A debate exists over whether (and to what degree) the democratic peace is explained by joint democracy or by a lack of motives for conflict between states that happen to be democratic. Gartzke (1998) applies expected utility theory to the democratic peace and shows that an index of states' preference similarity based on United Nations General Assembly roll-call votes (affinity) accounts for much of the lack of militarized interstate disputes (MIDs) between democracies. Oneal and Russett (1997b, 1998, 1999) respond by arguing that UN voting is itself a function of regime type—that democracy ‘causes’affinity. Oneal and Russett seek to demonstrate their thesis by regressing affinity on democracy and other variables from a standard model of the democratic peace. I replicate results reported by Oneal and Russett and then extend the analysis in several ways. I find that the residuals from Oneal and Russett's regression of affinity remain highly significant as a predictor of the absence of MIDs. Further, significance for democracy is shown to be fragile and subject to variable construction, model specification, and the choice of estimation procedure.