Third-party intervention and escalation in Kosovo: Does moral hazard explain it?

ABSTRACT

This piece provides a critique of the moral-hazard theory of third-party interventions. While the author agrees with the proponents of the theory that threats of intervention may have radicalizing effects on a minority's behaviour by reducing its risks in carrying out violence, he argues that this is an incomplete causal argument, if what we are trying to explain is escalated violence and not just minority radicalization. This criticism is based on two claims in particular. First, if the minority and the state observe the same third-party threat, they should both adjust their bargaining calculations. This means that such threats should affect the terms of settlement, and not the likelihood of violence. Second, he argues that cases where the minority radicalization results from explicit or implicit encouragement by third parties should not count as evidence supporting the moral-hazard theory. An internally consistent moral-hazard account requires an explanation of why third parties were unable to make their threats of intervention conditional on the minority's responsible behaviour. The piece also provides an alternative, informal analysis of escalated violence in response to third-party interventions focusing on the effects of incomplete information regarding the target-state's and the third party's preferences over a set of outcomes. The author uses evidence from Kosovo to compare the two arguments.

Acknowledgements

An earlier version of this paper was presented at the annual conference of the American Political Science Association, Chicago, IL, September 2004. For their helpful comments I would like to thank Robert Jervis, Ron Hassner, Jack Snyder, R. Harrison Wagner and the participants of the Columbia University International Politics Seminar. I owe a special debt of gratitude to Alan Kuperman and Tim Crawford, who have been particularly generous with their time and wisdom. I would also like to acknowledge the support of the Olin Institute for Strategic Studies and the Belfer Center for Science and International Affairs at Harvard University, which made the research for this paper possible.

Notes

1. The argument in Jatras (2000) and Daalder and O'Hanlon (2000) is not that moral hazard was the principal cause of the conflict in Kosovo. Their studies do mention moral hazard as one of the problems, however.

2. For studies that dispute the existence of the ‘CNN effect’ as an independent cause, and argue instead that the ‘CNN effect’ is manufactured either by the government or the elites, see Herman and Chomsky (1988) and Carpenter (1995).

3. Cetinyan's argument assumes complete information, but the argument stands even if we assume incomplete information regarding the third party's preferences. In particular, if both the minority and the target-state are incompletely informed about the preferences of the third party, the assumption of rationality demands that their estimate of intervention be the same. For the general treatment of this claim for all interactions characterized by incomplete information, see Harsanyi (1967, pp. 163–166).

4. Ernest Gellner (1983) makes the observation regarding the relative rarity of demands for independence given the number of groups which could potentially press such demands. See also Fearon and Laitin (1996).

5. This evidence is presented in Burg and Shoup (1999, p. 113), and Kuperman (2003, p. 62).

6. On the problems associated with selecting on the dependent variable in general see Geddes (1990), and King et al., (1994, pp. 129–137).

7. This is actually what happened in Kosovo as well, when the KLA arrived on the scene in 1996 and radically changed the situation without any visible shift in the behaviour of third parties.

8. Daalder and O'Hanlon (2000) make the most sophisticated case along these lines.

9. The most important problem in this account is the claim that NATO responded to an ongoing or planned mass expulsion. While the Serbian policy in Kosovo was indeed very repressive, and while there was quite a bit of violence in Kosovo before NATO's bombing campaign, there is little evidence that what the Serbs were doing was a systematic campaign of mass expulsion. The one potentially serious piece of evidence that, even if Serbs were not engaged in a campaign of systematic ethnic cleansing they planned to become so, was the so-called ‘Operation Horseshoe’, which was supposedly the blueprint for the operation of mass expulsion of Kosovo Albanians that would have been put in motion even if NATO did not intervene. But the plan turned out to be a German forgery, as Human Rights Watch discovered. See Human Rights Watch (2001, p. 59). Moreover, operational plans do not necessarily reveal political intentions. Finally, I will present evidence later in the paper that NATO leaders not only did not think there was ongoing ethnic cleansing in Kosovo, they thought Milosevic was unlikely to embark on such a course even if painted into a corner.

10. Serbs were the most dispersed group in Yugoslavia, with large numbers of them living in Bosnia and Croatia, as well as other parts of the federation. For that reason Serbs were particularly sensitive to any weakening of the federation and had strongly discouraged all expressions of nationalism in Yugoslavia, including their own.

11. There was also a general rise of separatism in other parts of Yugoslavia, particularly in Slovenia and Croatia in the 1980s, which also contributed to the weakening of the Serbian commitment to ‘Yugoslavism’, and the rise of Serbian nationalism. For a detailed and excellent analysis of the causes of the rise of nationalism in Yugoslavia in the 1980s, see Woodward (1995, chs 2–4).

