746 Power or Plenty? The Direct Effects of Political and Economic Interests on Intervention Table 1 presents the results of four models of American intervention in civil wars and international crises.There are good reasons to prefer a parsimonious specifi- cation (e.g.,Achen 2002;Clarke 2005;Ray 2003),but it is also possible that esti- mates from such a minimalist specification reflect related influences on intervention decisions that were omitted from the model.Models 1 and 3 include only the variables of primary theoretical interest here:alliance commit- ments,rival intervention,and exports.Models 2 and 4 add several additional independent variables reflecting considerations that might create omitted vari- able bias in the more parsimonious specification.For example,the United States might be less likely to intervene in conflicts involving militarily powerful states, but more likely to intervene in those involving its neighbors,or democratic gov- ernments.The United States might have acted more aggressively during the Cold War,or on behalf of its NATO allies.Military action might also be more likely in times of economic distress.All of these considerations might also be related to the primary independent variables,especially trade,creating omitted variable bias.In practice,including these additional independent variables has a small effect on inferences about the variables of primary theoretical interest.The coef ficient estimates for alliance commitments and rival intervention are somewhat larger;trade remains statistically insignificant. The most important empirical finding from this stage of the analysis is that security concerns about alliance commitments and rival intervention had large direct effects on American intervention,while exports were not statistically signif icant.The two security concerns had very similar effects on intervention in both civil wars and international crises.Holding the continuous independent variables at their mean value,the predicted probability of civil war intervention in the absence of an alliance or intervention by a rival state was.11.The addition of an alliance would increase this probability to .45.Rival intervention would increase it to.32.For international crises,the baseline probability of intervention was.14. TABLE 1.Probit Models of Civil War and International Crisis Intervention International Crisis Civil War Intervention Intervention Rival intervention 0.81(0.30)* 0.94(0.33)* 0.44(0.16)* 0.61(0.17)* Alliance partners involved 1.09(0.31)* 1.12(0.44)* 0.32(0.10)* 0.35(0.14)* Log of exports in previous year -0.04(0.06) 0.11(0.09) 0.01(0.01) 0.04(0.02) (millions of 2,000 dollars) Other states CINC scores/U.S. -1.02(2.07) -0.74(0.38) CINC score Log of capital to capital distance 0.73(0.42)* -0.01(0.21) from U.S. Log of population of states in -0.31(0.15)* 0.01(0.04) civil war or crisis 21-point Polity IV democracy score -0.01(0.02) -0.02(0.01) Cold War 0.85(0.37)* 0.28(0.18) U.S.GNP growth rate -0.08(0.05) -0.00(0.02) U.S.inflation rate -0.03(0.05) -0.02(0.02)◆ Number of crisis actors -0.18(0.31) Number of NATO members involved -0.22(0.18) Constant -0.88(0.29)* -10.46(6.14)* -1.09(0.12)* -1.16(1.82) Observations 150 149 393 390 Pseudo R2 0.11 0.21 0.06 0.09 Note.Asterisks indicate statistical significance at the p<05 level.One civil war and three interational crises had to be omitted from estimation in the second model because the states involved were not system members,creating missing data on many independent variables.The Direct Effects of Political and Economic Interests on Intervention Table 1 presents the results of four models of American intervention in civil wars and international crises. There are good reasons to prefer a parsimonious specifi- cation (e.g., Achen 2002; Clarke 2005; Ray 2003), but it is also possible that esti￾mates from such a minimalist specification reflect related influences on intervention decisions that were omitted from the model. Models 1 and 3 include only the variables of primary theoretical interest here: alliance commit￾ments, rival intervention, and exports. Models 2 and 4 add several additional independent variables reflecting considerations that might create omitted vari￾able bias in the more parsimonious specification. For example, the United States might be less likely to intervene in conflicts involving militarily powerful states, but more likely to intervene in those involving its neighbors, or democratic gov￾ernments. The United States might have acted more aggressively during the Cold War, or on behalf of its NATO allies. Military action might also be more likely in times of economic distress. All of these considerations might also be related to the primary independent variables, especially trade, creating omitted variable bias. In practice, including these additional independent variables has a small effect on inferences about the variables of primary theoretical interest. The coef- ficient estimates for alliance commitments and rival intervention are somewhat larger; trade remains statistically insignificant. The most important empirical finding from this stage of the analysis is that security concerns about alliance commitments and rival intervention had large direct effects on American intervention, while exports were not statistically signif￾icant. The two security concerns had very similar effects on intervention in both civil wars and international crises. Holding the continuous independent variables at their mean value, the predicted probability of civil war intervention in the absence of an alliance or intervention by a rival state was .11. The addition of an alliance would increase this probability to .45. Rival intervention would increase it to .32. For international crises, the baseline probability of intervention was .14. Table 1. Probit Models of Civil War and International Crisis Intervention Civil War Intervention International Crisis Intervention Rival intervention 0.81 (0.30)* 0.94 (0.33)* 0.44 (0.16)* 0.61 (0.17)* Alliance partners involved 1.09 (0.31)* 1.12 (0.44)* 0.32 (0.10)* 0.35 (0.14)* Log of exports in previous year (millions of 2,000 dollars) )0.04 (0.06) 0.11 (0.09) 0.01 (0.01) 0.04 (0.02) Other states CINC scores⁄ U.S. CINC score )1.02 (2.07) )0.74 (0.38) Log of capital to capital distance from U.S. 0.73 (0.42)* )0.01 (0.21) Log of population of states in civil war or crisis )0.31 (0.15)* 0.01 (0.04) 21-point Polity IV democracy score )0.01 (0.02) )0.02 (0.01) Cold War 0.85 (0.37)* 0.28 (0.18) U.S. GNP growth rate )0.08 (0.05) )0.00 (0.02) U.S. inflation rate )0.03 (0.05) )0.02 (0.02)* Number of crisis actors )0.18 (0.31) Number of NATO members involved )0.22 (0.18) Constant )0.88 (0.29)* )10.46 (6.14)* )1.09 (0.12)* )1.16 (1.82) Observations 150 149 393 390 Pseudo R2 0.11 0.21 0.06 0.09 Note. Asterisks indicate statistical significance at the p < .05 level. One civil war and three international crises had to be omitted from estimation in the second model because the states involved were not system members, creating missing data on many independent variables. 746 Power or Plenty?