19
year.
10
I then conducted the same regression as shown in model 2 of table four (pre-
World War I sample; monadic research design; all MIDS in the coding of the dependent
variable). This control variable did not alter the coefficient on democracy in the period
prior to 1914. The democracy term was still positive and statistically significant
(p<0.001). The coefficient on the system variable behaved as previous research suggests
(Hegre and Gleditsch 1997; Mitchell et al 1999; Kadera et al 2003). Similar to the first
phase of the inverted U, the coefficient on this systemic variable was positive and
significant. The initial spread of democracy in the nineteenth century encouraged states
to initiate military conflict.
Despite the addition of this control variable, the coefficient on democracy
remained positive and significant. While systemic pressures were spurring all states to
act more aggressively during the nineteenth century, such systemic effects do not
completely account for all the democratic pressures toward war. Because the coefficient
on the democracy term was positive and significant, distinct internal processes in
democratic states still stimulated conflict during the nineteenth century.
Internal explanations
Evolutionary approaches (e.g. Cederman, 2001) have explicitly criticized much of
the literature for assuming that democracy has acted as a consistent constraint on the
outbreak of war over the past two centuries. Their empirical findings, along with those
presented here tracing the emergence of a dyadic democratic peace to the turn of the
twentieth century (or later), challenge many of the standard explanations in the literature.
For example, the simple constraints based model argues that citizen aversion to paying
10
A state was counted as a democracy if it achieved a score of 6 on the combined democracy and autocracy
score from POLITY 4.
Convention