Population Shifts and Civil War: A Test of Power Transition Theory

Abstract

Do shifts in the distribution of ethnic group populations within a multinational state make civil war more likely? This article tests the proposition that they do using the competing logic of two core theories of interstate politics: power transition (PTT) and balance of power theory (BPT). The universe of potential population transition types are reduced to nine, and the logic of each of the competing explanations of war likelihood are reduced to four testable hypotheses. Overall, PTT fares better than BPT; although the article concludes that, as is the case at the interstate level, the key determinate of war likelihood rests more with how power is perceived than with raw changes in its distribution across the spectrum of meaningful political actors. Finally, the article offers a useful framework for further specifying the conditions under which population shifts alter the likelihood of an escalation to civil war.

KEYWORDS:

• population
• civil war
• power transition
• balance of power
• demography
• shift

Notes

1. Assumptions in the sense that they are not evaluated or tested. The following are paraphrases of “hard core” assumptions outlined in DiCicco and Levy, 2001, pp. 119–120.

2. This study does not seek to address the vast literature on inequality and civil wars, greed, grievance, development, agent base, or the question of historical identity differences.

3. Numbers in bold indicate where civil war is likely, and italic numbers indicate where civil war is unlikely per power transition theory.

4. PTT is an evolving theory. Tammen et al. (2000) hypothesize that slower rate of change (i.e. a longer period of parity or transition) increases the likelihood of war.

5. Horowitz (1985), among others, posits that the closer the ethnic groups are in size, the more likely the violence. Ellingsen's (2000) hypothesis is that the larger the size of the largest minority, the more likely that domestic conflict will occur (p. 234).

6. PTT also concerns itself with, but does not answer, the question of who is most likely to escalate a conflict of interests to violence, given a rapid and decisive shift in group power. This argument is not tested here.

7. The data set is available at http://www.svt.ntnu.no/iss/Tanja.Ellingsen/Default.htm

8. See Appendix A for cases and Appendix B for summary statistics. The full dataset is available electronically from the author.

9. Because demographic shifts take a long time, I thought it best to have as large a window for violence as possible. Here I use 20 years: most countries hold decennial censuses, so this allows for two counts, and 20 years allows for two generations of births and their rates to be assessed.

10. See Appendix D.

11. There were a total of 33 wars, but 45 distinct transition dynamics. This is because some wars lasted so long that population shifts occurred as the war was fought.

12. This is according to cross-tabulation calculations between expected and actual frequencies. See Appendix C.

13. The population percentages are from the year the war started.

14. Type 5 was included in the testing of models 3 and 4, but was dropped by STATA for too few observations.

15. Type 7 was included in the testing of models 5 and 6, but was dropped by STATA for too few observations.

16. This lends some support to Tammen et al.'s (2000) revision of PTT, which posits that slow transitions lead to a greater likelihood of war.

17. Stephen Walt's Origins of Alliances (1987) made a similar point years earlier when he argued that it was not power but threat against which states sought to balance in order to enhance their security. Also see Mercer, 1996.