CAPITALIST PEACE 175 their governments are preoccupied with existing terri- sus theory of national preferences,nor is one likely to tory.Development brings with it the ability to project be constructed in a reasonable time.Empirical research power,encouraging contests over both policy and re- on conflict must thus choose between measuring inter- sources,while the richest states lose much of their will- ests imperfectly,and not measuring them at all.I have ingness to steal resources associated with territory. chosen the former,while being mindful of the many Early quantitative studies of the democratic peace potential pitfalls involved in this approach.The argu- included GDP/pop(the consensus measure of develop- ment supplied here is consistent with other research ment),but the variable was not found to be significant in arguing that variable state interests are an impor- (Maoz and Russett 1992).I argue that per capita GDP tant indicator of foreign policy behavior(cf.Bueno de has contrasting effects on disputes.Gartzke and Rohner Mesquita 1981;Voeten 2000).If we cannot know the (2006b)examine this argument directly by splitting the myriad causes of preferences,we can at least go some sample of disputes between territorial and nonterrito- way in measuring their manifestation and their effects. rial conflicts,and by looking at initiators and targets. Measuring interests provides a number of empiri- Here,however,I need to adhere to an established re- cal challenges.Preferences are not directly observable, search design.To parse out the contrasting effects of so one must identify conditions that appear to reflect development on war and peace,I include two variables. state preferences.Using data on United Nations Gen- GDPPC (Low)measures the lower of the two monadic eral Assembly voting available for the period covered population weighted gross domestic product statistics by the Oneal and Russett (1999a)data (1946-96),I for a given dyad(Gleditsch 2002).I also examine the construct an AFFINTrY index.Data on "revealed"pref- natural log of GDPPC(Low)to limit multicollinearity erences are an imperfect representation of an actor's among variables.42 A second variable isolates the effect real ranking over outcomes.Still,UN voting arguably of wealth on likely subjects of territorial aggression. distorts preferences less than available alternatives such GDPPC x CONTIG (Low)interacts contiguity and the as alliance portfolios(Gartzke 1998,2000).I examine development variable.It is most likely that a decline in other indicators in the appendix(I also use the residuals the value of conquest will manifest itself in relations of AFFINITY as a proxy,after regressing the interest vari- with neighbors,where territorial claims are most com- able on democracy and other variables,and show that mon and aggression most practical.43 the residuals have similar effects).The Affinity index Interest Similarity:Many students of international re- reports the similarity of dyadic UN voting patterns,us- lations reject as excessively narrow the realist empha- ing the "S"coding(Signorino and Ritter 2001).Values sis on uniform,monolithic interests and argue instead range between one,"most similar,and negative one, that state objectives vary with a complex variety of "least similar."I expect a threshold effect of interests. factors (cf.Moravcsik 1997).Relations between the AFFINITY should be negatively associated with disputes, United States and Israel,and between the United States with the more dissimilar values (closer to-1)being and India have been quite different in the post-World disproportionately likely to fight.44 War II period,even accounting for capabilities,geog- raphy,regime type,and so on.National interests also Additional Variables change over time;elections in Bolivia and Germany re- sulted in two very different leaders,one who is moving Iinclude the same"control"variables as Onealand Russett her country closer to the United States,and one who is (1999a)to facilitate comparison of results.45 moving farther away. Ideally,researchers in international relations would possess a model of state interests that would estimate 4Readers who prefer can ignore the interest argument without altering the effects of development or markets.Russett and Oneal the effects of a number of relevant causal variables. (2001)suggest that UN voting patterns are explained by democracy. The same could be said for democracy,however,and Gartzke (2000)reports that even the residuals from a regression of for measures of national capabilities,economic devel- democracy and other variables on AFFINTTY account for the effect of democracy on conflict in politically relevant dyads.Regressing both opment,alliance ties,and so on.There exists no consen- monadic DEMOCRACY variables on AFFINTTY in the all dyads sample yields an R2 of 0.0658(93%of the variance is left unexplained), 42None of the key variables correlates at above 0.38(Democracy while adding an indicator of liberalization,GDPPC (Low),and [Low]and GDP per capita).Results are available from the author. dummies for NATO and Warsaw Pact membership increase R2to See Vasquez(1993)for a discussion of the close relationship be- only0.0936. tween contiguity and territorial disputes or wars.Data measuring 45There are reasons to be cautious about the arbitrary effects of con- territorial conflict cannot be used in the research design required trol variables (Achen 2005;Clarke 2005;Ray 2003,2005).However, to replicate existing democratic peace research (Tir et al.1998). a study of this type needs to replicate existing canonical models.CAPITALIST PEACE 175 their governments