174 ERIK GARTZKE (Oneal and Russett 1999c;Russett and Oneal 2001),or Markets:Democratic peace research examines trade on extending the temporal domain (Oneal and Russett interdependence (Oneal et al.1996,2003;Oneal and 1999b;Oneal,Russett,and Berbaum 2003). Russett 1997,1999a,1999b,1999c;Russett and Oneal I estimated coefficients using logit and GEE in Stata 2001).Capital and monetary integration may be more (v.8)with Huber/White robust standard errors,though relevant to conflict than trade (Gartzke and Li 2003; for brevity only the logit estimates are reported.Results Gartzke,Li,and Boehmer 2001).Liberalization cre- using GEE are generally equivalent or more favorable to ates valuable linkages and institutional constraints on the hypotheses.Independent variables are lagged one year a state's ability to intervene in market processes.Be- behind the dependent variable to control for endogeneity. cause states may be tempted to interfere with market The Beck,Katz,and Tucker(1998)method of temporal responses to interstate crises,both robust markets and spline variables was adopted to control for duration de- laissez-faire policies matter. pendence.37 The International Monetary Fund (IMF)pro- vides several indicators of market size,robustness,and Dependent Variable liberalization.The IMF publication Annual Reports on Exchange Arrangements and Exchange Restrictions Zeev Maoz's construction of dyadic militarized interstate (AREAER)lists a series of variables measuring eco- disputes(DYMID)is used as the dependent variable,with nomic openness.I use an index evaluated in previous the standard dichotomous coding of"1"for the initial studies that takes the difference between eight and the year of a MID in the dyad and "0"otherwise(Gochman sum of eight types of government restrictions on for- and Maoz 1984;Jones,Bremer,and Singer 1996).38 The eign exchange,current,and capital accounts(Gartzke Maoz data are intentionally formatted in dyads.Maoz also and Li 2003;Gartzke,Li,and Boehmer 2001).IMF FIN. corrects for coding errors in the MID 2.1 dataset.39 OPEN.(Low)reports the lower monadic score in the dyad.40 High values of IMF FIN.OPEN.(Low)are thus expected to reduce the likelihood of militarized dis- Key Independent Variables putes.The IMF only reports data on member countries, Democracy:Researchers differ over how to measure systematically reducing variance and biasing against democracy,both monadically and dyadically.I rely on statistical significance. three different datasets and three variable construc- I also include indicators of trade to assess whether tions to represent dyadic democracy.The standard in trade influences militarized disputes independent of democratic peace research is the Gurr Polity IV data capital liberalization.I use both the trade data pro- (Jaggers and Gurr 1995).I first prepared monadic val- vided by Oneal and Russett,and data from Gleditsch ues by combining Polity democracy(DEMOC)and au- (2002).41 Again,I follow the Oneal and Russett op- tocracy (AUTOC)scales as follows:[(DEMOC;-AUTOC;) erationalization.Monadic values are first constructed 10]/2,(where i [A,B]).The variable differs mod- using a ratio of bilateral trade over GDP to measure the estly from Oneal and Russett in that I add 10 so that importance of trade relative to a state's total economy. all values are nonnegative and divide by 2 to yield TRADE DEP.(Low)denotes the lower trade dependence the 0-10 range of Polity variables.DEMOCRACY (Low) statistic in the dyad(Bliss and Russett 1998;Oneal and and DEMOCRACY(HIGH),respectively,report the lower Russett 1997,1999a,1999b).Trade interdependence is and higher of democracy values in the dyad.DEMOC- expected to modestly decrease MID propensity. RACY Ax B is the product of monadic values.BoTH Development:Economic development leads to a secular DEMOC.(>7)equals one ("1")if each dyad member decline in the valuation of conquerable resources while has a monadic score of at least seven and zero ("0") intellectual and financial capital critical to productiv- otherwise. ity in modern economies must be enticed rather than coerced.Conversely,wealth and the technology effect 37A Stata"do"file is available from the author replicating all aspects allow for greater power projection.Poor countries sel- of data construction and analysis. dom fight abroad because they cannot,and because 38MID coding rules produce a selection bias(Smith 1998).For example,a state that threatens and then uses force is only coded as using force.Whether threats become uses also depends on whether 40 Bilateral data on financial openness is not available for alarge sam- a target acquiesces or resists. ple of countries.Signaling should occur monadically,though two integrated economies probably increase the effect.Unlike democ- 3DYMID is available at http://spirit.tau.ac.il/poli/faculty/maoz/ dyadmid.html.In subsequent research,I will extend analysis to the racy,economic freedom has a monadic effect (Gartzke 2005a). new MID 3 data(once other variables are coded or made available), 4Oneal,Russett,and Berbaum(2003)advocate use of the Gleditsch and the ICB data. (2002)data since these data contain fewer missing observations.174 ERIK GARTZKE (Oneal and Russett 1999c; Russett and