Suicidal rebellions and the moral hazard of humanitarian intervention

Abstract

This piece argues that the emerging norm of humanitarian military intervention, which is intended to prevent genocide and ethnic cleansing, perversely causes such violence through the dynamic of moral hazard. The norm, intended as a type of insurance policy against genocidal violence, unintentionally encourages disgruntled sub-state groups to rebel because they expect intervention to protect them from retaliation by the state. Actual intervention, however, is often too late or too feeble to prevent such retaliation. Thus, the norm causes some genocidal violence that otherwise would not occur. The piece starts by documenting the fact that historically most genocidal violence—unlike that of the Holocaust—has been retaliation against groups who threaten state authority. It then develops a framework based on deterrence theory to explain why groups vulnerable to genocidal retaliation might provoke that outcome. Next, it illustrates this dynamic in two cases from the 1990s: Bosnia and Kosovo. Following this, it explores analogous problems in economics and discusses potential remedies. The piece concludes by presenting a critique of the putative moral responsibility to intervene.

The author would like to thank all those who offered comments, particularly Timothy Crawford for his close reading of the penultimate version.

Notes

1. In both the Webster's and American Heritage dictionaries, only one of four definitions includes intentionality.

2. Subsequently Harff (1994) broadened these definitions to include policies sponsored by non-state actors in the case of civil war. For a critique of their methods, see Fein (1990). In addition to disputing their broad definition, she notes that, by counting each case by perpetrator rather than victim, they lump together several cases in the USSR and Iran.

3. Some of the other cases in Harff and Gurr's database would also satisfy this definition if re-coded properly. For example, the authors erroneously categorize the killing of Tutsi in Rwanda in 1963–64 as retributive on the grounds that it was perpetrated by new Hutu leaders in retaliation for years of Tutsi oppression. In fact, such retributive violence ended in Rwanda soon after the Hutu seized power in 1959. The subsequent killing of 1963–64 was rather a response by the Hutu nationalist state to fresh challenges to its authority from invading Tutsi refugee rebels and their domestic Tutsi allies—a fairly typical case of repressive politicide. Accordingly, 68% may represent a conservative (i.e. low-end) estimate of the proportion of cases from 1943–87 in which the ultimate victim group provoked its own demise, based on Harff and Gurr's database.

4. Fein operationalizes this distinction not on the basis of any objective definition, but rather by selecting cases identified as genocide by at least two of three prominent expert studies. The experts she relies on are Ezell, Kuper and Harff and Gurr. In a subsequent study, Fein (1993) identifies only 16 cases during the same period without acknowledging or explaining the discrepancy with her earlier study. In the latter study she also distinguishes genocide from “genocidal massacres” or “pogroms”, which are briefer or more episodic, and from “mass political killings”, a term she does not define clearly but which appears to refer to killings of civilians during civil wars. It is not clear if she operationalizes these distinctions by rigorous standards.

5. She is not absolutely precise about which cases are retributive. However, she does identify seven cases precisely as ideological, developmental or despotic, which leaves 12 rather than 11 cases as retributive. In addition, she says that one of the cases identified as despotic, Uganda, included periods of retributive genocide, which potentially raises the number of retributive cases to 13.

6. I include the targeting of political groups, even though this category of victim was excluded from the UN convention's definition of genocide under pressure from member states including the USSR. Other authors have chosen to coin new terms to indicate such a broader definition—for example, “mass killing”, in Valentino (2004) and “democide” in Rummel (1992).