are preoccupied with existing terri￾tory. Development brings with it the ability to project power, encouraging contests over both policy and re￾sources, while the richest states lose much of their will￾ingness to steal resources associated with territory. Early quantitative studies of the democratic peace included GDP/pop (the consensus measure of develop￾ment), but the variable was not found to be significant (Maoz and Russett 1992). I argue that per capita GDP has contrasting effects on disputes. Gartzke and Rohner (2006b) examine this argument directly by splitting the sample of disputes between territorial and nonterrito￾rial conflicts, and by looking at initiators and targets. Here, however, I need to adhere to an established re￾search design. To parse out the contrasting effects of development on war and peace, I include two variables. GDPPC (LOW) measures the lower of the two monadic population weighted gross domestic product statistics foragiven dyad (Gleditsch 2002). I also examine the natural log of GDPPC (LOW) to limit multicollinearity among variables.42 A second variable isolates the effect of wealth on likely subjects of territorial aggression. GDPPC × CONTIG (LOW) interacts contiguity and the development variable. It is most likely that a decline in the value of conquest will manifest itself in relations with neighbors, where territorial claims are most com￾mon and aggression most practical.43 • Interest Similarity: Many students of international re￾lations reject as excessively narrow the realist empha￾sis on uniform, monolithic interests and argue instead that state objectives vary with a complex variety of factors (cf. Moravcsik 1997). Relations between the United States and Israel, and between the United States and India have been quite different in the post–World War II period, even accounting for capabilities, geog￾raphy, regime type, and so on. National interests also change over time; elections in Bolivia and Germany re￾sulted in two very different leaders, one who is moving her country closer to the United States, and one who is moving farther away. Ideally, researchers in international relations would possess a model of state interests that would estimate the effects of a number of relevant causal variables. The same could be said for democracy, however, and for measures of national capabilities, economic devel￾opment, alliance ties, and so on. There exists no consen- 42None of the key variables correlates at above 0.38 (Democracy [Low] and GDP per capita). Results are available from the author. 43See Vasquez (1993) for a discussion of the close relationship be￾tween contiguity and territorial disputes or wars. Data measuring territorial conflict cannot be used in the research design required to replicate existing democratic peace research (Tir et al. 1998). sus theory of national preferences, nor is one likely to be constructed in a reasonable time. Empirical research on conflict must thus choose between measuring inter￾ests imperfectly, and not measuring them at all. I have chosen the former, while being mindful of the many potential pitfalls involved in this approach. The argu￾ment supplied here is consistent with other research in arguing that variable state interests are an impor￾tant indicator of foreign policy behavior (cf. Bueno de Mesquita 1981; Voeten 2000). If we cannot know the myriad causes of preferences, we can at least go some way in measuring their manifestation and their effects. Measuring interests provides a number of empiri￾cal challenges. Preferences are not directly observable, so one must identify conditions that appear to reflect state preferences. Using data on United Nations Gen￾eral Assembly voting available for the period covered by the Oneal and Russett (1999a) data (1946–96), I construct an AFFINITY index. Data on “revealed” pref￾erences are an imperfect representation of an actor’s real ranking over outcomes. Still, UN voting arguably distorts preferences less than available alternatives such as alliance portfolios (Gartzke 1998, 2000). I examine other indicators in the appendix (I also use the residuals of AFFINITY as a proxy, after regressing the interest vari￾able on democracy and other variables, and show that the residuals have similar effects). The Affinity index reports the similarity of dyadic UN voting patterns, us￾ing the “S” coding (Signorino and Ritter 2001). Values range between one, “most similar,” and negative one, “least similar.” I expect a threshold effect of interests. AFFINITY should be negatively associated with disputes, with the more dissimilar values (closer to −1) being disproportionately likely to fight.44 Additional Variables Iinclude the same “control” variables as Oneal and Russett (1999a) to facilitate comparison of results.45 44Readers who prefer can ignore the interest argument without altering the effects of development or markets. Russett and Oneal (2001) suggest that UN voting patterns are explained by democracy. Gartzke (2000) reports that even the residuals from a regression of democracy and other variables on AFFINITY account for the effect of democracy on conflict in politically relevant dyads. Regressing both monadic DEMOCRACY variables on AFFINITY in the all dyads sample yields an R2 of 0.0658 (93% of the variance is left unexplained), while adding an indicator of liberalization, GDPPC (LOW), and dummies for NATO and Warsaw Pact membership increase R2 to only 0.0936. 45There are reasons to be cautious about the arbitrary effects of con￾trol variables (Achen 2005; Clarke 2005; Ray 2003, 2005). However, a study of this type needs to replicate existing canonical models