Oneal 2001), or on extending the temporal domain (Oneal and Russett 1999b; Oneal, Russett, and Berbaum 2003). I estimated coefficients using logit and GEE in Stata (v. 8) with Huber/White robust standard errors, though for brevity only the logit estimates are reported. Results using GEE are generally equivalent or more favorable to the hypotheses. Independent variables are lagged one year behind the dependent variable to control for endogeneity. The Beck, Katz, and Tucker (1998) method of temporal spline variables was adopted to control for duration de￾pendence.37 Dependent Variable Zeev Maoz’s construction of dyadic militarized interstate disputes (DYMID) is used as the dependent variable, with the standard dichotomous coding of “1” for the initial year of a MID in the dyad and “0” otherwise (Gochman and Maoz 1984; Jones, Bremer, and Singer 1996).38 The Maoz data are intentionally formatted in dyads. Maoz also corrects for coding errors in the MID 2.1 dataset.39 Key Independent Variables • Democracy: Researchers differ over how to measure democracy, both monadically and dyadically. I rely on three different datasets and three variable construc￾tions to represent dyadic democracy. The standard in democratic peace research is the Gurr Polity IV data (Jaggers and Gurr 1995). I first prepared monadic val￾ues by combining Polity democracy (DEMOC) and au￾tocracy (AUTOC) scales as follows: [(DEMOCi – AUTOCi) + 10]/2, (where i ∈ [A,B]). The variable differs mod￾estly from Oneal and Russett in that I add 10 so that all values are nonnegative and divide by 2 to yield the 0–10 range of Polity variables. DEMOCRACY (LOW) and DEMOCRACY (HIGH), respectively, report the lower and higher of democracy values in the dyad. DEMOC￾RACY A × B is the product of monadic values. BOTH DEMOC. (≥7) equals one (“1”) if each dyad member has a monadic score of at least seven and zero (“0”) otherwise. 37A Stata “do” file is available from the author replicating all aspects of data construction and analysis. 38MID coding rules produce a selection bias (Smith 1998). For example, a state that threatens and then uses force is only coded as using force. Whether threats become uses also depends on whether a target acquiesces or resists. 39DYMID is available at http://spirit.tau.ac.il/poli/faculty/maoz/ dyadmid.html. In subsequent research, I will extend analysis to the new MID 3 data (once other variables are coded or made available), and the ICB data. • Markets: Democratic peace research examines trade interdependence (Oneal et al. 1996, 2003; Oneal and Russett 1997, 1999a, 1999b, 1999c; Russett and Oneal 2001). Capital and monetary integration may be more relevant to conflict than trade (Gartzke and Li 2003; Gartzke, Li, and Boehmer 2001). Liberalization cre￾ates valuable linkages and institutional constraints on a state’s ability to intervene in market processes. Be￾cause states may be tempted to interfere with market responses to interstate crises, both robust markets and laissez-faire policies matter. The International Monetary Fund (IMF) pro￾vides several indicators of market size, robustness, and liberalization. The IMF publication Annual Reports on Exchange Arrangements and Exchange Restrictions (AREAER) lists a series of variables measuring eco￾nomic openness. I use an index evaluated in previous studies that takes the difference between eight and the sum of eight types of government restrictions on for￾eign exchange, current, and capital accounts (Gartzke and Li 2003; Gartzke, Li, and Boehmer 2001). IMF FIN. OPEN. (LOW) reports the lower monadic score in the dyad.40 High values of IMF FIN. OPEN. (LOW) are thus expected to reduce the likelihood of militarized dis￾putes. The IMF only reports data on member countries, systematically reducing variance and biasing against statistical significance. I also include indicators of trade to assess whether trade influences militarized disputes independent of capital liberalization. I use both the trade data pro￾vided by Oneal and Russett, and data from Gleditsch (2002).41 Again, I follow the Oneal and Russett op￾erationalization. Monadic values are first constructed using a ratio of bilateral trade over GDP to measure the importance of trade relative to a state’s total economy. TRADE DEP. (LOW) denotes the lower trade dependence statistic in the dyad (Bliss and Russett 1998; Oneal and Russett 1997, 1999a, 1999b). Trade interdependence is expected to modestly decrease MID propensity. • Development: Economic development leads to a secular decline in the valuation of conquerable resources while intellectual and financial capital critical to productiv￾ity in modern economies must be enticed rather than coerced. Conversely, wealth and the technology effect allow for greater power projection. Poor countries sel￾dom fight abroad because they cannot, and because 40Bilateral data on financial openness is not available for a large sam￾ple of countries. Signaling should occur monadically, though two integrated economies probably increase the effect. Unlike democ￾racy, economic freedom has a monadic effect (Gartzke 2005a). 41Oneal, Russett, and Berbaum (2003) advocate use of the Gleditsch (2002) data since these data contain fewer missing observations