Statistics of Democide
Chapter 1: Summary and Conclusions [Why Democide?...]
Other Democide Related Documents On This Site
The question is then whether my results so far are conditioned or indeed caused by underlying socio-economic and geographic patterns. This is the same question I asked about cultural patterns and the answer was no at the global level. But conceivably the empirical relationship between Power and democide may be due to their correlation with whether a society is rich or poor, large or small, or its people well educated or not. (On the nature of correlation, see Understanding Correlation.)
To measure the wealth, education, size and other dimensions of society is straight forward. Beginning in the 1950s extensive component and factor analyses of social, political, and cultural variables have been carried out.
The second pattern is that of politics, particularly democratic versus authoritarian and totalitarian systems of government. And the third is national power (or size), which is a state's area, population, and physical and technological resources, the latter as measured by its consumption of energy. Wealth, politics, and national power span the most important aspects of society, accounting usually for over 50 percent of the variation in hundreds of national attributes. Other patterns delineate national cultures, density, domestic conflict, and foreign conflict.
I will use these results to encompass the major socioeconomic and geographic similarity and differences among national societies and thus to quantify the context within which regimes do or do not commit democide. Based on the literature,
For what years should data be collected on these socio-economic indicators? The natural response is for the years of a regime, such as for that of the Russian Czar 1900-1917. This is certainly correct, as for data on population, refugees, and famine. But what about measures of wealth and national power? Consider GNP per capita, for example. Contemporary figures for most state regimes for the early part of this century will rank them with many of the poorest states of the 1970s and 80s. Even the most comparatively developed states of 1905 and 1910 would rank with the poorer states of sixty or seventy years later. Since this is a comparative analyses, with data intercorrelated across all regimes, this surely would create misleading results. Even a few years can make a big difference. For example, were we to select contemporaneous data, then the mid-year of these data for the British regime would be around 1944, but since this was a war-year, postwar 1946 would be better. Great Britain's GNP per capita for that year was $482, which compared to GNP per capita twenty-one-years later would class Great Britain below many less developed states, such as Bulgaria ($747), Libya ($814), Panama ($500), Trinidad ($662), and Uruguay ($564).
Because of this problem with many of the measures, regardless of the years for which a regime existed, I have defined their data as for the late 1970s and early 1980s. Thus for France, for example, data on energy consumption per capita, one of the best indicators of national wealth, is collected for 1980 and applied to the pre-1941 French regime; that for the Soviet Union in 1980 is applied to the Czar's Russia; and that for West Germany is applied to the Kaiser's Germany 1900 to 1918. This is, of course, not entirely satisfactory, since some states have developed at a greater pace than others. The development of Japan is a case in point. But the alternative of collecting such data for years contemporaneous to the regime is even less satisfactory. In any case, I have labeled all measures in Table 20.1 with the prefix Reg- if they are contemporary, that is collected for the regime's mid-year. Those not so labeled are for the 1970s and 80s.
Moreover, several of the measures were highly skewed with a very few high scores. Because this would cause any correlations for these measures to hang on these high scores they were all logged. The measures for which this was done are shown in the table, as for the democide types, by the suffix L.
Now turning to the patterns among these measures, these will not necessarily be the same as generally found for cross-national data. This is because the data are not for all states for some particular year, but for all 141 state regimes that have committed democide in addition to the seventy-three that have not. Moreover, because for some states successive regimes have committed democide, such as the French Third Republic (1900-1940), Vichy (1940-1944), provisional government (1944-1946), and Fourth and Fifth Republics (1946-1987), a state may be represented in the data several times.
Table 20.2 presents the patterns among these measures and Table 20.3 gives the pattern interpretation and indicators. The first three patterns are indeed what has generally been found in cross-national data and their indicators are in line with those selected by other studies.
The next step is to do a component analysis of these indicators along with those democide, politics, and culture. The results of this are shown in Table 20.4.
An independent pattern of foreign democide is still there also, but now within the context of the larger society there is a correlation between whether a regime is Central or South American and foreign democide--Latin regimes tend to commit less of it. Population is also related to this pattern. Since this is an indicator of national power and its correlation with the pattern is positive, then the results show that the greater the national power controlled by a regime, the more likely it is to commit foreign democide.
Different forms of power are now seen to play a global role in both democide patterns. Totalitarian and political power are correlated with domestic democide as before (Chapter 17), national power with foreign democide. Let us see if this holds up at different levels of development--do relatively rich nations behave differently than relatively poor ones on these patterns?
I divided the 214 states into 89 richer and 125 poorer states at the average of energy consumption per capita, the indicator of wealth or economic development. Table 20.5 compares the result of doing a separate component analysis on each.
The magnitude of domestic democide is now a separate pattern from the domestic democide rate, with the latter having only minor relationships with the non-democide measures. The domestic democide indicator forms a pattern alone with totalitarian power and non-Central and South American states. This means that the poorer Latin states tend to have less democide and less totalitarian power, but even then there remains a relationship between democide and Power.
The foreign democide pattern (Factor 2) is almost lost and what there is includes a number of other measures. These imply that poorer nations that commit foreign democide tend to a minor extent to be more homogenous, have less English cultural influence, be less dominated by clans, and have higher national power.
In sum, we find that the fundamental relationship between Power and democide is little altered by enlarging the context of a regime's behavior to include socio-economic (which includes educational) and geographic conditions. Moreover, and importantly, we have extended the relationship of power to democide to include that of national power to the foreign democide pattern. Nor do we find that looking at the richer versus the poorer states changes these equations for richer states. But for poorer ones this relationship is more complex, although the role of Power in the magnitude of domestic democide is still clear.
What has been left out of all these analyses is the possible role of domestic and foreign violence, specifically war and rebellion. I will now look at this in detail.
* From the pre-publisher edited manuscript of Chapter 20 in R.J. Rummel, Statistics of Democide, 1997. For full reference to Statistics of Democide, the list of its contents, figures, and tables, and the text of its preface, click book.
1. Rummel (1994).
2. See Rummel (1970) for an annotated bibliography of the earlier component and factor analysis of states. In Rummel (1972) I present a 236 variable component analysis of states for 1955, and Rummel (1979) extends this to five different time periods. All such analyses I have seen since confirm the findings reported in these books.
3. See endnote 2.
4. Figures from Banks (1971, segment 8).
5. See endnote 2.
6. At an eigenvalue one criteria, I would have rotated nine factors. However, the analysis here is not exploratory but theoretical and comparative, particularly of the relationship between Power and democide. I rotated only four factors to conform to the structure found for the democide and political patterns in Table 17.5
7. For richer states I was able to define the democide patterns with the standard four-factor rotation. However, for the poorer states, the patterns are more diffuse and spread out, requiring a six factor rotation in order to best define them.