster, without exception, has libertarian type political systems. In other words, the clusters of states in the space of political attributes, and the clusters of state-societies in the space of diverse sociocultural attributes are nearly identical.
In final summary, then, exchange, authoritarian, and coercive societies and their corresponding political balances of power do exist. The theoretical balances of power-structures of expectations do manifest themselves. The next set of questions now must deal with these distinct common structures of expectations and social conflict.
* Scanned from Chapter 34 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. One such analysis is given in Rummel (1972). I am interpreting "in line with" to include the component analysis model and cross-sectional analysis.
2. All technical terms used here are defined and discussed in Rummel (1970).
3. The reader referring to my cited empirical work may be struck by a difference in terminology. In all previous analyses I have used the term nation; here, for the same society, the term state. The reason lies in the meaning of nation. Previously, I was only concerned with "nation" denotatively as standing for the state which is sovereign in international law. Sometimes this society is called a nation-state, to mean both the sovereign legal group and the historical evolution towards a state encompassing one nation, or cultural system.
Connotatively, the nation as a cultural unity is a confusing concept when referring to the sovereign state, for many states today are not unified nations, but indeed are conglomerations of sometimes hostile cultures, as in Nigeria, India, or Burundi. National unity is then a matter of political coercion, and not of an inner sharing of communal norms, beliefs, and values. Indeed, it is this lack of connotative congruence between state and nation that leads to violence in many of these states.
Because culture and state are not congruent in the modern world, I have dropped the term nation and used the classical concept "state" to refer to the sovereign group.
4. These are given in Rummel (1979), which also includes the data.
5. "Wealth," however, best captures the general nature of this component and the distribution of states on it. See Rummel (1979, Table 5.1).
6. The phrase "relative to" is crucial. The principal axes of the probability ellipsoid for all the attributes were rotated to a new set of axes which experience has shown (Rummel, 1970, Chapters 16 and 17) to better reflect the separate clusters of covariance. Incidentally, I did both orthogonal and oblique rotations, and the results I am sketching here take both into account. See Rummel (1979).
7. That is, the relationships were defined by a product-moment correlation matrix (on this matrix and the correlation coefficient, see Understanding Correlation). By covariance, I mean the relationship defined by the family of correlation coefficients which include the covariance and product-moment coefficients.
8. Throughout this chapter power means coercive power.
9. From Rummel (1970, Table 9.11, p. 232) we know that this component is also manifested through a high proportion of typhoid deaths, of female students in school, and of students in law.
10. Guerilla war is involved in the periphery of the cluster and therefore does not appear in Table 9.1 (Rummel, 1972).
11. A comparison of cross-sectional studies is given in Rummel (1972, Chapter 10). These studies have not employed image analysis and only Cattell and colleagues have used common factor analysis. Possibly for this reason others have found authoritarianism and totalitarianism combined into one political orientation component, with liberal democracies at one end and authoritarian and totalitarian states mixed at the other. See also Adelman and Morris (1967).
12. It appears as a growth or development common component.
13. See, for example, Berry (1965) and Hadden and Borgatta (1965). Many additional references are given in Rummel (1970, Chapter 24).
14. I mean this comparatively. In absolute terms, U.S. society is regulated to a considerable extent.
15. On the authoritarian side of the average, these would include for 1965 in roughly decreasing degree: Haiti, Portugal, and Spain; Ethiopia, Saudi Arabia, Yemen, Burma, Ghana, and Guinea; South Vietnam, Egypt, Cambodia, Indonesia, Afghanistan, Central African Republic, Guatemala, Libya, and Mali; Jordan, Syria, Paraguay, and Algeria; and Tanzania, Mauritania, Ivory Coast, Pakistan, Taiwan, South Korea, Brazzaville Congo, Leopold Congo, Dahomey, El Salvador, Iran, Iraq, Liberia, and Niger.
Moreover, we find Burma, Ghana, Guinea, South Vietnam, Egypt, and Algeria tending towards totalitarianism.
16. The data were from Banks and Textor (1963, Appendix A). The component analysis is reported in Rummel (1970, pp. 140, 353, 379).
17. I am using common here to mean more than one characteristic was loaded on the component. Since this was a component, and not a common factor analysis, components defining unique variance can emerge, as did the third component shown in Rummel (1970). The third component, in any case, had an eigenvalue less than one (.97).
18. The grouping analysis was done on the three components reported in Rummel (1970, p. 353). The third component involves only the monarchy attribute, and no doubt contributes to the traditional-authoritarian grouping.
19. This is factor V given in Gregg and Banks (1965, Table VIII, p. 610).
20. Interestingly, Gregg and Banks title this component legitimation, a concept which in my theoretical analysis underlies authoritarian power and regimes.
21. Because the political component was fuzzy, I included more political variables in the subsequent 1950-1965 space-time analyses reported in Section (34.1).
22. This is the same data used for the results of sections 34.1 and , except time, and Geographic X, Y, and Z are omitted. In the analysis, the data on each attribute was standardized before the Q-analysis was done. This eliminated the possibility of a species factor (Rummel, 1970, pp. 294-95). The data are given in Rummel (1979).
23. Many libertarian type attributes, as freedom of group opposition do not distinguish between coercive and authoritative societies.
24. I am referring to what he calls families, which are determined by mean correlations within a grouping of states on their attributes.
25. This can be seen by comparing the polyarchic, elitist, centrist (totalitarian), personalist, and traditional political clusters found by Banks and Gregg (1965) with the clusters of states found by Russett (1967) or reported in Section 34.3.1 for the 1950-1965 space-time.