BENJAMIN O.FORDHAM 747 Rival intervention would increase this probability to .32.The involvement of one American ally would increase the probability to .23.The involvement of two allies,something that happened in roughly 10 percent of the cases,would increase it to.35.Exports had no statistically significant effect on either civil war or crisis intervention. Applying the model to some well-known cases helps to show the plausibility of the relationships that it implies.Consider first the 1979 conflict between Uganda and Tanzania,included in the ICB data as crisis 296.Following a border incur- sion by Ugandan forces.Tanzanian forces invaded Uganda and forced its presi- dent,Idi Amin,from power.The United States remained largely uninvolved. The model places the probability of American intervention at.29.This crisis can be contrasted to the incursion by Cuban-backed rebels into Zaire's Shaba prov- ince in 1977,an event that prompted an upsurge in American assistance to the Zairian military (ICB 277).In this case,the predicted probability of American intervention was.42.The model implies that the different American response to these two crises was due mainly to Cuban and Soviet intervention in the Shaba case.The probability of American intervention if the Cubans had intervened in the Uganda-Tanzania dispute would have been .52.Without such intervention in the Shaba case,the predicted probability of American intervention would have fallen to .21.This inference makes sense in light of the high priority placed on preventing communist gains anywhere in the world during the Cold War.Ameri- can policy makers,perhaps not always wisely,regarded intervention by their rivals as evidence that something important was at stake. A similar comparison between American decisions about the civil wars in Morocco and El Salvador helps illustrate the effect of alliances.These long-running civil wars began within a few years of one another,and were quite similar in terms of their values on the other independent variables.The United States intervened in El Salvador,but not Morocco.The predicted probability of intervention in El Salvador is.32 compared to.13 in the Moroccan case.In the context of the model, the difference is due to the fact that the United States maintained a military alli- ance with El Salvador,the Rio Treaty,but not with Morocco.The model suggests that the probability of intervention in El Salvador would have fallen to .08 in the absence of the alliance,and risen to.43 in Morocco if an alliance had existed. The importance of an alliance like the Rio Treaty in these two civil war inter- ventions is not as easy to justify as the claim about the role of Cold War rivalries in the two African crises in the preceding paragraph.In spite of the fact that the Soviets and Cubans did not intervene in El Salvador under the definition used here,American intervention in the civil war was justified primarily in terms of anti-communist ideology rather than the need to uphold American alliance com- mitments.From a theoretical standpoint,alliance commitments are not easily enforceable under anarchy,and adhering to them can be costly.The effect of an alliance commitment in the model makes more sense if one recognizes that alli- ances represents a durable assessment of the American stake in the security of the allied state.This assessment is more fundamental than the formal agree- ment.American foreign policy makers have long had a special concern about the Western hemisphere.A satisfying explanation of American intervention in cases like El Salvador thus requires an account of how the underlying interests that produced the alliance developed,an issue to which I will return later. If an analysis of the effects of economic and security concerns on intervention ended with the direct effects estimated in Table 1,one might conclude that secu- rity concerns trump economic interests.Alliances and rival behavior have had much larger effects on the probability of American intervention in civil wars and international crises than have exports.The need to account for the interests indicated by alliance commitments suggests that such a conclusion would be premature.A more complete understanding of American intervention requiresRival intervention would increase this probability to .32. The involvement of one American ally would increase the probability to .23. The involvement of two allies, something that happened in roughly 10 percent of the cases, would increase it to .35. Exports had no statistically significant effect on either civil war or crisis intervention. Applying the model to some well-known cases helps to show the plausibility of the relationships that it implies. Consider first the 1979 conflict between Uganda and Tanzania, included in the ICB data as crisis 296. Following a border incur￾sion by Ugandan forces, Tanzanian forces invaded Uganda and forced its presi￾dent, Idi Amin, from power. The United States remained largely uninvolved. The model places the probability of American intervention at .29. This crisis can be contrasted to the incursion by Cuban-backed rebels into Zaire’s Shaba prov￾ince in 1977, an event that prompted an upsurge in American assistance to the Zairian military (ICB 277). In this case, the predicted probability of American intervention was .42. The model implies that the different American response to these two crises was due mainly to Cuban and Soviet intervention in the Shaba case. The probability of American intervention if the Cubans had intervened in the Uganda-Tanzania dispute would have been .52. Without such intervention in the Shaba case, the predicted probability of American intervention would have fallen to .21. This inference makes sense in light of the high priority placed on preventing communist gains anywhere in the world during the Cold War. Ameri￾can policy makers, perhaps not always wisely, regarded intervention by their rivals as evidence that something important was at stake. A similar comparison between American decisions about the civil wars in Morocco and El Salvador helps illustrate the effect of alliances. These long-running civil wars began within a few years of one another, and were quite similar in terms of their values on the other independent variables. The United States intervened in El Salvador, but not Morocco. The predicted probability of intervention in El Salvador is .32 compared to .13 in the Moroccan case. In the context of the model, the difference is due to the fact that the United States maintained a military alli￾ance with El Salvador, the Rio Treaty, but not with Morocco. The model suggests that the probability of intervention in El Salvador would have fallen to .08 in the absence of the alliance, and risen to .43 in Morocco if an alliance had existed. The importance of an alliance like the Rio Treaty in these two civil war inter￾ventions is not as easy to justify as the claim about the role of Cold War rivalries in the two African crises in the preceding paragraph. In spite of the fact that the Soviets and Cubans did not intervene in El Salvador under the definition used here, American intervention in the civil war was justified primarily in terms of anti-communist ideology rather than the need to uphold American alliance com￾mitments. From a theoretical standpoint, alliance commitments are not easily enforceable under anarchy, and adhering to them can be costly. The effect of an alliance commitment in the model makes more sense if one recognizes that alli￾ances represents a durable assessment of the American stake in the security of the allied state. This assessment is more fundamental than the formal agree￾ment. American foreign policy makers have long had a special concern about the Western hemisphere. A satisfying explanation of American intervention in cases like El Salvador thus requires an account of how the underlying interests that produced the alliance developed, an issue to which I will return later. If an analysis of the effects of economic and security concerns on intervention ended with the direct effects estimated in Table 1, one might conclude that secu￾rity concerns trump economic interests. Alliances and rival behavior have had much larger effects on the probability of American intervention in civil wars and international crises than have exports. The need to account for the interests indicated by alliance commitments suggests that such a conclusion would be premature. A more complete understanding of American intervention requires Benjamin O. Fordham 747