BENJAMIN O.FORDHAM 751 TABLE 3.Differenced Gravity Models of the Influence of Alliances on Trade Flows Alliance:-Alliance 0.10(0.25) Alliance -0.08(0.05) Alliance formed in preceding 5 years -0.08(0.14) In(GDPust)-In(GDPust) -1.67(1.11) -1.67(1.11) -1.67(1.11) In(GDPOTHER)-In(GDPOTHERI) 0.32(0.41) 0.36(0.41) 0.33(0.41) In(populationus)-In(populationust) 24.69(8.99)* 26.12(9.05)* 25.34(9.23) In(populationoTHER)-In(populationoTHER1) -0.71(1.48) -0.80(1.48) -0.75(1.48) Lagged dependent variable -0.23(0.04)* -0.23(0.04)* -0.23(0.04)* Constant -0.11(0.12) -0.10(0.12) -0.11(0.12) Observations 7.025 7.025 7.025 0.06 0.06 0.06 Note The dependent variable is In(exportsusoTeR)-In(exportsusorER).Exports are in millions of constant 2,000 dollars.The asterisk indicates significant at the <05 level.Panel-corrected standard errors are reported in parentheses. (1993,411-12)contend that an alliance should have a greater influence on trade under bipolarity because the risk of exit by an ally is lower than it would be under multipolarity.If they are correct,then an alliance effect on trading pat- terns should be more likely during this bipolar period than in the multipolar era before World War II. Although the models estimated here constitute a very generous test of the alliances-increase-trade hypothesis,the empirical results do not support it.Nei- ther ongoing alliances nor those signed in recent years have statistically signifi- cant effects on the change in trade.Failure to reject the null hypothesis is not evidence that the null hypothesis is true,so these results should not be inter- preted as demonstrating that alliances do not increase trade.In the context of the analysis presented here,the purpose of these results is to show that endoge- neity is not likely to be a serious problem in the analysis of the impact of trade on alliance formation and termination.The evidence considered here suggests that trade has influenced American alliance decisions rather than the other way around. Assessing the Total Effect of Economic and Security Interests on Intervention The evidence considered thus far indicates that alliances and the behavior of rival states have had large effects on the probability of American intervention in civil wars and international crises.While the volume of American exports has not had comparable direct effects,it has affected the probability of alliance forma- tion,and thus indirectly influenced intervention decisions.The next step in the analysis is to assess the substantive importance of this indirect effect. The probability of intervention can be computed from conditional probability estimates given trade and alliance formation as follows: pr(intervention)=pr(interventionalliance)*pr(alliance) pr(intervention~alliance)*pr(~alliance) The probability of alliance formation in a given year is small,regardless of the value of exports.When exports were one standard deviation below their mean, the probability of alliance formation was roughly .0001.This probability rises to only about .0049 when exports were one standard deviation above their mean. Given these small numbers,the immediate effect of an increase in American exports on the probability of intervention was tiny.An increase in exports from one standard deviation below their mean to one standard deviation above it would increase the probability of intervention in a civil war that broke out the(1993, 411–12) contend that an alliance should have a greater influence on trade under bipolarity because the risk of exit by an ally is lower than it would be under multipolarity. If they are correct, then an alliance effect on trading pat￾terns should be more likely during this bipolar period than in the multipolar era before World War II. Although the models estimated here constitute a very generous test of the alliances-increase-trade hypothesis, the empirical results do not support it. Nei￾ther ongoing alliances nor those signed in recent years have statistically signifi- cant effects on the change in trade. Failure to reject the null hypothesis is not evidence that the null hypothesis is true, so these results should not be inter￾preted as demonstrating that alliances do not increase trade. In the context of the analysis presented here, the purpose of these results is to show that endoge￾neity is not likely to be a serious problem in the analysis of the impact of trade on alliance formation and termination. The evidence considered here suggests that trade has influenced American alliance decisions rather than the other way around. Assessing the Total Effect of Economic and Security Interests on Intervention The evidence considered thus far indicates that alliances and the behavior of rival states have had large effects on the probability of American intervention in civil wars and international crises. While the volume of American exports has not had comparable direct effects, it has affected the probability of alliance forma￾tion, and thus indirectly influenced intervention decisions. The next step in the analysis is to assess the substantive importance of this indirect effect. The probability of intervention can be computed from conditional probability estimates given trade and alliance formation as follows: pr(intervention) ¼ pr(interventionjallianceÞ pr(alliance) þ pr(interventionj alliance) prð allianceÞ The probability of alliance formation in a given year is small, regardless of the value of exports. When exports were one standard deviation below their mean, the probability of alliance formation was roughly .0001. This probability rises to only about .0049 when exports were one standard deviation above their mean. Given these small numbers, the immediate effect of an increase in American exports on the probability of intervention was tiny. An increase in exports from one standard deviation below their mean to one standard deviation above it would increase the probability of intervention in a civil war that broke out the Table 3. Differenced Gravity Models of the Influence of Alliances on Trade Flows Alliancet – Alliancet-1 0.10 (0.25) Alliance )0.08 (0.05) Alliance formed in preceding 5 years )0.08 (0.14) ln(GDPUSt) – ln(GDPUSt-1) )1.67 (1.11) )1.67 (1.11) )1.67 (1.11) ln(GDPOTHERt) – ln(GDPOTHERt-1) 0.32 (0.41) 0.36 (0.41) 0.33 (0.41) ln(populationUSt) – ln(populationUSt-1) 24.69 (8.99)* 26.12 (9.05)* 25.34 (9.23)* ln(populationOTHERt) – ln(populationOTHERt-1) )0.71 (1.48) )0.80 (1.48) )0.75 (1.48) Lagged dependent variable )0.23 (0.04)* )0.23 (0.04)* )0.23 (0.04)* Constant )0.11 (0.12) )0.10 (0.12) )0.11 (0.12) Observations 7,025 7,025 7,025 R2 0.06 0.06 0.06 Note. The dependent variable is ln(exportsUS-OTHERt) – ln(exportsUS-OTHERt-1). Exports are in millions of constant 2,000 dollars. The asterisk indicates significant at the p < .05 level. Panel-corrected standard errors are reported in parentheses. Benjamin O. Fordham 751