Statistics of Democide
Chapter 1: Summary and Conclusions [Why Democide?...]
Other Democide Related Documents On This Site
Now, surely such social diversity does play a role. But the necessary cause is first and foremost the extent of power. Especially, if Power (totalitarian power) dominates and especially if it is absolute, then as a racial, ethnic, national, or religious group is a threat to Power, stands in the way of Power's aims, or its liquidation will further Power, genocide may be unleashed. This means that whether social diversity will promote genocide is contextual and not general; it is a triggering cause or aggravating condition, not a necessary or sufficient cause.
But this is theory. What do we in fact find? To answer this I must answer a prior question: how do we best measure social diversity? Many different measures of this have been proposed in the literature, with the major useful ones I could find given in Table 18.1. I have listed their data in Table 18A.1. Given these data, what are their patterns? Again, component analysis will be helpful on which method see "Understanding Factor Analysis").
As clear from the component analysis shown in Table 18.2, all but one of the measures listed in Table 18.1 are quantifying similar aspects of those societies governed by the 214 state regimes. They are thus highly intercorrelated in one overall diversity pattern (Factor 1). There also is a minor pattern (Factor 2) involving the minorities that are at risk as a percent of the population. This is not the same as the other measures of diversity (as its creating a secondary pattern shows), for the percent of a population comprising all minorities and those that are at risk of the "violations of its survival rights"
Although in the tradition of component (factor) analysis one should ignore patterns that involve single measures, here this singular pattern could be an important one in our analysis and will be included. We thus have the two patterns listed in Table 18.3. These indicators will now be the basic measures of social diversity in subsequent analysis.
I now can answer the two important questions motivating this analysis. What is the relationship of diversity to democide? More specifically, is democide in some way dependent upon diversity? Second, is the close relationship between Power and democide described in Chapter 17 weakened or eliminated by taking account of social diversity.
Considering the first question, Table 18.4 presents a component analysis of the indicators of democide and diversity. For those who have emphasized the general causal role of diversity in genocide, these are unbelievable. There are the two patterns of democide that we have already identified, domestic democide and foreign democide, and a third comprising diversity that is completely uncorrelated with the first two. Since these patterns are uncorrelated (orthogonal), there is no relationship between this diversity pattern and the democide ones (if these patterns were allowed to be correlated through oblique rotation, then as can be seen from the primary intercorrelations, the diversity pattern would be correlated no more than an absolute .12 with the democide ones--on the meaning of such coefficients, see Understanding Correlation). Second, the two diversity measures themselves have a correlation of no more than an absolute .12 with the democide patterns. Indeed, the highest absolute product moment correlation between the two diversity indicators and the five democide ones is only .14, or no more than 2 percent of the covariation of diversity and democide in common.
I also carried out multiple regression analyses for each of the democide indicators on the two diversity indicators. None were significant (F-test) and the highest of the multiple correlation coefficients was .15. These regressions further confirm the results of the component analysis.
It may be, however, that there is a curvilinear relationship. Therefore all the democide indicators were separately plotted against the two ones, but no curvilinear relationship was evident. Figure 18.1 gives the most important of these plots, that between domestic democide and the major diversity indicator. The distribution of points looks almost random.
By itself, then, diversity has virtually no general impact on any pattern of democide, including that of genocide. But I hasten to add that this does not mean that for particular societies and their regimes, diversity will not be important. We know, especially from the genocides of the Nazis, Young Turks, and Czechs (regarding their ethnic Germans) that ethnic and religious differences can be a source of mass murder. However, what the results do say is that we cannot predict from the mere existence of such diversity to the existence, amount, or rate of such killing. If there is a general cause at work, social diversity does not appear to be it.
But what about Power and diversity? It may be that the effects of social diversity are being hidden by Power. Moreover, this broaches my original question--is the apparent large role of Power simply due to the social context of that Power, that is, social diversity? Again I will resort to component analysis to help answer these questions, with the results given in Table 18.5.
Nothing changes. We still get the important relationship between Power and domestic democide (with a lesser role for political power), the independent foreign democide, the independent other political indicators, and the independent diversity pattern. The underlying bivariate correlations reflect this lack of relationship. The two social diversity indicators have no more than an absolute correlation of .16 with all the democide and political indicators, and no greater than a .14 absolute partial correlation. Nor do they have any significant effect on a regression of the democide indicators on the political ones (not shown).
None the less, there may be significant differences in the patterns of relationship between democide and Power if we only look at those regimes with high social diversity versus those with low. Accordingly, I divided the 214 regimes on the diversity indicator into two groups: those below the diversity average of 9.3 (coded as 0) and those above (coded as 1). The results of doing a component analysis separately on each of these groups is shown in Table 18.6.
First, there is a significant structural change in the relationship between low and high diversity. For less diverse societies, the relationship is virtually the same as it is for all 214 state regimes shown in Table 17.5. But for diverse regimes, foreign democide is now part of the domestic pattern. During their lifetime regimes of diverse societies are more likely to commit both domestic and foreign democide, but its foreign democide is less likely to involve democidal foreign bombing. But second and more important here, Power still has a relationship to domestic democide, although less central. It also is now correlated with an independent pattern involving authoritarian regimes and political power. In addition, political power now has no relationship to domestic democide while monarchical regimes and especially those having to contend with the power of a traditional elite are most likely to kill others through democidal bombing.
Thus, below average diversity has no meaningful influence on the patterns of democide and democidal role of Power. But those regimes operating in societies with above average diversity tend more to commit, characteristically, domestic and foreign democide together; and Power's relationship has now been extended to both. Clearly this should be looked at in more detail, for this does give some role to diversity, but I will do this in a later chapters in the context of other possible influences.
In general, then, the degree to which there is national disunity and racial, ethnic, religious, and linguistic divisions do not by themselves significantly cause or condition a regime's democide, nor do they effect the relationship between Power and democide. However, when we look only at only those societies with much diversity, we do find a change in democide patterns and a lessening of the influence of Power. All these results leave open another question. Is the context of democide and Power more the question of culture rather than diversity, or the interaction of both? This is the concern of the next Chapter.
* From the pre-publisher edited manuscript of Chapter 18 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. Kuper (1981)
2. Gurr and Scarritt (1989, p. 381).
3. I was surprised by this result. I expected low to moderate correlations, but not this near total lack of relationship. Thus I thoroughly checked the data and analyses to make sure there were no errors.
4. The eigenvalue 1.0 criteria recommends five factors, but the best number of factors was theoretically four, given the previous results (two for democide, one for independent political characteristics, and one for diversity).