Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

3. Democracy doesn’t cause peace – statistical models are spurious and don’t assume economic growth

Mousseau, 12

(Michael – Professor IR Koç University, “The Democratic Peace Unraveled: It’s the Economy” International Studies Quarterly, p 1-12)

Model 2 presents new knowledge by adding the control for economic type. To capture the dyadic expectation of peace among contract-intensive nations, the variable Contract- intensive EconomyL (CIEL) indicates the value of impersonal contracts in force per capita of the state with the lower level of CIE in the dyad; a high value of this measure indicates both states have contract-intensive economies. As can be seen, the coefficient for CIEL ()0.80) is negative and highly significant. This corroborates that impersonal economy is a highly robust force for peace. The coefficient for DemocracyL is now at zero. There are no other differences between Models 1 and 2, whose samples are identical, and no prior study corroborating the democratic peace has considered contractintensive economy. Therefore, the standard econometric inference to be drawn from Model 2 is the nontrivial result that all prior reports of democracy as a force for peace are probably spurious, since this result is predicted and fully accounted for by economic norms theory. CIEL and DemocracyL correlate only in the moderate range of 0.47 (Pearson’s r), so the insignificance of democracy is not likely to be a statistical artifact of multicollinearity. This is corroborated by the variance inflation factor for DemocracyL in Model 2 of 1.85, which is well below the usual rule-of-thumb indicator of multicollinearity of 10 or more. Nor should readers assume most democratic dyads have both states with impersonal economies: While almost all nations with contract-intensive economies (as indicated with the binary measure for CIE) are democratic (Polity2 > 6) (Singapore is the only long-term exception), more than half—55%—of all democratic nation-years have contract-poor economies. At the dyadic level in this sample, this translates to 80% of democratic dyads (all dyads where DemocracyBinary6 = 1) that have at least one state with a contract-poor economy. In other words, not only does Model 2 show no evidence of causation from democracy to peace (as reported in Mousseau 2009), but it also illustrates that this absence of democratic peace includes the vast majority—80%—of democratic dyad-years over the sample period. Nor is it likely that the causal arrow is reversed—with democracy being the ultimate cause of contract-intensive economy and peace. This is because correlations among independent variables are not calculated in the results of multivariate regressions: Coefficients show only the effect of each variable after the potential effects of the others are kept constant at their mean levels. If it was democracy that caused both impersonal economy and peace, then there would be some variance in DemocracyL remaining, after its partial correlation with CIEL is excluded, that links it directly with peace. The positive direction of the coefficient for DemocracyL informs us that no such direct effect exists (Blalock 1979:473–474). Model 3 tests for the effect of DemocracyL if a control is added for mixed-polity dyads, as suggested by Russett (2010:201). As discussed above, to avoid problems of mathematical endogeneity, I adopt the solution used by Mousseau, Orsun and Ungerer (2013) and measure regime difference as proposed by Werner (2000), drawing on the subcomponents of the Polity2 regime measure. As can be seen, the coefficient for Political Distance (1.00) is positive and significant, corroborating that regime mixed dyads do indeed have more militarized conflict than others. Yet, the inclusion of this term has no effect on the results that concern us here: CIEL ()0.85) is now even more robust, and the coefficient for DemocracyL (0.03) is above zero.7 Model 4 replaces the continuous democracy measure with the standard binary one (Polity2 > 6), as suggested by Russett (2010:201), citing Bayer and Bernhard (2010). As can be observed, the coefficient for CIEL ()0.83) remains negative and highly significant, while DemocracyBinary6 (0.63) is in the positive (wrong) direction. As discussed above, analyses of fatal dispute onsets with the far stricter binary measure for democracy (Polity = 10), put forward by Dafoe (2011) in response to Mousseau (2009), yields perfect prediction (as does the prior binary measure Both States CIE), causing quasi-complete separation and inconclusive results. Therefore, Model 5 reports the results with DemocracyBinary10 in analyses of all militarized conflicts, not just fatal ones. As can be seen, the coefficient for DemocracyBinary10 ()0.41), while negative, is not significant. Model 6 reports the results in analyses of fatal disputes with DemocracyL squared (after adding 10), which implies that the likelihood of conflict decreases more quickly toward the high values of DemocracyL. As can be seen, the coefficient for DemocracyL 2 is at zero, further corroborating that even very high levels of democracy do not appear to cause peace in analyses of fatal disputes, once consideration is given to contractintensive economy. Models 3, 4, and 6, which include Political Distance, were repeated (but unreported to save space) with analyses of all militarized interstate disputes, with the democracy coefficients close to zero in every case. Therefore, the conclusions reached by Mousseau (2009) are corroborated even with the most stringent measures of democracy, consideration of institutional distance, and across all specifications: The democratic peace appears spurious, with contract-intensive economy being the more likely explanation for both democracy and the democratic peace.

Sign in to recut, update, or save this card.