7. The quantitative thresholds of 50 000 total and 5000 annually are arbitrary. As with any such arbitrary definition, cases that fall marginally short of the standard could probably be included in the universe without significantly affecting its characteristics. Furthermore, an argument could be made for utilizing an alternative threshold based on the percentage, rather than absolute toll, of people killed within the victim group. However, this alternative would have two drawbacks. First, it could include some cases with relatively low death tolls (in cases where the target group was small), while excluding others with significantly higher death tolls (among big target groups). Second, determining the size of the target population in many cases would be subjective, because it could depend on whether an entire ethnic group were counted or only that portion within a state or region. While my definition is arbitrary, it does have the merit of being relatively objective, at least to the extent that existing death-count estimates are. It is possible that my high threshold may exclude some less violent examples of the very phenomenon I seek to examine, but that is the unavoidable price of seeking to exclude different phenomena such as terrorism or counter-insurgency campaigns that generally have lower death tolls. I accept this trade-off consciously, preferring to ensure that all cases in my universe represent the same phenomenon, rather than that the universe contain all examples of the phenomenon.

8. Darfur in the Sudan is too recent to be listed in published databases. The other three cases are the only ones that satisfy my definition in the database of genocides and politicides during 2001, in Harff (2003).

9. Despite the peaceful nature of the challenge, hardline Tutsi feared that peaceful Hutu consolidation of political power would lead to violence against them or threats to their way of life, and so they assassinated the new Hutu president and reclaimed power in 1993, triggering mutual ethnic violence and a Hutu rebel insurgency. The new Tutsi government then responded to the Hutu insurgency with a seven-year brutal counter-insurgency that included mass killing of Hutu civilians. Although this second, protracted wave of killing fits the typical pattern of a suicidal rebellion, the case as a whole cannot be coded as the victim group provoking its own demise. (This coding could change if evidence were found that the assassination itself was provoked by impending Hutu plans for violence.)

10. Despite its title, the book devotes considerable attention to the strategic nature of genocide.

11. Valentino (2000) makes a similar point about the potential misperceptions of state leaders: “A strategic approach to mass killing does not imply that leaders accurately assess the threats they face. Nor does it suggest that mass killing will always help leaders achieve their goals or solve their problems…Nevertheless, leaders ultimately act on the basis of their perceptions and beliefs.”

12. For example, in 1995 the Minorities at Risk database identified 268 “ethnic or communal groups” worldwide that were “disadvantaged by comparison with other groups in their society”, of which only 22 (8%) were engaged in violent rebellion at or above the level of intermediate-scale guerrilla activity. See Gurr (2000). Valentino (2000) notes that “Recent quantitative research on ethnic conflict and genocide has found little correlation between the severity of ethnic, social, economic, and cultural differences and the likelihood of large-scale violence between groups”. See also Fearon and Laitin (2003); Hoeffler and Collier (2004).

13. A good primer is Davies (1971b). Perhaps most prolific is Ted Robert Gurr (1970; 1993), who has published numerous versions of his theory.

14. Both quoted in Davies (1971a), which notes that the original theory of Marx and Engels posited revolution as the response of industrial workers to their progressive absolute deprivation under capitalism. However, Marx later wrote that revolution was still inevitable in the face of rising living standards of the proletariat, because “although the enjoyments of the workers have risen, the social satisfaction that they give has fallen in comparison with the increased enjoyments of the capitalist”. Thus, according to Marx, relative deprivation of material goods leads to absolute deprivation of social satisfaction, and thence to revolution.

15. His work draws on ‘frustration-aggression’ theory.

16. A recent version of this theory is found in Goodwin (2001). See also the more general argument in Hirschman (1970).

17. Similarly, Thornton (1964) writes that “terroristic acts often are committed with the express purpose of provoking reprisals”. Pye (1964) writes that “The initial decisions of a government confronted with the threat of internal war are usually the most fateful and long-lasting”.

18. As early as 1968 a study for the US military found that rebellions tend to occur in rural societies with rough terrain favourable to guerrilla warfare. D.M. Condit, cited in Orlansky (1970). A similar finding was recently made by Fearon and Laitin (1999). Other researchers have discovered that areas of ethnic geographic concentration—that is, when a group is a majority in a local region but a minority in the state as a whole—also favour rebellion, presumably by facilitating mobilization but also possibly by exacerbating the security dilemma with other groups in the state. On this point, Gurr (2000) cites the work of Erik Melander, Monica Duffy Toft, Barry Posen and Stephen Van Evera.