12. The KLA had held its first meeting in 1993, but it had no visible political presence in Kosovo until 1996.

13. Kuperman and Crawford have raised questions regarding the relevance of this issue in personal correspondence with the author. The relevant issue, according to them, is whether the the intervenor—the USA in this case—creates perverse incentives for risky behaviour by the minority—the Kosovo Albanians—regardless of whether the intervener could threaten abandonment or not. I disagree for two reasons. First, that claim is only correct if the object of analysis is the minority's behaviour rather than the outcome of an interaction between the intervenor and the minority. If we are trying to explain why conflicts between states and minorities escalate, however, rather than why minorities behave provocatively, the intervenor's choices are more than relevant. Second, the most basic puzzle underlying the moral hazard theory, and the reason we study it in this as well as other contexts, is why principals—third-party intervenors in this case—do things that produce outcomes that are less than optimal for them. The answer usually is some condition that either prevents the principal from observing the agent's—in this case the minority's—undesirable behaviour, or makes punishing the agent even costlier than absorbing the costs of the agent's undesirable behaviour, as I mentioned above. We cannot describe a problem as one of moral hazard if the principal knowingly and intentionally encourages reckless behaviour on the part of the agent.

14. The ‘School pact’ negotiations are interesting for another reason. It seems that Milosevic displayed a readiness to negotiate after his bargaining position against the Kosovo Albanians had improved considerably. More specifically, the ‘School pact’ negotiations took place after Dayton, where the West had refused to make a settlement in Kosovo a part of the comprehensive deal and had told the Kosovo Albanians that any settlement had to respect rump Yugoslava's territorial integrity (Vickers, 1998, pp. 286–294). This is definitely not sufficient to conclude that more NATO pressure on the Albanians would also have had a simultaneous moderating effect on Milosevic, but it is quite suggestive.

15. This rationalist account of violence between states is not universally shared either. But there is a tendency to see violence against minorities as fundamentally irrational even among those who do not see interstate wars as irrational.

16. I develop this argument formally and more extensively in Arman Grigorian, ‘Third-party intervention and escalation of violence in state–minority disputes’, PhD dissertation in progress, Columbia University. What follows is the more condensed and informal version.

17. The resolution passed with 219 votes in favour and 191 against, but the debates in the House left little doubt about the extent of the opposition to the intervention. The resolution also attached a large set of conditions to any deployment. It required, for instance, that the president report to Congress about what national interests were involved in Kosovo, what the costs were going to be, what the exit strategy was, etc. See Jessica Lee, ‘GOP backs Clinton on Kosovo vote', USA Today, 12 March 1999, p. A1.

18. The interview is available on line at: http://www.pbs.org/newshour/bb/europe/jan-june99/ albright_3-24.html.

19. In the end he gave up without a ground invasion. There was a gathering threat of ground invasion during the latter part of the air campaign, however, which in all likelihood strongly affected his decision to give up. Studies that make this claim include Daalder and O'Hanlon (2000), Byman and Waxman (2000) and Hosmer (2001). The best defence of the position that air power alone should be credited with victory is Stigler (2002/03). In any case, the threat of escalation to a ground war was not the only reason for Milosevic's decision to quit. The withdrawal of Russia's diplomatic support almost certainly played a major role as well. It should also be mentioned that the terms of his surrender were lighter than those of the Rambouillet ultimatum. On this, see Mandelbaum (1999) and Hosmer (2001, ch. 4).

20. Sandy Berger, who was Clinton's National Security Adviser, told Daalder and O'Hanlon that maintaining the fragile alliance and avoiding a congressional debate on the issue was why it was necessary to rule out a ground war and why Clinton had to make the aforementioned statement. See Daalder and O'Hanlon (2000, p. 97).

21. Such criticism has come both from the left and the right. See Chomsky (1999) and Carpenter (2000).

22. Milosevic stated that he could empty Kosovo in a week if there were an intervention. See Greenhill (2003, p. 214).

23. One such flaw was extrapolating from Milosevic's decision to go to Dayton and accept compromise there. The extrapolation was problematic because at Dayton Milosevic got carrots along with the sticks. Moreover, the air strikes that led to Dayton were accompanied by the Croatian conquest of Krajina. See Daalder and O'Hanlon (2000, p. 93). Some experts indeed predicted escalation before the fact. See Kuperman (1998a; 1998b).

24. One possible answer is incompetence, of course. Incompetence, however, is not an explanation, but a substitute for one.