19. See Gurr (2000) on “feedback effects” and the “dynamics of protracted conflict”. Unfortunately, a linear causal diagram (p. 70) masks the endogeneity of his theory. A diagram in his previous book (Gurr, 1993, p. 125), containing feedback loops, was more confusing but more explicit about this endogeneity.

20. The main variables, and their underlying variables, are also summarized in the diagram in Gurr (2000, p. 70). Salience of identity is a function of: 1) the extent of cultural differentials; 2) relative deprivation; and 3) intensity of past and present conflicts with the state and other groups. Incentives for collective action (based on various types of relative deprivation) include: 4) overcoming collective disadvantage; 5) regaining political autonomy; and 6) resisting repression. Group capacity for action (i.e. mobilization) is a function of all the preceding variables and 7) territorial concentration (including terrain features); 8) pre-existing group cohesion; 9) intra-group coalition building; and 10) legitimacy of group leaders. Political opportunities for action are opened by: 11) state creation or destruction; 12) regime transition including democratization; and 13) leadership transition. International factors that can affect all of the above variables include: 14) global norms of group rights; 15) diasporas; 16) diffusion and contagion of ideas and resources between similarly situated but ethnically distinct groups in different states; and 17) other external political and material support. Domestic political factors that determine whether ethnopolitical action will be peaceful or militant include: 18) institutions of democracy or authoritarianism; 19) extent of state resources to accommodate group demands; and 20) state traditions of accommodating or repressing group demands.

21. This theory, from the field of psychology, posits that the frustration from unfulfilled aspirations or expectations is the root cause of aggression. The theory originates with Freud and was formalized and tested originally by Dollar et al. (1939). Subsequent studies include Feierabend and Feierabend (1966) and Tanter and Midlarsky (1967). In addition, Gurr (1971) explicitly roots itself in this theory.

22. Some versions of this theory assume rational action but by multiple actors in a non-unitary state to maximize their own utility rather than that of the state. Allison and Halperin (1972).

23. These case studies are drawn from, and fully documented in, Kuperman (2002).

24. As noted by Lebow and Stein (1989), “the reconstructions of participants after the fact…[are] subject to well-known biases”. Accordingly, Lebow and Stein “look for convergent evidence from several participants from each side, and for historical documentation as well”.

25. The journalist quoted is Veton Surroi.

26. A similar phenomenon may explain a few earlier suicidal rebellions, for example that in the 19th century by Christian groups in the Balkans against Ottoman authorities, aimed at provoking retaliation that would attract intervention from Christian Europe.

27. A recent report on this phenomenon is discussed in Blustein (2004).

28. On requirements, see Quinlivan (1995). On resources, see O'Hanlon (2003).

29. For suggestions on how to expand global capacity for timely and effective humanitarian military intervention, see Kuperman (2001; 2004a); White House (2004); O'Hanlon (2003).

30. Advocates of this norm do not appear to draw a meaningful distinction between provoked and unprovoked state violence. Interestingly, the ‘responsibility to protect’ was based on the two cases in my study—Bosnia and Kosovo—plus Rwanda and Somalia. See also Chopra and Weiss (1992); Deng (1995); and Deng et al. (1996).

31. In his contribution to this volume Crawford (2005) apportions responsibility for genocidal retaliation between the rebels who provoke it and the international intervenors who create moral hazard that encourages rebellion. He argues that more responsibility adheres to the rebels if moral hazard has been created by a longstanding international norm of intervention rather than an ad hoc proximal threat of intervention, because the rebels can anticipate the domestic and international reactions to their rebellion.

The original version of this study was presented at the annual meeting of the International Studies Association, Chicago, IL, 21–24 February 2001. Revised versions have been presented at more than a dozen international conferences and university seminars.